PDA

View Full Version : How competitive do you want it to be?



Pages : [1] 2 3

ursogr8
06-01-2004, 08:48 AM
How competitive do you want it to be?

The old ACF Bulletin Board has previously debated the merits and demerits of 'junk' rounds in SWISS tournaments.
Junk rounds have the drawback of discouraging stronger players from entering all-in SWISS tournaments.
Strong players are more likely to be attracted to those SWISS tournaments where modification is made to the structure such as holding a Championship and Reserves. In this way the effect of junk games is reduced by separating the field into two distinct tournaments. However, the separation of the B-quartile from the top players creates its own resentment in the underclass. You only have to read the vigorous debate in firegoats “Championship qualifier criteria” to see that B-quartile players like to have the chance to play against A-quartile players.
So this is the dilemma for the tournament designer: how to create a pairings arrangement that pairs players of comparable ratings without going to the extreme of shunting the B-quartile and C-quartile players into a separate tournament.

Box Hill Chess Club has been experimenting for a couple of years with a permanent acceleration (of 2 bonus points) for the A division players in a large all-in SWISS.
These SWISSes have mainly been 7-evening events at (about) 90 minute time controls for 80-100 players.
The all-in field is split into 2 divisions. Prizes are available in each division; but the bottom division players cannot win a top division prize is set a tournament condition.

There are two benefits of this system
1) Junk games are reduced dramatically.
2) Prizes for C-quartile winners (that is the top players in the bottom division) are much fairer as these players are scoring 5 points out of 7 instead of 3 out of 7 in a normal SWISS. Fluke results and cinchy pairings are reduced in their effect on the prizes.

While debate on the merits of these various tournament formats has been spirited, we lacked a measure of the relative competitiveness of various tournament formats. The design of events is a major influence on whether players are attracted to the event. (The other major factor is obviously the rewards by way of prizes and titles).
After further BB debate we decided that the metric that serves the purpose of measuring competitiveness is the calculation of the mean of absolute differences of ratings for all pairings. This is calculated on a round by round basis.

Who knows...one day we might be able to advertise a tournament as
“Guaranteed for this tournament the mean of absolute differences of ratings for all pairings is less than 200” and this could attract strong players. Implicitly this is why Victorian INTERCLUB was always so attractive at the A grade level. It had a very low mean of absolute differences of ratings for all pairings.

In the next three posts I will give details of the metric for some key tournaments run in NSW and VIC in 2003.

starter

ursogr8
06-01-2004, 08:51 AM
I have calculated a competitive metric, in our recent Box Hill OPEN event, for each round.
It is a measure of whether the player was presented with a consistently competitive tournament,
or did he get only 2 or 3 reasonable games.

What would you have considered a satisfactory mean absolute deviation for you personally if you had entered the tournament?

My data is based on a 9 round tournament; that will influence your estimate of what is a satisfactory metric level for you.

Here is the data

List of average (absolute differences) for each of 9 rounds

Round 1 222.0
Round 2 218.9
Round 3 208.6
Round 4 197.1
Round 5 195.5
Round 6 204.0
Round 7 191.3
Round 8 196.1
Round 9 177.7
Average for tournament 201.2

Note >> this data was formed by calculating the absolute differences of ratings for all pairings for a given round. Then averaging the absolute differences in the round.
The final average for the tournament is just the average of the 9 figures.

starter

ursogr8
06-01-2004, 08:53 AM
Barry provided some data as below

I quickly did the same thing on the 2003 Doeberl Premier Division. This had a rating range of 1560 - 2544 and 72 players. Mean rating differences were:

Round 1 - 435
Round 2 - 323
Round 3 - 235
Round 4 - 249
Round 5 - 261
Round 6 - 214
Round 7 - 172

Overall - 270

Interesting thing is it took till the end of the tournament before the mean difference got below the Box Hill "junk rounds" level. Interestingly the last round was comparable. However, this is purely anecdotal.

Here comes the data for the Doeberl Major. 54 rated players in the range 1169-1975.

Round 1 - 222
Round 2 - 206
Round 3 - 181
Round 4 - 170
Round 5 - 157
Round 6 - 156
Round 7 - 148

Overall - 177

And the Doeberl Minor. Rating range 295 - 1598, 76 rated players.

Round 1 - 545
Round 2 - 371
Round 3 - 298
Round 4 - 346
Round 5 - 336
Round 6 - 385
Round 7 - 355

Overall - 377

The wide variability on the Minor might be an indication of underrated juniors having an impact on the event. Leading to a lot of seemingly rating based mismatches but actually quite a competitive tournament.

ursogr8
06-01-2004, 08:58 AM
Fairfield (NSW) data has now been analysed and presented here for the first time. (Note, I excluded all games involving un-rated players).
Round 1 - 734
Round 2 - 593
Round 3 - 383
Round 4 - 386
Round 5 - 374
Round 6 - 326

Fairfield had Ian Rogers in the field and his rating is at the extreme of Aus. ratings. It naturally has an upwards effect on the metric.

starter

PHAT
06-01-2004, 10:06 AM
Good work, starter. Now, 5 data sets, what do I see?

1. If I take a guess at the % of juniors in each field, I see a positive correlation between the proportion of juniors and the mean rating difference (MRD) in non-junk rounds (NJR).

2. Where the field is a ratings based subset of the playing population the MRD is generallylower.

3. The larger the field the larger the MRD in the junk rounds (JR).


Where to now? If we want to pursue a lower MRD - and I am not convinced of this - we should:

1. Keep juniors out of opens (or make sure people are more accuritely rated)

2. Run parallel swiss events based on grades, like the Doeberl and/or the Aust Open festival.

3. Keep fields as small as posible - eg 8 player round robins determined by rating. However, with this protocol, some players would never meet OTB.

ursogr8
06-01-2004, 12:11 PM
Where to now? If we want to pursue a lower MRD - and I am not convinced of this - we should:

1. Keep juniors out of opens (or make sure people are more accuritely rated)

2. Run parallel swiss events based on grades, like the Doeberl and/or the Aust Open festival.

3. Keep fields as small as posible - eg 8 player round robins determined by rating. However, with this protocol, some players would never meet OTB.


I am recommending
4. Give the A-quartile and B-quartile players a 2-point bonus in SWISSPerfect, and leave the bonus in place for all rounds of the SWISS. All players (A, B, C, D-quartiles) play in this SWISS, and C, D players cannot win A, B prizes as a condition of entry.

I am not in favour of your 1. And probably you are not either.
Your 2.gets the underclasses unhappy (e.g fg7 argument)
Your 3 is essentially our Interclub (your grades) competitions. And very well liked thank you. It has its place in the panoply.

PHAT
06-01-2004, 12:31 PM
I am recommending
4. Give the A-quartile and B-quartile players a 2-point bonus in SWISSPerfect, and leave the bonus in place for all rounds of the SWISS.



Was there evr the following concidered:
A grade +1.5
B +1.0
C +0.5

ursogr8
06-01-2004, 12:43 PM
I am recommending
4. Give the A-quartile and B-quartile players a 2-point bonus in SWISSPerfect, and leave the bonus in place for all rounds of the SWISS.



Was there evr the following concidered:
A grade +1.5
B +1.0
C +0.5

Matt

Yes. We did this in the recent Christmas Friday SWISS; almost exactly how you describe. We had 8 grades in fact.
I had resisted getting into this detail while I was concentrating on establishing the point that 'competitiveness' is a key attraction for stronger players. And I want you to eventually yield on this. You need to forsake supporting junk games.

starter

ursogr8
08-01-2004, 12:05 PM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 195.1


A comparison with earlier posts shows this is nearly as competitive as you can get in a tourney. And are the players attracted to it? Or is it the title and prize money.

starter

Bill Gletsos
08-01-2004, 12:15 PM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 195.1
Is that based on their ACF or FIDE rating.
The reason I ask is that in the round by round results the rating shown is FIDE.

ursogr8
08-01-2004, 12:32 PM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 195.1
Is that based on their ACF or FIDE rating.
The reason I ask is that in the round by round results the rating shown is FIDE.

Bill

Sounds like the FIDE then, because I just went to the round by round results URL.
Happy to do the ACF calculation. Are they consolidated in any handy spot?

starter

Bill Gletsos
08-01-2004, 12:36 PM
Yes, just go to the round by round results on the web page and scroll down to standings.

You will then see the players FIDE and ACF ratings listed.

ursogr8
08-01-2004, 12:59 PM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 195.1 for FIDE ratings and 257.1 for ACF ratings.

A comparison with earlier posts shows this is nearly as competitive as you can get in the early rounds of a tourney. And are the players attracted to it? Or is it the title and prize money.

starter

........................

Kevin Bonham
08-01-2004, 05:03 PM
A comparison with earlier posts shows this is nearly as competitive as you can get in a tourney. And are the players attracted to it? Or is it the title and prize money.

Are you using the top division field only or the Reserves as well? Whichever is the case, the competitiveness is obviously caused by the field being divided into two sections, making large rating differentials uncommon. Also ratings at the top end are probably more accurate than those lower down - an 1100-rated player who has only played a dozen rated games may conceivably be 700 or 1500 strength in some cases, but a 2200 would virtually never be 1800 or 2600 strength.

ursogr8
08-01-2004, 05:12 PM
A comparison with earlier posts shows this is nearly as competitive as you can get in a tourney. And are the players attracted to it? Or is it the title and prize money.

Are you using the top division field only or the Reserves as well? Whichever is the case, the competitiveness is obviously caused by the field being divided into two sections, making large rating differentials uncommon. Also ratings at the top end are probably more accurate than those lower down - an 1100-rated player who has only played a dozen rated games may conceivably be 700 or 1500 strength in some cases, but a 2200 would virtually never be 1800 or 2600 strength.
Kevin
Top division only, because that is one tournament.
All previous data posted has been for one tournament.
Sometimes there is more than one tourney at the venue; like Doeberl ...had three

starter

ursogr8
09-01-2004, 10:17 AM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 195.1 for FIDE ratings and 257.1 for ACF ratings.

A comparison with earlier posts shows this is nearly as competitive as you can get in the early rounds of a tourney. And are the players attracted to it? Or is it the title and prize money.

starter
Round 2 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of 137.2 for FIDE ratings and 177.1 for ACF ratings.

ursogr8
09-01-2004, 11:37 AM
Are you using the top division field only or the Reserves as well? Whichever is the case, the competitiveness is obviously caused by the field being divided into two sections, making large rating differentials uncommon. Also ratings at the top end are probably more accurate than those lower down - an 1100-rated player who has only played a dozen rated games may conceivably be 700 or 1500 strength in some cases, but a 2200 would virtually never be 1800 or 2600 strength.

Kevin

Round 2 of the Reserves has a mean absolute deviation of rating differences at 179.4. (This discounts a couple of pairings with unrateds and byes.)
This compares very favourably for competitiveness with any events examined in 2003.

starter

ursogr8
15-01-2004, 11:27 AM
Round 1 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>195.1 for FIDE ratings and 257.1 for ACF ratings.

A comparison with earlier posts shows this is nearly as competitive as you can get in the early rounds of a tourney. And are the players attracted to it? Or is it the title and prize money.

starter
Round 2 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>137.2 for FIDE ratings and 177.1 for ACF ratings.

Round 3 of the current Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>101.6 for FIDE ratings

starter

ursogr8
20-01-2004, 10:01 AM
Round 1 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>195.1 for FIDE ratings and 257.1 for ACF ratings.

A comparison with earlier posts shows this is nearly as competitive as you can get in the early rounds of a tourney. And are the players attracted to it? Or is it the title and prize money.

starter
Round 2 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>137.2 for FIDE ratings and 177.1 for ACF ratings.

Round 3 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>101.6 for FIDE ratings

starter

Round 4 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>140.6 for FIDE ratings

ursogr8
23-01-2004, 10:18 AM
Round 1 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>195.1 for FIDE ratings and 257.1 for ACF ratings.

A comparison with earlier posts shows this is nearly as competitive as you can get in the early rounds of a tourney. And are the players attracted to it? Or is it the title and prize money.

starter
Round 2 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>137.2 for FIDE ratings and 177.1 for ACF ratings.

Round 3 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>101.6 for FIDE ratings

starter

Round 4 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>140.6 for FIDE ratings
Round 5 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>138.3 for FIDE ratings

starter

Bill Gletsos
23-01-2004, 10:46 AM
Since the figures you previously posted for other tournaments were based on ACF ratings I would have thought for comparison purposes it would be more worthwhile if you had continued with ACF ratings here instead of FIDE.

ursogr8
23-01-2004, 11:05 AM
Since the figures you previously posted for other tournaments were based on ACF ratings I would have thought for comparison purposes it would be more worthwhile if you had continued with ACF ratings here instead of FIDE.

Bill
Whew. Someone else reads this thread. For a moment there I thought I was posting to myself. Thank you. Thank you. :D

Yes, after your previous post here I went back and did the calculation in ACF ratings, and these are the large font figures in blue-bold earlier in the thread. I did it for two rounds and then I judged that I had bench-marked the relativity, (like that turn of phrase?).
I have not calculated ACF figures beyond round 2 because it was labour intensive given that the source of the data available to me is on unichess.org; and on that site the pairings only have FIDE ratings that I can easily paste into the spreadsheet for calculation.

starter

PS. I will check the bench-marked relativity, in a round or so.

Bill Gletsos
23-01-2004, 11:21 AM
I have not calculated ACF figures beyond round 2 because it was labour intensive given that the source of the data available to me is on unichess.org; and on that site the pairings only have FIDE ratings that I can easily paste into the spreadsheet for calculation.
That is easily resolved. ;)
Send me your email address and I'll email you the SP files. :D

ursogr8
23-01-2004, 11:28 AM
I have not calculated ACF figures beyond round 2 because it was labour intensive given that the source of the data available to me is on unichess.org; and on that site the pairings only have FIDE ratings that I can easily paste into the spreadsheet for calculation.
That is easily resolved. ;)
Send me your email address and I'll email you the SP files. :D


Bill
I have sent a PM 12.28
starter

ursogr8
27-01-2004, 10:31 AM
Round 5 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>138.3 for FIDE ratings

starter

Round 6 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>120.0 for FIDE ratings .

starter

PS (Bill, I have received the SP files you sent and will produce RESERVES and ACF metrics).

skip to my lou
27-01-2004, 10:36 AM
Im editing those font sizes.

arosar
27-01-2004, 10:50 AM
Im editing those font sizes.

What font sizes? And why? I thought we all agreed that you won't interfere in this section of the board?

AR

arosar
27-01-2004, 10:51 AM
test quick reply

arosar
27-01-2004, 10:53 AM
Whoa! Now that was silly. Why do I have to click a message to post a quick reply - and the post I clicked wasn't even directly quoted. Seems an unnecessary action to me.

AR

ursogr8
27-01-2004, 10:54 AM
Im editing those font sizes.

Jeo

OK by me.
I had a go at reducing them myself, intuitive approach only, (I don't read instructions for software), and it didn't do what I hoped for.
So, thanks. :cool:

starter

skip to my lou
27-01-2004, 11:16 AM
Whoa! Now that was silly. Why do I have to click a message to post a quick reply - and the post I clicked wasn't even directly quoted. Seems an unnecessary action to me.

AR
When you press that quick reply button, hit quote message in reply option and it will quote that message.

It is required incase you press enter accidentally.

skip to my lou
28-01-2004, 07:41 PM
Quoted from Admin Panel:


Require Click for Quick Reply
If you set this option to 'no', users will not have to click the Quick Reply icon in order to type a message in the quick reply box.

However, this will be at the expense of the usefulness of the Threaded and Hybrid view modes, as the system will not know to which post a user is replying, so the threaded view will become non-sensical.

Set this option to 'no' at your own discretion.

So thats another reason.

ursogr8
30-01-2004, 08:32 AM
Round 6 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>120.0 for FIDE ratings .

starter


Round 7 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>108 for FIDE ratings .This is the most competitive round yet according to our metric.

starter

ursogr8
03-02-2004, 12:44 PM
Round 7 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>108 for FIDE ratings .This is the most competitive round yet according to our metric.

starter
Round 8 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>78 for FIDE ratings .This is the most competitive round yet according to our metric, and the first time I have seen the metric under 100 for any tournament analysed. :clap:

A summary of the metric to date is
Round 1 Metric 195
Round 2 Metric 137
Round 3 Metric 101
Round 4 Metric 141
Round 5 Metric 138
Round 6 Metric 120
Round 7 Metric 108
Round 8 Metric 78 :hmm:

Bill Gletsos
03-02-2004, 12:51 PM
I'd be more interested in seeing the ACF metrics rather than the FIDE ones.
Also using the ACF ones would allow comparison to other tournament metrics you previously had posted.

ursogr8
03-02-2004, 01:57 PM
I'd be more interested in seeing the ACF metrics rather than the FIDE ones.
Also using the ACF ones would allow comparison to other tournament metrics you previously had posted.

Yeh, OK Bill

I know that.
And I did appreciate the SP files you sent me.
But, mate, I am deprived...I don't have SP. So I have to wait until I visit the Club, where our computer with SP is loaded, then I have to export to text files, and then I have come back here where I have EXCEL to do what you want me to do.

starter

Bill Gletsos
03-02-2004, 02:04 PM
Nice try but no cigar. :doh:
You could download and use an evaluation copy of SP from the SP web site.
It is valid for 30 days. :whistle:

ursogr8
03-02-2004, 03:15 PM
Round 1 of the current BH AUTUMN CUP has a calculated mean absolute deviation of rating differences, over the 37 boards, of >>>>>>> 431 .

starter

ursogr8
06-02-2004, 08:32 AM
Round 9 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>116 for FIDE ratings .

A summary of the metric to date is
Round 1 Metric 195
Round 2 Metric 137
Round 3 Metric 101
Round 4 Metric 141
Round 5 Metric 138
Round 6 Metric 120
Round 7 Metric 108
Round 8 Metric 78
Round 9 Metric 116

^^^^^^^^^^^

ursogr8
09-02-2004, 02:25 PM
Round 10 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>87 for FIDE ratings .

A summary of the metric to date is
Round 1 Metric 195
Round 2 Metric 137
Round 3 Metric 101
Round 4 Metric 141
Round 5 Metric 138
Round 6 Metric 120
Round 7 Metric 108
Round 8 Metric 78
Round 9 Metric 116
Round 10 Metric 87

ursogr8
10-02-2004, 11:25 AM
Round 11 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>123 for FIDE ratings .

A summary of the metric to date is
Round 1 Metric 195
Round 2 Metric 137
Round 3 Metric 101
Round 4 Metric 141
Round 5 Metric 138
Round 6 Metric 120
Round 7 Metric 108
Round 8 Metric 78
Round 9 Metric 116
Round 10 Metric 87
Round 11 Metric 123

ursogr8
10-02-2004, 11:32 AM
Round 1 of the current BH AUTUMN CUP has a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431 .
Round 2 of the current BH AUTUMN CUP has a calculated mean absolute deviation of rating differences, over the 35 boards, of >>>>>>> 259 . Unrated-opponents games excluded.

starter

Bill Gletsos
10-02-2004, 11:37 AM
Round 11 of the recent Australian Championship has a calculated mean absolute deviation of rating differences, over the 15 boards, of >>>>>>>123 for FIDE ratings .

A summary of the metric to date is
Round 1 Metric 195
Round 2 Metric 137
Round 3 Metric 101
Round 4 Metric 141
Round 5 Metric 138
Round 6 Metric 120
Round 7 Metric 108
Round 8 Metric 78
Round 9 Metric 116
Round 10 Metric 87
Round 11 Metric 123
Well now that you have finally finished wasting you time with the FIDE calcs perhaps you can do the more useful ACF calcs. ;)

ursogr8
10-02-2004, 11:53 AM
Well now that you have finally finished wasting you time with the FIDE calcs perhaps you can do the more useful ACF calcs. ;)
Bill

We have a lot to learn from you.
You know, don't just be a 'starter', finish the opponent off. (How are you BTW ChessLover; groggy, lost track, yielding)?

So, I started on the FIDE ratings, and I have finished with the FIDE ratings. I will leave it to others to judge if it was a waste of time.
Post #18 (editted from 33) did benchmark the relativity, of sorts; for those who were impatient for the metric on ACF ratings.

starter

Bill Gletsos
10-02-2004, 12:08 PM
Bill

We have a lot to learn from you.
You know, don't just be a 'starter', finish the opponent off. (How are you BTW ChessLover; groggy, lost track, yielding)?

So, I started on the FIDE ratings, and I have finished with the FIDE ratings. I will leave it to others to judge if it was a waste of time.
Post #33 did benchmark the relativity, of sorts; for those who were impatient for the metric on ACF ratings.
What has post #33 got to do with ACF ratings.
The last post where you mentioned an ACF metric appears to be post #18 when you gave the round 2 ACF figure.

ursogr8
10-02-2004, 12:22 PM
What has post #33 got to do with ACF ratings.
The last post where you mentioned an ACF metric appears to be post #18 when you gave the round 2 ACF figure.

Just a little test I planted? :oops:

ursogr8
16-02-2004, 10:05 AM
Rounds 1 to 4 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220 ,308 .
Unrated-opponents games excluded.

starter

ursogr8
26-02-2004, 07:38 AM
Rounds 1 to 4 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220 ,308 .
Unrated-opponents games excluded.

starter

Rounds 1 to 5 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220, 308, 247.
Unrated-opponents games excluded.

After 5 rounds, of the specially invented Box Hill SWISS, we are suddenly confronted with a game where there is a 1000 point difference in the ratings. Is this a flaw in the invention?
Probably not since the overall mean absolute deviation is still looking very good (i.e. competitive pairings).
How do the local tournaments you run at your Club stack up for competitiveness? :hmm:
Details of the Box Hill pairings at
http://www.boxhillchess.org.au/e2004/e0401fac/round5.htm

starter

ursogr8
12-03-2004, 07:15 AM
Rounds 1 to 6 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220, 308, 247, 261.
Unrated-opponents games excluded.

After 6 rounds, of the specially invented Box Hill SWISS, we still have seven games with a rating difference in excess of 450 points. How do the local tournaments you run at your Club stack up for competitiveness? :hmm:

Details of the Box Hill pairings at
http://www.boxhillchess.org.au/e2004/e0401fac/round6.htm

starter

Garvinator
12-03-2004, 08:42 AM
Rounds 1 to 6 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220, 308, 247, 261.
Unrated-opponents games excluded.

After 6 rounds, of the specially invented Box Hill SWISS, we still have seven games with a rating difference in excess of 450 points. How do the local tournaments you run at your Club stack up for competitiveness? :hmm:

Details of the Box Hill pairings at
http://www.boxhillchess.org.au/e2004/e0401fac/round6.htm

starter

ill make an attempt at in a couple of weeks when the city of brisbane is over.

ursogr8
14-03-2004, 12:58 PM
Rounds 1 to 7 of the current BH AUTUMN CUP have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>> 431, 259, 220, 308, 247, 261, 255.
Unrated-opponents games excluded.

After 7 rounds, of the Box Hill SWISS, permanently accelerated A and B divisions, we still have four games with a rating difference in excess of 450 points. How do the local tournaments you run at your Club stack up for competitiveness? :hmm:

Details of the Box Hill pairings at
http://www.boxhillchess.org.au/e2004/e0401fac/round7.htm

starter

ursogr8
21-03-2004, 08:16 AM
Rounds 1 to 7 of the 2004 Begonia/Ballarat week-ender have a calculated mean absolute deviation of rating differences, of >>>>>>>>>>>>>>>>>>>>>>>>>
764, 506, 419, 348, 344, 273, 302

I will post the same figures on the Tournament >Ballarat/Begonia thread and debate (on Ballarat) should continue there.

starter




Unrated-opponents games excluded.

How do the local tournaments you run at your Club stack up for competitiveness?

ursogr8
24-03-2004, 07:13 AM
I'd be more interested in seeing the ACF metrics rather than the FIDE ones.
Also using the ACF ones would allow comparison to other tournament metrics you previously had posted.

Bill,
You will have noticed that I posted the results of Ballarat/Begonia, analysed to calculate the competitveness metric. :clap:
No comment by you on those. :hmm:

The real reason for my post > um, err :oops:
I lost those files you sent me on the Aus Reserves, ACF ratings and pairings.
Could you send again? :oops:

starter

Garvinator
26-03-2004, 04:06 PM
You will be pleased about these results starter :hmm: i think.

City of Brisbane Championships just finished.

Results can be found here: http://www.caq.org.au/htm/cobct.htm

As for the competitive index, here we go:

Round 1: 420
Round 2: 338
Round 3: 256
Round 4: 242
Round 5: 277
Round 6: 272
Round 7: 174

unrated players not included. 10 upset wins out of 76 games.

ursogr8
26-03-2004, 04:25 PM
You will be pleased about these results starter :hmm: i think.

City of Brisbane Championships just finished.

Results can be found here: http://www.caq.org.au/htm/cobct.htm

As for the competitive index, here we go:

Round 1: 420
Round 2: 338
Round 3: 256
Round 4: 242
Round 5: 277
Round 6: 272
Round 7: 174

unrated players not included. 10 upset wins out of 76 games.

Nice set of figures gg''.
Even round 1 escapes the criteria of junk round, viz >450.

BTW, bit odd about rounds 5 and 6. Any theory?

starter

Garvinator
26-03-2004, 04:29 PM
Nice set of figures gg''.
Even round 1 escapes the criteria of junk round, viz >450.

BTW, bit odd about rounds 5 and 6. Any theory?

starter
how come you keep changing your thoughts about what the competitive index should be.

You have said this in the past:

Bill, have you calculated the metric? Is this InterLeagues Rapid, (a non accelearated swiss), an event that has a competitive index of less than 300 for all rounds?

Garvinator
26-03-2004, 05:14 PM
by me:


As for the competitive index, here we go:

Round 1: 420
Round 2: 338
Round 3: 256
Round 4: 242
Round 5: 277
Round 6: 272
Round 7: 174

unrated players not included. 10 upset wins out of 76 games.


Nice set of figures gg''.
Even round 1 escapes the criteria of junk round, viz >450.

BTW, bit odd about rounds 5 and 6. Any theory?

starter

the 'upsets' came equally in each round. One of two upsets in each round. I used a criteria for upsets of roughly: a person who wins the game despite being rated more than 200 points less than their opponent.

I actually thought rounds 1 and 2 could be classed as junk rounds. As for rounds 5 and 6, not sure. I do know that their were only 6 players rated above 1600 and about 10 less than 1250. This might have something to do with it.
Over performing low rated meeting underperforming high raters in round 5 and 6 might account for it. not sure really :hmm: what i did notice was not many games went for the entire 4 hours or went past 60 moves. the timer was 40 moves in 90 minutes and then 30 mins to finish.

Bill Gletsos
26-03-2004, 05:34 PM
I would think that its not only important to publish the mean rating difference per round but also the standard deviation.

Garvinator
26-03-2004, 05:38 PM
I would think that its not only important to publish the mean rating difference per round but also the standard deviation.
probably is but i have no clue how to publish the standard deviation, so thought it was better that i didnt, than to post incorrect data.

Alan Shore
26-03-2004, 07:45 PM
Personally, I don't ever think rounds 1 and 2 are a waste.. I've beaten an IM in round 1 before, so don't say it can never be done! (I also drew with someone 700 pts below me, so don't say it can't happen in both directions!)

Garvinator
26-03-2004, 08:08 PM
Personally, I don't ever think rounds 1 and 2 are a waste.. I've beaten an IM in round 1 before, so don't say it can never be done! (I also drew with someone 700 pts below me, so don't say it can't happen in both directions!)
yes, that might be true, but we are talking about what happens 99% of the time :whistle:

Alan Shore
26-03-2004, 08:13 PM
Maybe so, but I enjoy the balance between given the opportunity of playing (and beating) the top players and given the opportunity to win prizes.

Garvinator
26-03-2004, 08:19 PM
Maybe so, but I enjoy the balance between given the opportunity of playing (and beating) the top players and given the opportunity to win prizes.
in my opinion, it is more about playing against players that you have a realistic chance of having a decent game against. I also have the opinion that getting constant beatings as a new player could be one reason why our sport is not expanding as much as it should.

Alan Shore
26-03-2004, 08:31 PM
in my opinion, it is more about playing against players that you have a realistic chance of having a decent game against. I also have the opinion that getting constant beatings as a new player could be one reason why our sport is not expanding as much as it should.

I have to disagree - there are sufficient 'restricted tournaments' for those who want a genuine chance of coming first in an event, in australian chess. Tell me, how are players supposed to improve their chess by eeking noobs all the time rather than playing stronger players and learning? Both on the net and OTB, I've improved my ratings from playing stronger players and not eeking patzers.

Garvinator
26-03-2004, 08:34 PM
I have to disagree - there are sufficient 'restricted tournaments' for those who want a genuine chance of coming first in an event, in australian chess. Tell me, how are players supposed to improve their chess by eeking noobs all the time rather than playing stronger players and learning? Both on the net and OTB, I've improved my ratings from playing stronger players and not eeking patzers.
and when we refer to junk rounds, we are talking about rounds where each game had a average differential of 400 points. this means that the result is a statistical certainty.

Alan Shore
26-03-2004, 08:38 PM
and when we refer to junk rounds, we are talking about rounds where each game had a average differential of 400 points. this means that the result is a statistical certainty.

Even so, I enjoy the discrepancies.. Of course I'm always in the top half of the draw these days, so it's nice to have an easy first round when one is still kinda sleepy, hehe. On the flipside, I've been seeded in the bottom-half of the draw and come outright third in an open tournament before.. ANYONE is potentially beatable, and on occasions, are.

Garvinator
26-03-2004, 08:39 PM
Even so, I enjoy the discrepancies.. Of course I'm always in the top half of the draw these days, so it's nice to have an easy first round when one is still kinda sleepy, hehe. On the flipside, I've been seeded in the bottom-half of the draw and come outright third in an open tournament before.. ANYONE is potentially beatable, and on occasions, are.
i think ill pass the buck to starter here, he was the first one to notice this effect that im talking about.

Kevin Bonham
26-03-2004, 09:08 PM
and when we refer to junk rounds, we are talking about rounds where each game had a average differential of 400 points. this means that the result is a statistical certainty.

No it doesn't, a 400-point difference is an 8% average for the bottom half of the draw, which in practice (with draws) may mean that say 12% of the top half do not win. For a first round, I think that's quite acceptable. Normally I quite like having the first round to warm up in, although if I lose in the first round I find it difficult to work back into contention knowing that I must win every game.

It's first rounds where the average differences are something like 800 points (where each game has say a 1% chance of an upset) that I think are silly and should be dealt with by acceleration.

ursogr8
27-03-2004, 10:39 PM
No it doesn't, a 400-point difference is an 8% average for the bottom half of the draw, which in practice (with draws) may mean that say 12% of the top half do not win. For a first round, I think that's quite acceptable. Normally I quite like having the first round to warm up in, although if I lose in the first round I find it difficult to work back into contention knowing that I must win every game.

It's first rounds where the average differences are something like 800 points (where each game has say a 1% chance of an upset) that I think are silly and should be dealt with by acceleration.

Kevin

450-points is the figure that I proposed as the definition of a junk round, not 400.
What would be the corresponding % for this defintion?

Just a point, could you and Bruce please stop rabbiting on about how you quite enjoy an easy first round. We all suspected you enjoy winning, but that is not what this thread was about. You are just confusing the objective of the thread which was to explore what is the tolerance level of a low rated player who will choose not enter a tournament if the oppponents his paired with have too high a rating difference. Now, you have declared it self-evident at 800. I am declaring it at 450 based on conversations I have had with experienced juniors who have stopped going to Ballarat/ Begionia. And Bruce has declared nothing; so, I am presuming he thinks it does not harm the entries into tournaments whatever the rating differences are.

starter

Garvinator
27-03-2004, 10:53 PM
starter, i have wondered how much self interest is involved when ppl say, i like getting an easy game in the first round. I bet they havent considered that their opponent isnt getting an easy game at all.

Alan Shore
27-03-2004, 11:08 PM
I really don't care - as I said before, I've been in the bottom half of the draw before and welcomed the chance to play strong players - get over it. It works both ways and I believe the current swiss system is the best available. You can whinge all you like but I won't be supporting you at all and I'm sure Kevin won't either.

Alan Shore
27-03-2004, 11:11 PM
And, I just might add for starter's benefit, there are u1850, u1750 and u1450 tourns around if you are really scared of those strong players.

Bill Gletsos
27-03-2004, 11:12 PM
Kevin

450-points is the figure that I proposed as the definition of a junk round, not 400.
What would be the corresponding % for this defintion?
Actually Kevin appears to be using the normal scale. The logistic scale is better, especially with larger differences.
Irrespective of all that on the normal scale a 400 difference is 7.6%, 450 is 5.5%, 500 is 3.9% and 600 is 1.7%..
On the logistic scale, 400 is 9.1% and 450 is 7%, 500 is 5.3% and 600 is 3.1%.

Kevin Bonham
27-03-2004, 11:37 PM
Just a point, could you and Bruce please stop rabbiting on about how you quite enjoy an easy first round.

Sorry.


You are just confusing the objective of the thread which was to explore what is the tolerance level of a low rated player who will choose not enter a tournament if the oppponents his paired with have too high a rating difference. Now, you have declared it self-evident at 800.

No I haven't, I've declared that 800 point differences are one point where I personally think the first round is stupid and will use acceleration instead as an organiser. 800 need not be the threshold either, but it is a figure at which I certainly feel comfortable making that decision. I make no comment about the tolerance levels of low rated players - which will vary enormously, as has been obvious from many comments here. There are those who hate playing players even 300 pts above them, and those who would pay to play players 1500 pts up.

We had quite a long debate at our Club about whether to split our club champs (traditionally a round robin with c.15 entries rated from c.600 to c.2000) into two divisions because of the length of the event and the cost of rating all the junk games. (The motion to split ended up being tied so a casting vote retained the status quo, and karma inflicted a field of 21 on us as punishment, but I digress). One argument made there, by one of the stronger players, was that even if weaker players do not like playing strong players, doing so is the main way they will significantly improve. I think that's a relevant point, but I'm very willing to be convinced that significant numbers of weaker players really quit as opposed to complain (or move to another club) about losing junk rounds.


I am declaring it at 450 based on conversations I have had with experienced juniors who have stoppedgoing to Ballarat/ Begionia.

Is this where we start to haggle? :owned:

Rincewind
27-03-2004, 11:40 PM
Actually Kevin appears to be using the normal scale. The logistic scale is better, especially with larger differences.

You've picked me up on this in the past too. :oops:

Actually Glickman says in "A Comprehensive Guide to Chess Ratings" that stats Professor Hal Stern in a 1992 article titled "Are all linear paired comparison models empirically equivalent" showed "it makes virtually no difference whether one assumes the logistic distribution or the normal distribution for differences in players' strengths". The reason Glickman prefers the logisitic (Bradley-Terry model) is that "it tends to be more tractable to work with".

Caveat: I haven't read Stern's piece. If anyone has the journal Mathematical Social Sciences, 23, 103-117 and could send me a copy of this article please PM me.

Bill Gletsos
28-03-2004, 12:33 AM
You've picked me up on this in the past too. :oops:
Oh yeah so I did. :doh:


Actually Glickman says in "A Comprehensive Guide to Chess Ratings" that stats Professor Hal Stern in a 1992 article titled "Are all linear paired comparison models empirically equivalent" showed "it makes virtually no difference whether one assumes the logistic distribution or the normal distribution for differences in players' strengths". The reason Glickman prefers the logisitic (Bradley-Terry model) is that "it tends to be more tractable to work with".
Although that appears to be true, I thought I read somewhere where the logistic actually gave a closer fit to observed results than the normal. Unfortunately I cannot remember where I read it. :hmm:

Rincewind
28-03-2004, 12:56 AM
Apparently (again from Glickman's article) the Bradley-Terry model assumes the distribution of an individual's playing strength (ie values in the players' boxes) follows an "extreme value distribution". This assumes players rarely play substatially below their average ability. The Thurstone-Mosteller model on the other hand assume the normal distribution, meaning it assume players are equally likely to play above or below their average playing strength.

Perhaps it is felt the Bradley-Terry model more closely reflects reality, especially when the rating difference is approaching the extremes, but I guess Stern may argue otherwise. ;)

ursogr8
28-03-2004, 09:19 AM
I make no comment about the tolerance levels of low rated players - which will vary enormously, as has been obvious from many comments here. There are those who hate playing players even 300 pts above them, and those who would pay to play players 1500 pts up.


Kevin, I agree there is a spread of tolerance over all individuals. No argument on that. But somewhere along the line a chess tournament has to be designed and a decision made as to the competitiveness of pairings. Seems to be that you are now backing down to a modified Bruce D. position >> just play a single SWISS and tell the players to get over it if they don't like junk games; except that you will fine-tune if the the figure is 800. BTW what is the % for 800?



We had quite a long debate at our Club about whether to split our club champs (traditionally a round robin with c.15 entries rated from c.600 to c.2000) into two divisions because of the length of the event and the cost of rating all the junk games. (The motion to split ended up being tied so a casting vote retained the status quo, and karma inflicted a field of 21 on us as punishment, but I digress).


Nice to see the debate is being held.
But what about the $8 (or so) per junk game that the player forks out. Your Committee may have saved themselves a rating fee, but what about the players $8 entry fee/round...say Ballarat for example?



One argument made there, by one of the stronger players, was that even if weaker players do not like playing strong players, doing so is the main way they will significantly improve.



This irrational rationale is like the old justification for LATIN, as a curriclum subject >> it will be good for you son.


I think that's a relevant point, but I'm very willing to be convinced that significant numbers of weaker players really quit as opposed to complain (or move to another club) about losing junk rounds.

Is this where we start to haggle? :owned:

Not haggle Kevin.
This is where we post our personal views on how competitive we want it to be. What is your figure (somewhere between 450 and 800 it sounds like).

regards,and thanks for your post
starter

ursogr8
28-03-2004, 09:27 AM
You can whinge all you like but I won't be supporting you at all and I'm sure Kevin won't either.

Bruce

Sorry you feel my posts on this thread are regarded by you as whinges.
I have a genuine objective to increase player participation by improving the design of tournament pairings. You obviously don't agree.
Do you need assistance to learn how to not read this thread so that you avoid reading the whinges?
Or, if you actually like the thread but don't like my posts, do you need assistance on how to set a filter so that you don't get to see my posts?

starter

ursogr8
28-03-2004, 09:29 AM
And, I just might add for starter's benefit, there are u1850, u1750 and u1450 tourns around if you are really scared of those strong players.

hi Bruce
You have called me a whinger.
And now scared.

Do you have any debating points coming eventually?

starter

Alan Shore
28-03-2004, 10:52 AM
hi Bruce
You have called me a whinger.
And now scared.

Do you have any debating points coming eventually?

starter

Go back and read what I've said, they are my points, you seemed to have just ignored them and instead singled out throwaway remarks I added for emphasis only and not directed at you at all. If I believe in the current system, the onus is on you to suggest a better one. So far, it's only been your subjective opinion the games are too easy/hard. I for one enjoy the range of ratings in a weekender and the opportunity to play and beat titled players. Therefore I'm unimpressed with the subjectivity of your argument - it's still a good thing you raised it since it gives others a chance to think about it but we'll have to agree to disagree.

ursogr8
28-03-2004, 11:22 AM
and not directed at you at all.

Bruce
The words used were “you are really scared’ and “you can whinge all you like”. Seems to be directed at me.



If I believe in the current system, the onus is on you to suggest a better one.



And I have done so at length.



Therefore I'm unimpressed with the subjectivity of your argument - it's still a good thing you raised it since it gives others a chance to think about it but we'll have to agree to disagree.


On the BB various posters have laboriously debated what is an appropriate METRIC since October of last year. I have published the calculated metric from various tournaments and invited comment from the BB as to what each reader thinks is appropriate competitiveness as measured by the METRIC. My personal view is 450 is the thresh-hold.
I have evidence that previous visitors to tournaments don’t return because the travel costs don’t stack against the waste of time on the first two junk rounds.

Labelling my posts as subjective is off the mark given I am trying to make the debate objective.

starter

Alan Shore
28-03-2004, 11:34 AM
My personal view is 450 is the thresh-hold.
starter

Why arbitrarily this amount? What's wrong with Kevin's '800'? What do your metric calculations prove?


I have evidence that previous visitors to tournaments don’t return because the travel costs don’t stack against the waste of time on the first two junk rounds.

Where is this evidedence? Who says?


P.S. I'm going home now and will take me a long time to travel so I'm unable to reply back any time soon if you respond.

Kevin Bonham
28-03-2004, 12:40 PM
Seems to be that you are now backing down to a modified Bruce D. position >> just play a single SWISS and tell the players to get over it if they don't like junk games; except that you will fine-tune if the the figure is 800. BTW what is the % for 800?

Less than 1 on the normal scale, perhaps Bill can say what it is on the other one.

I would certainly want to use acceleration to avoid a round with mean 800 point mismatches. I would consider using acceleration to avoid a round with smaller mismatches, but that would depend on the number of rounds and the number of players. i.e. will the first round get much in the way of the most efficient sorting of the field?


Nice to see the debate is being held.
But what about the $8 (or so) per junk game that the player forks out. Your Committee may have saved themselves a rating fee, but what about the players $8 entry fee/round...say Ballarat for example?

Yes in weekenders this is something to be considered (not so much of an issue in our club champs where the cost is effectively 25c per player per junk game - plus the time commitment to play it, which IMHO is more important.)


This irrational rationale is like the old justification for LATIN, as a curriclum subject >> it will be good for you son.

Nice try but your analogy fails. Latin will be good for you if you are bad at Latin and want to be better at it, whether you think it's all too hard or not.


This is where we post our personal views on how competitive we want it to be. What is your figure (somewhere between 450 and 800 it sounds like).

See above.

ursogr8
28-03-2004, 01:00 PM
I would certainly want to use acceleration to avoid a round with mean 800 point mismatches. I would consider using acceleration to avoid a round with smaller mismatches, but that would depend on the number of rounds and the number of players. i.e. will the first round get much in the way of the most efficient sorting of the field?


hi Kevin
Should I read into this, that in the hierarchy of values, you are putting ‘efficient sorting’ ahead of ‘enjoyable pairings’?
My view is that if the first round is junk, do something else so that the majority of players are not browned-off by the unbalanced pairings. The ‘sorting’ actually gets better the more competitive games/rounds are played.




Yes in weekenders this is something to be considered (not so much of an issue in our club champs where the cost is effectively 25c per player per junk game - plus the time commitment to play it, which IMHO is more important.)

Good point. Time to travel to Ballarat, or Drouin, or wherever, and cost per round; are both reasons for us to design tournaments that maximise the enjoyment.



Nice try but your analogy fails. Latin will be good for you if you are bad at Latin and want to be better at it, whether you think it's all too hard or not.


Not going to debate you on this point. It was just an indulgence of mine to trot out an old curriculum argument.


starter

Bill Gletsos
28-03-2004, 01:04 PM
Less than 1 on the normal scale, perhaps Bill can say what it is on the other one.
For normal it is 0.23% for the logistic its 0.99%.

Bill Gletsos
28-03-2004, 01:07 PM
Of course the question could come down to is a mean difference of 400 with a SD of 150 better than a mean diff of 500 with a SD of 50. :hmm:

ursogr8
28-03-2004, 01:11 PM
Why arbitrarily this amount?

Bruce
You have already declared you don’t think any fine-tuning of a SWISS is justified to produce more enjoyable pairings. So, why are you asking this question?


What's wrong with Kevin's '800'?


Nothing is wrong with his figure. It is just his personal criteria, based on experience, where it becomes pointless to play the junk round. BTW why didn’t you ask him where he got 800 from; like you challenged me where I got 450 from?


What do your metric calculations prove?


Bruce, from your position where you think a normal SWISS cannot be improved then I have no hope of demonstrating that other factors should come into the parings process for a tournament. I am never going to be able to ‘prove’ things to you because you have declared a position where you would not ‘tinker’ with the normal SWISS.






Where is this evidence? Who says?



Why do you want the evidence Bruce, you have already made up your mind?


starter

ursogr8
28-03-2004, 01:20 PM
Of course the question could come down to is a mean difference of 400 with a SD of 150 better than a mean diff of 500 with a SD of 50. :hmm:

Bill

Thanks. A sensible observation, from a mathematicians point of view.
But not practical from a marketing point of view.
Compare
>>> Come play at Balberll where we have designed for you a pairings regime that has at worst a M.A.D of pairing differences of 450,
with
>>>> Come play at Balberll where we have designed for you a pairings regime that has at worst a SD of M.A.D of pairing differences of 50.

Which is more evocative?
Which would you like to explain to the usual punter?

(And don't give me a hard time that they would understand neither from my phrasing :shhh: ).

starter

Bill Gletsos
28-03-2004, 01:39 PM
Bill

Thanks. A sensible observation, from a mathematicians point of view.
But not practical from a marketing point of view.
Compare
>>> Come play at Balberll where we have designed for you a pairings regime that has at worst a M.A.D of pairing differences of 450,
with
>>>> Come play at Balberll where we have designed for you a pairings regime that has at worst a SD of M.A.D of pairing differences of 50.

Which is more evocative?
Which would you like to explain to the usual punter?
I would argue thats not the point.
Whats the use of saying you have a M.A.D of 450 when perhaps it can be proved that a M.A.D of 500 is actually better.

ursogr8
28-03-2004, 01:49 PM
Of course the question could come down to is a mean difference of 400 with a SD of 150 better than a mean diff of 500 with a SD of 50. :hmm:

Bill
OK, I go back to this post of yours and I say I agree (for the second time), it is a good thought of yours.
Now what do you actually do with the observation. You seem to have ruled out use as a marketing phrase.

Just how are you going to get the pairings with a 500:50 profile instead of 450:150?

starter

PHAT
28-03-2004, 01:51 PM
Some months ago I posted a bit on starters push for accelerated pairings for the first one or two rounds as a means to have more games more closely matched. I would like to reiterate my personal position on it and then pose a question.

I like the junk round(s) - usually have 3 before I am in my place. I enjoy scaring the sh.t out of 1800's and sometimes embarrasing one.

The question that must be answered is: Do the majority of players want to retain the junk rounds or have all rounds more closely matched? Until you/I/we answer that question, we should stop bickering.

ursogr8
28-03-2004, 02:01 PM
Some months ago I posted a bit on starters push for accelerated pairings for the first one or two rounds as a means to have more games more closely matched. I would like to reiterate my personal position on it and then pose a question.

I like the junk round(s) - usually have 3 before I am in my place. I enjoy scaring the sh.t out of 1800's and sometimes embarrasing one.

The question that must be answered is: Do the majority of players want to retain the junk rounds or have all rounds more closely matched? Until you/I/we answer that question, we should stop bickering.

Matt
Yes, yes, yes.
That is the central question (your bolded italics words).

Thanks for re-phrasing.

And the follow-on question >> how do we find out or estimate what the majority want.


BTW.........which posters are bickering? Me with BD? I can defend that.

starter

Garvinator
28-03-2004, 02:35 PM
As for the competitive index, here we go:

Round 1: 420
Round 2: 338
Round 3: 256
Round 4: 242
Round 5: 277
Round 6: 272
Round 7: 174

unrated players not included. 10 upset wins out of 76 games.

I posted these figures a couple of days ago and have followed the debate since. Starter, i should have said with the figures above, that the highest rated player was 1965 and there were only 6 players above 1600.

ursogr8
28-03-2004, 02:47 PM
I posted these figures a couple of days ago and have followed the debate since. Starter, i should have said with the figures above, that the highest rated player was 1965 and there were only 6 players above 1600.

hi 'ra'g'y'

I responded to your post...see my #55. A few questions for you there.

And another > in which rounds did the upsets occur?

starter

ursogr8
28-03-2004, 02:54 PM
I posted these figures a couple of days ago and have followed the debate since. Starter, i should have said with the figures above, that the highest rated player was 1965 and there were only 6 players above 1600.

And with such a low highest-rated player there is a natural lowering of the M.A.D. of pairing differences. I noticed the same effect (but in the other direction) when Geoff Saw entered the recent Box Hill Autumn Cup. In each round, Geoff's pairing difference was substantial, and had quite an effect on the average. Perhaps Bill or Baz can suggest a mathematical fine-tuning of the METRIC?

starter

Garvinator
28-03-2004, 03:14 PM
the 'upsets' came equally in each round. One of two upsets in each round. I used a criteria for upsets of roughly: a person who wins the game despite being rated more than 200 points less than their opponent.

I actually thought rounds 1 and 2 could be classed as junk rounds. As for rounds 5 and 6, not sure. I do know that their were only 6 players rated above 1600 and about 10 less than 1250. This might have something to do with it.
Over performing low rated meeting underperforming high raters in round 5 and 6 might account for it. not sure really :hmm: what i did notice was not many games went for the entire 4 hours or went past 60 moves. the timer was 40 moves in 90 minutes and then 30 mins to finish. i posted this reply previously.

Alan Shore
28-03-2004, 03:16 PM
Some months ago I posted a bit on starters push for accelerated pairings for the first one or two rounds as a means to have more games more closely matched. I would like to reiterate my personal position on it and then pose a question.

I like the junk round(s) - usually have 3 before I am in my place. I enjoy scaring the sh.t out of 1800's and sometimes embarrasing one.

The question that must be answered is: Do the majority of players want to retain the junk rounds or have all rounds more closely matched? Until you/I/we answer that question, we should stop bickering.

I want to see what other people say too Matthew - I've given my opinion, I too enjoy the "junk round".



Why do you want the evidence Bruce, you have already made up your mind?

Stop stalling and give me the evidence - If it's good enough I may change my mind, I am always willing to listen, if the argument is good! Up to this point, you've given me only codswallop. And you can stop the veiled snideness please.

ursogr8
28-03-2004, 03:32 PM
I want to see what other people say too Matthew - I've given my opinion, I too enjoy the "junk round".




Stop stalling and give me the evidence - If it's good enough I may change my mind, I am always willing to listen, if the argument is good! Up to this point, you've given me only codswallop. And you can stop the veiled snideness please.

hi Bruce

So far you have called me
>a whinger
>> scared
>>> who writes codswallop
>>>> and now, snide.

Please put me on your ignore list so that my further posts are of no concern to you.

tks
starter

ursogr8
28-03-2004, 03:38 PM
Could those high-ranked players, say >1800 for example, list their last 5 games against opponents who have a rating 800 points higher than themselves. I would like to see evidence of how long it took to play a reasonable number of opponents of higher rating in a 'junk' round.

tks
starter

Alan Shore
28-03-2004, 03:47 PM
I wish you would not be so sensitive starter - I'd hate to think what it would do to you if you ever got Bill offside..

And no, I'm not putting anyone on ignore.

For the last time, pretty please, post your evidence and I promise not to call you those ghastly :eek: things again.

Bill Gletsos
28-03-2004, 04:52 PM
hi Bruce

So far you have called me
>a whinger
>> scared
>>> who writes codswallop
>>>> and now, snide.

Actually starter I find all those terms rather mild and not worthy of worrying about.
Its the geese, morons, dipsticks and clowns that lead to real entertainment. :lol:

Kevin Bonham
28-03-2004, 07:07 PM
hi Kevin
Should I read into this, that in the hierarchy of values, you are putting ‘efficient sorting’ ahead of ‘enjoyable pairings’?

Well, I know that pairings with too great a rating difference are inefficient from a sorting perspective, unless your event has a large number of rounds for the number of players, that's mathematical reality.

But on the other, I don't know whether "uncompetitive" rounds annoy more weaker players than the absence of them would annoy others.

(Also, a side-comment - in many games with a rating difference of c.500 points, the lower rated player is still quite competitive, even though they will almost always end up losing.)


The ‘sorting’ actually gets better the more competitive games/rounds are played.

Some simulations I did (albeit with fairly small sample sizes) suggested that acceleration improves the sorting at the top - and hence finding the most deserving winner - but makes rather little difference to the sorting in the middle. This is because after 3 rounds of an Accelerated Swiss, most of the field are where they would have been anyway.

I suspect your double-acceleration system also has benefits for your "competitiveness" but not a lot of benefits for sorting.

Kevin Bonham
28-03-2004, 07:14 PM
Could those high-ranked players, say >1800 for example, list their last 5 games against opponents who have a rating 800 points higher than themselves. I would like to see evidence of how long it took to play a reasonable number of opponents of higher rating in a 'junk' round.

Do you mean historically, including when our ratings were much lower?

Even so I can't give you 800, but I can say the last time I played a player over 500 points above myself at the time in a normal-length game was in 1995. I won.

ursogr8
28-03-2004, 08:16 PM
I wish you would not be so sensitive starter - I'd hate to think what it would do to you if you ever got Bill offside..

And no, I'm not putting anyone on ignore.

For the last time, pretty please, post your evidence and I promise not to call you those ghastly :eek: things again.

Bruce,

First, I have been off-side with Bill a few times. Most times I am wrong.
Second, he has never bothered with name-calling, even for emphasis.

I am not adding sensitive to the list.

starter

ursogr8
28-03-2004, 08:32 PM
Well, I know that pairings with too great a rating difference are inefficient from a sorting perspective, unless your event has a large number of rounds for the number of players, that's mathematical reality.

But on the other, I don't know whether "uncompetitive" rounds annoy more weaker players than the absence of them would annoy others.

Kevin
That is a big call to speculate that players would be annoyed by the 'lack of uncompetitive rounds'.




(Also, a side-comment - in many games with a rating difference of c.500 points, the lower rated player is still quite competitive, even though they will almost always end up losing.)


True, there are competitive games between individuals who are 500 rating points apart.
But when you look around the hall and see 60 boards playing, and the average rating split wherever you look is 500 (row after row, table after table, room after room, and you see that most games are finished in less than half-time, and you see there are 2 upsets in the first 50 boards; you have to imagine that it could be better managed/designed.




I suspect your double-acceleration system also has benefits for your "competitiveness" but not a lot of benefits for sorting.

Somewhere along this track I will be able to persuade you to look at the points received by a C quartile player ( more like 6.5 out of 8 or 9, instead of 3.5 out of 8), then you will see the wisdom of my claim that the C-quartile prizes are more fairly earned rather than a fluke pairing in the last round.

starter

ursogr8
28-03-2004, 08:41 PM
Do you mean historically, including when our ratings were much lower?

Even so I can't give you 800, but I can say the last time I played a player over 500 points above myself at the time in a normal-length game was in 1995. I won.

Kevin
Thanks for the response.
You say that you have not faced a 800 rated difference for 8 years, where you have been the lower rated player.

I have 3 juniors who got 2 games out of 7 in Ballarat. Of course it is easy to overlook that a 500 rated player versus a 1300 rated player is a ratings difference of 800. But anyhow, I can assure you they get browned off long before the difference is 800 points. The cost involved is i) travel to the Country, ii) accomodation for junior and parent, and iii) pro-rata entry fee of about $5 per round.
Not picking on you, but you really have not experienced the downside for 8 years.

I wait with interest for a few others to respond with their 'experience'. Maybe even BD.


starter

ursogr8
07-04-2004, 08:49 PM
I was discussing the Box Hill policy of permanent acceleration of the top half of a SWISS last night with Arthur Goudy, a stalwart of the Club. Arthur is the typical C quartile player who has been around for years but always thinks improvement is just around the corner, or in the next book bought from Marcus Raine.

Arthur spoke highly of the system as it gave him many competitive games. He particularly like the avoidance of the any junk round pairings early in the tournament. He said a third quartile player entering into the non-accelerated SWISSes at other Clubs frequently had to wait until they were 0/5 before getting on the scoreboard.

starter

Garvinator
07-04-2004, 08:59 PM
question for you starter, can someone from the D quartile win the tournament?

Kevin Bonham
08-04-2004, 02:18 AM
Kevin
That is a big call to speculate that players would be annoyed by the 'lack of uncompetitive rounds'.

No it isn't, we've had strong players here say that they like an easy start and weak players say that they like to play the really strong players.


True, there are competitive games between individuals who are 500 rating points apart.
But when you look around the hall and see 60 boards playing, and the average rating split wherever you look is 500 (row after row, table after table, room after room, and you see that most games are finished in less than half-time, and you see there are 2 upsets in the first 50 boards; you have to imagine that it could be better managed/designed.

There will be a bit more than 2 on average. More like 3 or 4, including draws - for what it's worth. Better managed - from what specific perspective?


Somewhere along this track I will be able to persuade you to look at the points received by a C quartile player ( more like 6.5 out of 8 or 9, instead of 3.5 out of 8), then you will see the wisdom of my claim that the C-quartile prizes are more fairly earned rather than a fluke pairing in the last round.

I suspect that who draws the top-quartile players in the last round and who doesn't will be a big part of which C-quartile player gets the prize anyway, prizes below the very top of the tree in a Swiss always have a degree of dodginess about them. I really don't know how you stop the "leapfrog" problem except having lots and lots of rounds.


He said a third quartile player entering into the non-accelerated SWISSes at other Clubs frequently had to wait until they were 0/5 before getting on the scoreboard.

Only if they're going really badly. In the normal scheme of things a third-quartile player loses round 1 and wins round 2. The risk of getting an upset victim in round 2 is countered by the chance of scoring an upset in round 1. Even if a 3rd-quartile player goes loss, loss, their third round opponent is a fourth quartile player. If they still lose that they don't deserve an easy game. :lol:


You say that you have not faced a 800 rated difference for 8 years, where you have been the lower rated player.

Actually I said it was that long since I had a difference over 500. It would be at least 16 years since I had an 800-point difference.


I have 3 juniors who got 2 games out of 7 in Ballarat. Of course it is easy to overlook that a 500 rated player versus a 1300 rated player is a ratings difference of 800. But anyhow, I can assure you they get browned off long before the difference is 800 points. The cost involved is i) travel to the Country, ii) accomodation for junior and parent, and iii) pro-rata entry fee of about $5 per round.
Not picking on you, but you really have not experienced the downside for 8 years

Hang on, I did say that I considered 800 point differences to be too much, didn't I? And that, depending on circumstances, I might consider drawing the mark lower for particular tournaments? But you've got work to do if you want to haggle me down to 400.

Is there a crosstable for that Ballarat event up anywhere readily accessible? Might have some comments on how they came to get two 800+ - point mismatches, it sounds unusual.

Alan Shore
08-04-2004, 02:23 AM
No it isn't, we've had strong players here say that they like an easy start and weak players say that they like to play the really strong players.

Count me in as one of those players, and a number of others that I've spoken to... I've been in both categories and enjoy the "mismatches".

ursogr8
08-04-2004, 07:57 AM
question for you starter, can someone from the D quartile win the tournament?

Good morning grayggray

It is a condition of entry that C and D quartile players cannot win the B and A Division prizes. Therefore, the answer is no; D quartile players cannot win the tournament. But I add that in 20 years of running normal SWISSes, no D quartile player ever got more than 3.5/9.

starter

ursogr8
08-04-2004, 08:10 AM
Count me in as one of those players, and a number of others that I've spoken to... I've been in both categories and enjoy the "mismatches".

hi Bruce

How about re-reading post #100 on this thread and giving us some details.


starter

ursogr8
08-04-2004, 08:33 AM
I suspect that who draws the top-quartile players in the last round and who doesn't will be a big part of which C-quartile player gets the prize anyway, prizes below the very top of the tree in a Swiss always have a degree of dodginess about them.


I agree with your point that the opponent paired against a C-quartile player will have an element of luck attached to it. Yes some C's will be paired against a B and some might even get a well-performed D.
The extra point I am making is that (under the BHCC SWISS) the C's, who get to the end of round 8 with a chance of winning the C prize, will be sitting on 5.5/8 rather than 2.5/8, and will have done more to 'earn' the position as 'ready to win C Divn'. Hence the prizes are won in C in a fairer, more reliable, way.




Only if they're going really badly. In the normal scheme of things a third-quartile player loses round 1 and wins round 2. The risk of getting an upset victim in round 2 is countered by the chance of scoring an upset in round 1. Even if a 3rd-quartile player goes loss, loss, their third round opponent is a fourth quartile player. If they still lose that they don't deserve an easy game. :lol:


Yes but. Yes but, the player I was interviewing is one of the old-and-bold. They get creamed by the under-rated juniors in C and D. So all your analysis is based on the presumption of ratings representing current ability and we know that a strong lag effect is operating at BHCC for these youngsters.




Hang on, I did say that I considered 800 point differences to be too much, didn't I? And that, depending on circumstances, I might consider drawing the mark lower for particular tournaments? But you've got work to do if you want to haggle me down to 400.


I think I have moved to 450 as the cut-off point for declaring a junk round.

starter

Alan Shore
08-04-2004, 08:34 AM
Could those high-ranked players, say >1800 for example, list their last 5 games against opponents who have a rating 800 points higher than themselves. I would like to see evidence of how long it took to play a reasonable number of opponents of higher rating in a 'junk' round.

tks
starter

Well starter, I'll go from my last tournament.. First round I played David Smerdon (2400) and won that game. Second round, I played Kevin Sheldrick (2100) and won that game. Third round I played Giang Nguyen (2100) and lost.

Now, playing under accelerated Swiss, I would not have had the opportunity to play these players, and win. The lowest rated player I played I think was around 1750 in that tourn and I got 4.5/7. Under your system, I would most likely not have had a shot at those players. You must understand, the best way to improve is to play stronger players than those in your category - your skill level really steps up a couple of notches.

Similarly, this works for those rated well below you - First round of the Uni Open last year, I played Michael Hill, rated about 800. I appreciated the 'easy round' too since I hadn't slept much the night before. However, I found it wasn't easy at all and had to play myself out of a worse position and fortunately win the endgame. The point there was we both had a good game with chances to win, despite the rating difference - the same with my games against higher opponents.

Now the people I've spoken to are in agreement with me, they like either the opportunity to play a very strong player or to have a comfortable first round. Therefore I'd advise you, just because you're in an administrative position, to listen to the players point of view first, without launching into your spiel about so-called junk rounds.

That's the beauty of chess - anyone can win if they play the right moves, against any given opponent.

Alan Shore
08-04-2004, 08:37 AM
Good morning grayggray

It is a condition of entry that C and D quartile players cannot win the B and A Division prizes. Therefore, the answer is no; D quartile players cannot win the tournament. But I add that in 20 years of running normal SWISSes, no D quartile player ever got more than 3.5/9.

starter

When you say they can't win the rating group prizes of the division above this is fair enough but surely they can still feature in the open 1st, 2nd or 3rd placings? I've come 3rd in a tournament before where I've been seeded in the bottom half of the draw (would have come second too, but for the result on the top board).

jeffrei
08-04-2004, 12:27 PM
When you say they can't win the rating group prizes of the division above this is fair enough but surely they can still feature in the open 1st, 2nd or 3rd placings?

No, they can't. And realistically this wouldn't happen in unaccelerated tournaments at Box Hill anyway because the fields are large (almost always 80+, and often much higher) and because the fields are well-spread across ratings.

Basically the only time acceleration is problematic is if you have a junior in the C quartile rated 1400 but well capable of playing 1850+ strength (e.g. Zhigen Lin and Chris Wallis). In any case I coach both those boys and I'm quite happy that they miss out for the greater good of the club. Their ratings have been going up in due course anyway, and now both of them are in the top half.

ursogr8
08-04-2004, 02:33 PM
Well starter, I'll go from my last tournament.. First round I played David Smerdon (2400) and won that game. Second round, I played Kevin Sheldrick (2100) and won that game. Third round I played Giang Nguyen (2100) and lost.

Bruce my original request was>>


Could those high-ranked players, say >1800 for example, list their last 5 games against opponents who have a rating 800 points higher than themselves. I would like to see evidence of how long it took to play a reasonable number of opponents of higher rating in a 'junk' round.

tks
starter
Now I have not been following the detective story as to your true identity, and therefore I don’t know your rating. And I could not find your (handle) surname in the CAQ list.
But anyhow, if you followed the qualifiers in my post then your rating is 1300 or lower
(that is 2100-800=1300). If this correct then you have had a sensational tournament and it is surprising that the Press or the BB has not remarked on these fantastic results.
More likely though, you did not follow the qualifiers in my post, in which case I am only guessing at your rating.
If your OTB rating is 1800 then you have won games 600 and 300 rating points higher than your rating. Well done under any circimstances.
However, the second result (300) is not even near the category I am calling a junk round. I call a junk round where the average pairing differential is 450. I have no desire to create tournaments that preclude you from meeting players as close to your rating as 300 higher. And your win against this player is something to cherish.

Can I ask you to go back to the drawing board and find games where you have played some-one 800 points higher?
(Can I just remind as to why we are looking at 800 points….some youngsters travelled to a large country tournament in 2003 and obtained two early pairings against players 800 points higher. The inequity of their chances against such opponents put the kids off bothering to travel in 2004. The same kids have been willing to travel to Perth in DEC2003, and Canberra this week. They are keen. But not keen enough to take on a 800 point gap again, twice over, on day 1).




Now, playing under accelerated Swiss, I would not have had the opportunity to play these players, and win. The lowest rated player I played I think was around 1750 in that tourney and I got 4.5/7. Under your system, I would most likely not have had a shot at those players.



I don’t think your paragraph is correct Bruce; but I will await advice of the general area of you rating. For example, are you 1800-1900?



You must understand, the best way to improve is to play stronger players than those in your category - your skill level really steps up a couple of notches.



If you say I must Bruce, then I will understand. But actually, I did understand before you said I must understand.




Similarly, this works for those rated well below you - First round of the Uni Open last year, I played Michael Hill, rated about 800. I appreciated the 'easy round' too since I hadn't slept much the night before. However, I found it wasn't easy at all and had to play myself out of a worse position and fortunately win the endgame. The point there was we both had a good game with chances to win, despite the rating difference - the same with my games against higher opponents.


I need no convincing that ‘ the winner is always a grinner’ Bruce.
However, the evidence that would be relevant is whether Michael (the loser not the grinner) had travel far to have two such losses in a day. Would he put himself out the next year and enter a remote tourney?





Now the people I've spoken to are in agreement with me, they like either the opportunity to play a very strong player or to have a comfortable first round.



The question is one of degree Bruce. As Matthew Sweeney said in post #92, “Do the majority of players want to retain the junk rounds or have all rounds more closely matched? “
So, at what point does a round become a ‘junk’ round where all but the sadists and masochists lose interest?
As I understand your posts to date ‘there is no such round as a bad round without the possibility of improvement’.



Therefore I'd advise you, just because you're in an administrative position, to listen to the players point of view first, without launching into your spiel about so-called junk rounds.


A cheap shot by you Bruce.
I am a player first, not just an administrator.
And yes I do listen to the players. Five of them have their opinions on this thread.
Many more discuss tinkering with the bog-standard SWISS with me over the past 20 years. We experimented and sort feedback for 20 years. It is only in year 21 that this BB has become available for my spiel to unsettle you.





That's the beauty of chess - anyone can win if they play the right moves, against any given opponent.


Your statement is true Bruce.
And if it was the only criteria we were to use in the construction of parings then we would draw the names from a hat and let consecutive drawings rule as pairings.

But we don’t, even in a bog-standard SWISS.

Instead, the bog-standard SWISS creates pairings of similar rating difference, presumably to try to create as many competitive games as possible.

starter

Garvinator
08-04-2004, 02:41 PM
bruce, have you played in an accelerated swiss before?

Kevin Bonham
08-04-2004, 03:42 PM
Instead, the bog-standard SWISS creates pairings of similar rating difference, presumably to try to create as many competitive games as possible.

No, the aim of the bog-standard Swiss is to sort the field as efficiently as possible at all stages irrespective of any other considerations. I'm not saying it always succeeds in this aim, just that that is what it is designed to do. If it wanted to create as many competitive games as possible, the first round pairing might be 1-2, 3-4, 5-6 etc.

An interesting point here is that the Swiss is "blind" to the length of the event (except for the last round pairing rules). Thus the results after round 1 are meant to be the best possible sort after 1 round, those after round 2 the best possible sort after 2 rounds, etc.

Accelerated systems can only improve on the basic Swiss because they are not "blind"; the scores they produce after round 1 are a nonsensical sorting of the field but that's OK because the system corrects it later.

Your system provides for more competitive games by sacrificing the "sorting" of the top half vs the bottom half altogether, meaning a bottom-half player cannot win the tournament, and indeed cannot bother the outright contenders until very late in the show even if winning every single game. Whether it sorts each individual half better than a standard Swiss is an interesting question. Thinking about it more, I'd expect that it would be somewhat better at this because the first round of a standard Swiss sorts into halves (barring upsets) anyway.

ursogr8
08-04-2004, 05:29 PM
No, the aim of the bog-standard Swiss is to sort the field as efficiently as possible at all stages irrespective of any other considerations. I'm not saying it always succeeds in this aim, just that that is what it is designed to do. If it wanted to create as many competitive games as possible, the first round pairing might be 1-2, 3-4, 5-6 etc.



I think I accept what you say. But in an 16 player SWISS (for example) what is (the sub-objective) driving the pairings to
1v9,10v2,3v11,12v4 etc
instead of
1v16,15v2,3v14,13v4 etc?



Your system provides for more competitive games by sacrificing the "sorting" of the top half vs the bottom half altogether, meaning a bottom-half player cannot win the tournament, and indeed cannot bother the outright contenders until very late in the show even if winning every single game. Whether it sorts each individual half better than a standard Swiss is an interesting question. Thinking about it more, I'd expect that it would be somewhat better at this because the first round of a standard Swiss sorts into halves (barring upsets) anyway.

Thanks, some nice views on the how we achieve competitiveness.

starter

Kevin Bonham
08-04-2004, 11:57 PM
I think I accept what you say. But in an 16 player SWISS (for example) what is (the sub-objective) driving the pairings to
1v9,10v2,3v11,12v4 etc
instead of
1v16,15v2,3v14,13v4 etc?

I don't know if there's an intended theory justification for it, but here's some arguments that could be made.

It sorts the better players more effectively. For instance suppose you have 20 players staggered at 50 pt intervals so that 1-11, 2-12, 3-13 etc are all 500 pt gaps. The better players will, on average, get something like nine wins and a draw.

If you did 1-20, 2-19, 3-18 etc you'll get some massacres but you also get some close games. The bottom half might get say 1.5/10. Say #11 beats #10. Now it may turn out that #11 is playing better on the weekend in question and deserves to be ahead of #10, but one game doesn't prove that. It would be better to defer their clash til later when they've had more chances to prove they are (or are not) around the same standard on the weekend in question, rather than making them play each other right away.

Another, probably stronger, possible argument is fairness. As it is everyone in the top half has a roughly equal task in round 1, so they should all be equally exhausted (or not) for round 2. Whereas the game between #10 and #11 is going to be very close and difficult - and the winner will be fed to #1 next round. #1 not only gets a pile of the weakest opponents possible, but he is also going to keep getting relatively weak opponents who just won relatively difficult (for them) games. Even if #1 is way out of form, you're probably not going to find out until he plays #2 under these circumstances.

Alan Shore
09-04-2004, 12:19 AM
bruce, have you played in an accelerated swiss before?

I have - It was a ridiculous affair, where 6 players finished on 6/7. (Sorry to say I was on 6/6 but lost the last round, thus screwing everything, doh!)

Alan Shore
09-04-2004, 12:36 AM
Now I have not been following the detective story as to your true identity, and therefore I don’t know your rating. And I could not find your (handle) surname in the CAQ list.
But anyhow, if you followed the qualifiers in my post then your rating is 1300 or lower
(that is 2100-800=1300). If this correct then you have had a sensational tournament and it is surprising that the Press or the BB has not remarked on these fantastic results.
More likely though, you did not follow the qualifiers in my post, in which case I am only guessing at your rating.
If your OTB rating is 1800 then you have won games 600 and 300 rating points higher than your rating. Well done under any circimstances.
However, the second result (300) is not even near the category I am calling a junk round. I call a junk round where the average pairing differential is 450. I have no desire to create tournaments that preclude you from meeting players as close to your rating as 300 higher. And your win against this player is something to cherish.

Can I ask you to go back to the drawing board and find games where you have played some-one 800 points higher?
(Can I just remind as to why we are looking at 800 points….some youngsters travelled to a large country tournament in 2003 and obtained two early pairings against players 800 points higher. The inequity of their chances against such opponents put the kids off bothering to travel in 2004. The same kids have been willing to travel to Perth in DEC2003, and Canberra this week. They are keen. But not keen enough to take on a 800 point gap again, twice over, on day 1).


Starter, I'm not an idiot, I can read that you said 800 points. At the time, my rating was horribly underrated (my point) in the low 1400's. Given this is a difference of 700-800, I'm sure you wouldn't be so picky to say 'oh but that's not above 800'

As for the whole 'travel' and 'only 1 or 2 games a day' I could make such a concession for circumstances like this. Specifically I do not want the accelerated Swiss in weekenders or one-day tournaments. For tournaments like Australian Juniors, I am more flexible on the subject.


I don’t think your paragraph is correct Bruce; but I will await advice of the general area of you rating. For example, are you 1800-1900?

I improved approximately 400 points, from low 1400's to 1829, from a performance rating of 2100+ in that tournament.




I need no convincing that ‘ the winner is always a grinner’ Bruce.
However, the evidence that would be relevant is whether Michael (the loser not the grinner) had travel far to have two such losses in a day. Would he put himself out the next year and enter a remote tourney?

He was a local junior.

Another such instance I can give you, is Tristan Stevens travelling to a tournament from SA to (NSW or ACT?) and winning his first three rounds against players approximately 800 points higher (Brian Jones, Kerry Stead, etc.).




The question is one of degree Bruce. As Matthew Sweeney said in post #92, “Do the majority of players want to retain the junk rounds or have all rounds more closely matched? “
So, at what point does a round become a ‘junk’ round where all but the sadists and masochists lose interest?
As I understand your posts to date ‘there is no such round as a bad round without the possibility of improvement’.

I wouldn't say completely without possibility of improvement but I will say the accelerated swiss is a worse system than the current swiss - it does not properly separate players and can lead to no clear winners of categories as no one has had sufficient chance to prove themselves against stronger opposition.



A cheap shot by you Bruce.
I am a player first, not just an administrator.
And yes I do listen to the players. Five of them have their opinions on this thread.
Many more discuss tinkering with the bog-standard SWISS with me over the past 20 years. We experimented and sort feedback for 20 years. It is only in year 21 that this BB has become available for my spiel to unsettle you.

If this is the case then that is fine starter, as long as you did consult the players. I remember last year when Shaun Press DID NOT consult players, and went ahead with his ridiculous 'no drop-mate' rule for the ACT Transfer tournament, even though the standard for all other Transfer OTB tournaments in Australia, and Internationally and on FICS allows drop-mate.


I'm sure we both want what is best for chess starter. But I believe the swiss system is necessary for underrated players (particularly juniors) to have a chance to get near their rating as quickly as possible.

ursogr8
09-04-2004, 10:27 AM
Starter, I'm not an idiot, I can read that you said 800 points. At the time, my rating was horribly underrated (my point) in the low 1400's. Given this is a difference of 700-800, I'm sure you wouldn't be so picky to say 'oh but that's not above 800'

hi BD. Yes, I am picky, but not picky enough to discount your evidence. So, thanks for that.

I must confess to being staggered that your rating was in the low 1400’s and yet you beat 2 players whose rating was >2000. I was unable to conceive of such a thing when I was writing my previous post. In 30 years of chess administration I had not heard of such a great accomplishment. Sincerely…well done.

But how on earth did you get a rating of 1400? It seems like your true rating is above 1800. Anyway, I accept your anecdote and now move on in the debate.

Now I am going to say something controversial. I am going to say your anecdote should be disregarded, rejected as irrelevant, ignored, discarded. Sorry. Your rating improvement is a wonderful achievement. BUT, we have to design tournaments (and in particular pairing rules) to cater for the norm; not a once-in-a-lifetime distortion that your under-rating has caused. What this means is that I am still convinced that we should not adopt pairing rules that give an average pairing difference above 450. I say this even though you have produced out of left field an example of where it was justifiable to be paired against some-one 800 points higher. But your evidence is an exception that should be ignored. We should not design tournaments to cater for those who are grossly under-rated. We must instead use an assumption that all entrants have a rating that represents reasonably their playing strength.





As for the whole 'travel' and 'only 1 or 2 games a day' I could make such a concession for circumstances like this. Specifically I do not want the accelerated Swiss in weekenders or one-day tournaments. For tournaments like Australian Juniors, I am more flexible on the subject.



Good to see you move from your hard-line position. As the ladies who walk up and down Kings Cross say "Now we are just negotiating the price"; or in our case, how much to tweak and which tournaments to tweak.



Another such instance I can give you, is Tristan Stevens travelling to a tournament from SA to (NSW or ACT?) and winning his first three rounds against players approximately 800 points higher (Brian Jones, Kerry Stead, etc.)



Is this another example of a grossly under-rated?




…………………. but I will say the accelerated swiss is a worse system than the current swiss - it does not properly separate players and can lead to no clear winners of categories as no one has had sufficient chance to prove themselves against stronger opposition.



What we do is give the top-half a permanent bonus of 2 points for the duration of the tournament. For C quartile players this means they score up to 6.5/9, instead of the likely 3.5/9 in a bog-standard SWISS. I can assure you that a C quartile players thus has 9 competitive games with no junk rounds. The results are thus far more reliable; contrary to your post. (Similarly for B and D).





If this is the case then that is fine starter, as long as you did consult the players.


BD. Well I said I did.




I'm sure we both want what is best for chess starter. But I believe the swiss system is necessary for underrated players (particularly juniors) to have a chance to get near their rating as quickly as possible.


If under-rating is the issue then please debate with Bill to get the formula fixed. To me it seems an incorrect approach to be arguing for tournament design to be influenced by the need to correct rating-anomalies.

Thank you for your post.


starter

jay_vee
09-04-2004, 12:50 PM
Now I am going to say something controversial. I am going to say your anecdote should be disregarded, rejected as irrelevant, ignored, discarded. Sorry. Your rating improvement is a wonderful achievement. BUT, we have to design tournaments (and in particular pairing rules) to cater for the norm; not a once-in-a-lifetime distortion that your under-rating has caused. What this means is that I am still convinced that we should not adopt pairing rules that give an average pairing difference above 450. I say this even though you have produced out of left field an example of where it was justifiable to be paired against some-one 800 points higher. But your evidence is an exception that should be ignored. We should not design tournaments to cater for those who are grossly under-rated. We must instead use an assumption that all entrants have a rating that represents reasonably their playing strength.


Maybe not quite so. Why don't you include the possibility to 'opt-out' for a player who considers himself grossly under-rated? That way, the (very few, imho) players who actually are under-rated and are interested in competitive games based on their real strength not their rating, get a chance at that (as well as a faster increase in rating, if they perform above expectations). Of course, these players should be considered 'artificial' B-quartile players and barred from winning bottom-half prizes.

ursogr8
09-04-2004, 04:37 PM
Maybe not quite so. Why don't you include the possibility to 'opt-out' for a player who considers himself grossly under-rated? That way, the (very few, imho) players who actually are under-rated and are interested in competitive games based on their real strength not their rating, get a chance at that (as well as a faster increase in rating, if they perform above expectations). Of course, these players should be considered 'artificial' B-quartile players and barred from winning bottom-half prizes.

hi jay-vee

First, let me say that I am not a Director of Play, and I have not read the rules. So, what I about to say is from a position of ignorance.
If I took an entry from a player that I knew to be grossly under-rated, or who declared him(her)self under-rated then I would treat them as un-rated and they would be eligible for no prize (save the UNRATED prize if there is one).
In practice I would make an estimate if they were in the top half or the bottom half of the table. If in the top half, they get a permanent bonus of +2 for the duration of the tournament. They will then meet top players if they win early rounds.

starter

jay_vee
09-04-2004, 07:13 PM
If I took an entry from a player that I knew to be grossly under-rated, or who declared him(her)self under-rated then I would treat them as un-rated and they would be eligible for no prize (save the UNRATED prize if there is one).

I assume you meant to say 'no ratings-based prize' (i.e. aside from the unrated they could still win open, junior, etc. prizes). Still, I can't say I agree with you. Really, if you look at your average tournament, most winners of rating-prizes would have vastly out-performed their rating. So they could be considered to be under-rated as well, I suppose, which in turn should make them inelligible for the prize :-). I guess, the only case where I would support this drastic measure would be if someone was intentionally keeping his rating low. And even if I suspected something like that, instead I'd prefer a rule that prevented anyone from winning the same rating-prize more than once every, say, 3 years.
But if someone is genuinely under-rated, and it's not their fault (and I wouldn't blame anyone for improving too fast :-)), why should they not be able to win the prize?

ursogr8
09-04-2004, 09:37 PM
I assume you meant to say 'no ratings-based prize' (i.e. aside from the unrated they could still win open, junior, etc. prizes).

In the context of a BHCC SWISS (where the top half get a bonus of +2 for the whole tournament) if they were placed in A at the start of the tournament only then could they win the OEN prizes. We don’t have junior prizes any more.



Still, I can't say I agree with you. Really, if you look at your average tournament, most winners of rating-prizes would have vastly out-performed their rating. So they could be considered to be under-rated as well, I suppose


But I never suggested retrospective treatment of mis-rated players. I only indicated treating as un-rated at the point of entry.



But if someone is genuinely under-rated, and it's not their fault (and I wouldn't blame anyone for improving too fast :-)), why should they not be able to win the prize?


Fault does not come into it.
If a player does not have a rating that represents their playing strength, in their view, then they effectively don’t have a rating.

starter

Garvinator
09-04-2004, 11:20 PM
sorry starter for another question ;) if a player requested to be placed in the top half of the draw for the tournament, would BHCC do it?

eclectic
10-04-2004, 12:50 AM
and another question (slightly off topic)

rd 1 doeberl cup

can dragecivic (2009) defeating rogers (2668) be classed a "junk-round" upset?

eclectic

Rincewind
10-04-2004, 01:00 AM
It happens.

Garvinator
10-04-2004, 11:30 AM
and another question (slightly off topic)

rd 1 doeberl cup

can dragecivic (2009) defeating rogers (2668) be classed a "junk-round" upset?

eclectic
no, because it is only one result in one round. Junk rounds refer to the rating differential (roughly) for a whole round, not just one game. A rough guide for a junk round is where each pairing in a round averages 450.

Also junk rounds might not be as relevant because of the status of the event and because doebrel is divided into three divisions.

ursogr8
10-04-2004, 04:05 PM
sorry starter for another question ;) if a player requested to be placed in the top half of the draw for the tournament, would BHCC do it?

hi gg

Our previous solution was to run two tournaments (roughly A and B). And the top B's, in particular rising juniors in B, had the option of playing in A. Two problems emerged: > at the point of entry (7.15 pm for a 7.45 pm start) we had a noisy room of 100 entrants with at least 10 of them chopping and changing and arguing about which Division to play in. >> on evening #2 (or later) what happens when A Division has an odd number and B Division has an odd number? Remember, this is a weekly evening Club, not a week-ender; so the number who turn-up out of 100 can be a little variable. Hence the solution that repairs all problems is to put them in one tournament and give the top half 2 points bonus.
The answer to your question is no.

starter

Garvinator
10-04-2004, 04:09 PM
hi gg

Our previous solution was to run two tournaments (roughly A and B). And the top B's, in particular rising juniors in B, had the option of playing in A. Two problems emerged: > at the point of entry (7.15 pm for a 7.45 pm start) we had a noisy room of 100 entrants with at least 10 of them chopping and changing and arguing about which Division to play in. >> on evening #2 (or later) what happens when A Division has an odd number and B Division has an odd number? Remember, this is a weekly evening Club, not a week-ender; so the number who turn-up out of 100 can be a little variable. Hence the solution that repairs all problems is to put them in one tournament and give the top half 2 points bonus.
The answer to your question is no.

starter

i meant from the c quartile to the a or b quartile, so that player who wants the high rated games receives the 2 point acceleration.

ursogr8
10-04-2004, 04:10 PM
no, because it is only one result in one round. Junk rounds refer to the rating differential (roughly) for a whole round, not just one game. A rough guide for a junk round is where each pairing in a round averages 450.

Also junk rounds might not be as relevant because of the status of the event and because doebrel is divided into three divisions.

gg

What a beautiful answer. Spot on. Perfect.

gg 1 - eclectic 0

starter

eclectic
10-04-2004, 04:58 PM
no, because it is only one result in one round. Junk rounds refer to the rating differential (roughly) for a whole round, not just one game. A rough guide for a junk round is where each pairing in a round averages 450.

Also junk rounds might not be as relevant because of the status of the event and because doebrel is divided into three divisions.
gg,

yeah,

right about the three divisions

and this year too they've shut u1600 players out of the top division

but there might be more rd 1 games whose rating differential did exceed 450

however, rogers' rating is 100-150 more than the next player so that immediately causes a statistical aberration

anyway, it's a great term "junk-round"

it puts lower rated people in their place where they belong and conveys the idea that higher rated players perhaps wish they did not have to bother turning up for such rounds

whatever you do, micro juniors or precocious late starters, don't study too hard or grasp the finer points of the game too quickly, as we don't want the anomaly between your displayed rating and your actual performance upsetting desires for predictable outcomes in well massaged swiss tournaments

if there is an objection in some quarters to the "junk-round" caused by regular swisses why not (where feasible) organise events into round robin divisions and be done with it ?

that seems better than the super accelerated split swiss deals where lower rated people get "knobbled" two points and are told they're not entitled to win the event


:hmm:

eclectic

Garvinator
10-04-2004, 05:02 PM
I have suggested this idea previously but didnt get any support. My personal opinion is that different style of tournaments should be designed to promote variety, so staleness and tourney boredom does not set in. This would combine bog-standard/monster style swiss, accelerated versions, round robins, fischer random etc.

interesting, no comment from the pro monster swiss camp on this point/idea.

ursogr8
11-04-2004, 07:56 AM
gg,

yeah,

right about the three divisions

and this year too they've shut u1600 players out of the top division

but there might be more rd 1 games whose rating differential did exceed 450

however, rogers' rating is 100-150 more than the next player so that immediately causes a statistical aberration


hi eclectic. Nice morning to read your post.




anyway, it's a great term "junk-round"



Yes. It's bit of marketing in that one. But like all evocative terms it tends to polarise the debate and some folk react against it. Have you any alternate names to describe a round where the average rating difference is very large in contrast to a competitive round?




it puts lower rated people in their place where they belong and conveys the idea that higher rated players perhaps wish they did not have to bother turning up for such rounds


On the contrary eclectic. It simply means statistically that most games will result as per the rating difference; thus pleasing but a handful of a large field.




whatever you do, micro juniors or precocious late starters, don't study too hard or grasp the finer points of the game too quickly, as we don't want the anomaly between your displayed rating and your actual performance upsetting desires for predictable outcomes in well massaged swiss tournaments



More hyperbole eclectic. Mildly humorous, but hyperbole nonetheless.
And your use of ‘predictable outcomes’; care to expand. It is my contention that players like competitive games, which of course goes hand in hand with ‘unpredictable outcomes’, not ‘predictable outcomes’.






if there is an objection in some quarters to the "junk-round" caused by regular swisses why not (where feasible) organise events into round robin divisions and be done with it ?



Well we do. And best sellers they are too. For example, our FIDE round-robin. For example Grades and interclubs in various States.




that seems better than the super accelerated split swiss deals where lower rated people get "knobbled" two points and are told they're not entitled to win the event


I agree that round-robins and Grades are better than ‘super accelerated split swiss deals’.

(Listen eclectic, your phrase ‘super accelerated split swiss deals’ is a sensational description; do you mind if I use that in the marketing)?

regards
starter

Garvinator
11-04-2004, 12:42 PM
super accelerated split swiss deals’
sounds more like a dodgy swiss banking scheme ;) :whistle:

Alan Shore
11-04-2004, 03:27 PM
hi BD. Yes, I am picky, but not picky enough to discount your evidence. So, thanks for that.

I must confess to being staggered that your rating was in the low 1400’s and yet you beat 2 players whose rating was >2000. I was unable to conceive of such a thing when I was writing my previous post. In 30 years of chess administration I had not heard of such a great accomplishment. Sincerely…well done.

But how on earth did you get a rating of 1400? It seems like your true rating is above 1800. Anyway, I accept your anecdote and now move on in the debate.

Thanks starter. Not only that, I also drew against 2 other players over 2000 (I won't talk about the wins I missed in both those games though :wall: )


Now I am going to say something controversial. I am going to say your anecdote should be disregarded, rejected as irrelevant, ignored, discarded. Sorry. Your rating improvement is a wonderful achievement. BUT, we have to design tournaments (and in particular pairing rules) to cater for the norm; not a once-in-a-lifetime distortion that your under-rating has caused. What this means is that I am still convinced that we should not adopt pairing rules that give an average pairing difference above 450. I say this even though you have produced out of left field an example of where it was justifiable to be paired against some-one 800 points higher. But your evidence is an exception that should be ignored. We should not design tournaments to cater for those who are grossly under-rated. We must instead use an assumption that all entrants have a rating that represents reasonably their playing strength.

I'll give you another example of having the opportunity to play higher rated's that have paid divedends: In the 2001 Hervey Bay Open, Nathan Davidson (1500) defeated Myers (2000) Yu (2200) and Solomon (2400) so my own feat was not as incredibly remarkable as it may seem (I didn't even win a prize in my tournament either, lol).

And, I noticed GM Rogers has lost his first round in the Doeberl... how opportune for Dragicevic that we have the lovely swiss system to allow him the chance to do this :)

Also, I wouldn't call these upsets entirely 'out of left field' they happen quite often, especially where I come from where there is a high level of juniors participating. It gives these young players the opportunity and the exposure to, the higher rated's - even to PLAY against a titled player, or even someone over 2000 is a big deal for them.




Good to see you move from your hard-line position. As the ladies who walk up and down Kings Cross say "Now we are just negotiating the price"; or in our case, how much to tweak and which tournaments to tweak.

I won't make any comments on your choice of analogy, I know how sensitive you can be ;) But of course, I'm always willing to listen if anyone makes a good case for something.



Is this another example of a grossly under-rated?

Most likely - The Nathan Davidson example however is not - he remains about the same rating he has always held these past few years.



What we do is give the top-half a permanent bonus of 2 points for the duration of the tournament. For C quartile players this means they score up to 6.5/9, instead of the likely 3.5/9 in a bog-standard SWISS. I can assure you that a C quartile players thus has 9 competitive games with no junk rounds. The results are thus far more reliable; contrary to your post. (Similarly for B and D).

Giving them this many more points is ridiculous - it is not an accurate portrayal of their true performance and demeans higher rated players finishing on a lower score - not to mention how much it can clog a leaderboard when determining prizes.



If under-rating is the issue then please debate with Bill to get the formula fixed. To me it seems an incorrect approach to be arguing for tournament design to be influenced by the need to correct rating-anomalies.

He already has - it's called the Glicko. I can't praise it enough against the old and very slow ELO system...

Alan Shore
11-04-2004, 03:29 PM
Yes. It's bit of marketing in that one. But like all evocative terms it tends to polarise the debate and some folk react against it. Have you any alternate names to describe a round where the average rating difference is very large in contrast to a competitive round?

I'd call it the 'upset-opportunity' round for the lower player or the 'cake' round for the higher :)

Garvinator
11-04-2004, 06:17 PM
Giving them this many more points is ridiculous - it is not an accurate portrayal of their true performance and demeans higher rated players finishing on a lower score - not to mention how much it can clog a leaderboard when determining prizes.
i think a clarification/explaination is required here. From my understanding and playing around with swiss perfect, the adding of two points refers to how the accelerated players are paired. The computer recognises those accelerated players as having two points from the start compared to zero for the bottom half. The pairings are then made with this acceleration.

After each round when the standings are published, there isnt any two points added on to each player. they are all starting and finishing with no points added on.

the reference to two points added on is mainly for the swissperfect system so it can easily differentiate who is in which group.

Alan Shore
11-04-2004, 06:55 PM
i think a clarification/explaination is required here. From my understanding and playing around with swiss perfect, the adding of two points refers to how the accelerated players are paired. The computer recognises those accelerated players as having two points from the start compared to zero for the bottom half. The pairings are then made with this acceleration.

After each round when the standings are published, there isnt any two points added on to each player. they are all starting and finishing with no points added on.

the reference to two points added on is mainly for the swissperfect system so it can easily differentiate who is in which group.

I'm aware it's not literally adding points but it may as well be from the resultant pairing format. I maintain this creates imbalances in the system, not just 'differentiating the groups'. Statistically I believe there will be a skew in results, when comparing scores.

eclectic
11-04-2004, 07:01 PM
I'm aware it's not literally adding points but it may as well be from the resultant pairing format. I maintain this creates imbalances in the system, not just 'differentiating the groups'. Statistically I believe there will be a skew in results, when comparing scores.
it takes a special sleight of hand to be a skilled swiss masseur/masseuse

;)

eclectic

ursogr8
11-04-2004, 08:05 PM
Thanks starter. Not only that, I also drew against 2 other players over 2000 (I won't talk about the wins I missed in both those games though :wall: )




Very special performances BD.
But let me repeat (while not detracting from your accomplishments); > rare events in general, and >> the opportunity for such upsets should not be the raison detre for tournament format design.





I'll give you another example of having the opportunity to play higher rated's that have paid divedends: In the 2001 Hervey Bay Open, Nathan Davidson (1500) defeated Myers (2000) Yu (2200) and Solomon (2400) so my own feat was not as incredibly remarkable as it may seem (I didn't even win a prize in my tournament either, lol).



Same comment BD. Extreme cases. I am ignoring this evidence.




And, I noticed GM Rogers has lost his first round in the Doeberl... how opportune for Dragicevic that we have the lovely swiss system to allow him the chance to do this :)



BD
gg has already hit this half-court slow-pace lob for a winner. (See response to an eclectic post.




Also, I wouldn't call these upsets entirely 'out of left field' they happen quite often, especially where I come from where there is a high level of juniors participating.



Again I suggest the solution is to get the ratigs fixed; don't distort the tournament design for a 100 player field just to cater for one incorrectly rated player.



It gives these young players the opportunity and the exposure to, the higher rated's - even to PLAY against a titled player, or even someone over 2000 is a big deal for them.



An admirable objective. But at the cost of forcing the other 98 players to have a junk round; not a smart or equitable choice.




Giving them this many more points is ridiculous - it is not an accurate portrayal of their true performance and demeans higher rated players finishing on a lower score - not to mention how much it can clog a leaderboard when determining prizes.



And you were starting to get the hang of this DB until this little paragraph of yours. I challenge you to actually look at some results of tournaments played with this bonus system before you comment from your theoretical armchair. You are just wrong on this part.
Stay engaged though.
gg understands this topic exceptionally well.

starter

ursogr8
11-04-2004, 08:06 PM
I'd call it the 'upset-opportunity' round for the lower player or the 'cake' round for the higher :)

4/10 for both.
No cigar.

ursogr8
11-04-2004, 08:08 PM
i think a clarification/explaination is required here. From my understanding and playing around with swiss perfect, the adding of two points refers to how the accelerated players are paired. The computer recognises those accelerated players as having two points from the start compared to zero for the bottom half. The pairings are then made with this acceleration.

After each round when the standings are published, there isnt any two points added on to each player. they are all starting and finishing with no points added on.

the reference to two points added on is mainly for the swissperfect system so it can easily differentiate who is in which group.


There is an admin. job waiting for you at BHCC gg''.
You have graduated to mentor status.
Well done. Good explanation.

ursogr8
11-04-2004, 08:09 PM
I'm aware it's not literally adding points but it may as well be from the resultant pairing format. I maintain this creates imbalances in the system, not just 'differentiating the groups'. Statistically I believe there will be a skew in results, when comparing scores.


BD

Look at some evidence before you make these assertions. Results contradict your theory.

starter

eclectic
11-04-2004, 08:45 PM
get out of your theoretical armchair
if everyone here heeded that advice then this bulletin board would die such a death !!

:whistle:

eclectic

Garvinator
11-04-2004, 08:57 PM
if everyone here heeded that advice then this bulletin board would die such a death !!

:whistle:

eclectic
and no one would be volunteers either ;) :hmm:

Garvinator
11-04-2004, 08:59 PM
There is an admin. job waiting for you at BHCC gg''.
You have graduated to mentor status.
Well done. Good explanation.

how much does this job pay :lol: do i get airfares and accommodation too, how about meals :lol: :owned: :whistle: :uhoh:

eclectic
11-04-2004, 09:08 PM
how much does this job pay :lol: do i get airfares and accommodation too, how about meals :lol: :owned: :whistle: :uhoh:
gg,

don't use swiss air

though perhaps super accelerated the inherent price handicapping would mean you could go no further than sydney

;) :lol:

eclectic

ursogr8
11-04-2004, 10:54 PM
if everyone here heeded that advice then this bulletin board would die such a death !!

:whistle:

eclectic

eclectic
While your post was cute, the title was revealing. Now just how many people say 'OTB praxis ' in normal conversation. To be honest I have never heard it while on the elevator when I go to get my coffee, nor in the murmurings of the smokers standing just outside the front door. In fact the only person I heard talk like that was....

ursogr8
11-04-2004, 10:55 PM
gg,

don't use swiss air

though perhaps super accelerated the inherent price handicapping would mean you could go no further than sydney

;) :lol:

eclectic

Genuinely funny eclectic..........:lol:

ursogr8
11-04-2004, 10:58 PM
how much does this job pay :lol: do i get airfares and accommodation too, how about meals :lol: :owned: :whistle: :uhoh:

gg''

The answer lies in last years best post (as voted).
BYO
and
We will recognise you.

starter

Alan Shore
12-04-2004, 06:18 AM
http://www.gotfuturama.com/Desktop/wallpaper/farnsworthwallpape640.jpg


:D

Alan Shore
12-04-2004, 06:31 AM
BD is legendary.

:owned: Flattery will get you everywhere, hehe



Very special performances BD.
But let me repeat (while not detracting from your accomplishments); > rare events in general, and >> the opportunity for such upsets should not be the raison detre for tournament format design.

OK, if you won't listen, you won't listen, but they are there and they are spectacular when they happen - OF COURSE they're going to be rare when you consider the rating difference but it makes it so much more special and memorable when these upsets happen.



Same comment BD. Extreme cases. I am ignoring this evidence.

So be it, Jedi.


BD
gg has already hit this half-court slow-pace lob for a winner. (See response to an eclectic post.

Yes yes, exception to the rule, you've said it before. Why don't you ask Dragicevic himself though :p




Again I suggest the solution is to get the ratigs fixed; don't distort the tournament design for a 100 player field just to cater for one incorrectly rated player.


An admirable objective. But at the cost of forcing the other 98 players to have a junk round; not a smart or equitable choice.

I should add something here: Not only do juniors find it an experience to play against someone quite strong, but many adults do too! For one thing, there are pretty much as many juniors as adults in Gold Coast tournaments anyway, so your figure of '98' is simply ludicrous. Added to that, those players who enjoy the easy first round and you actually get a lower % of people opposed to your "junk round".



And you were starting to get the hang of this DB until this little paragraph of yours. I challenge you to actually look at some results of tournaments played with this bonus system before you comment from your theoretical armchair. You are just wrong on this part.
Stay engaged though.
gg understands this topic exceptionally well.

Show me a tournament played with acc. swiss then please, I'd like to thoroughly analyse the final standings for myself - I was merely going by the one tourn I had played in with the system and the clustering together was rampant in that instance.

Alan Shore
12-04-2004, 06:35 AM
4/10 for both.
No cigar.

Subjective opinion. :hand:

It's possible there may be better terms around, but you got in first with junk so I don't mind sticking with it to humour you :p

Alan Shore
12-04-2004, 06:39 AM
BD

Look at some evidence before you make these assertions. Results contradict your theory.

starter

Show me the money!


As for my theoretical armchair, it's very comfortable, with imported leather.. pretty hard to budge from sometimes ;)

Alan Shore
12-04-2004, 06:51 AM
I have suggested this idea previously but didnt get any support. My personal opinion is that different style of tournaments should be designed to promote variety, so staleness and tourney boredom does not set in. This would combine bog-standard/monster style swiss, accelerated versions, round robins, fischer random etc.

interesting, no comment from the pro monster swiss camp on this point/idea.

Well, sure, guess you could have some variety on occasion. If you want to have these kind of tournaments though then there should be no exorbitant entry fees.. the fischer-random tournament held in Brisbane a few months ago only attracted 8 players.. no wonder if you're going to charge $30-40 entry fees. I am all for chess variants.. more transfer, more transfer! :)

ursogr8
12-04-2004, 08:38 AM
Show me the money!


As for my theoretical armchair, it's very comfortable, with imported leather.. pretty hard to budge from sometimes ;)

Bruce
Here is the most recent
http://www.boxhillchess.org.au/e2004/e0401fac/crosstable.htm

regards
starter

ps BTW, > leather sounds very conservative
>> imported sounds like 'cultural cringe'.

ursogr8
12-04-2004, 08:39 AM
Well, sure, guess you could have some variety on occasion. If you want to have these kind of tournaments though then there should be no exorbitant entry fees.. the fischer-random tournament held in Brisbane a few months ago only attracted 8 players.. no wonder if you're going to charge $30-40 entry fees. I am all for chess variants.. more transfer, more transfer! :)

Yes
$20 tops is the usual, Box Hill entry cost.
starter

ps Transfer has been transferred to outside the playing Hall. Into the courtyard.

ursogr8
12-04-2004, 08:49 AM
Yes yes, exception to the rule, you've said it before. Why don't you ask Dragicevic himself though :p



BD
No problems
He has been a member at Box Hill for 5 years.
But then the question doesn't come up in our environment for Dragicevic because we have no way of offering the prize-money that Canberra offers, hence we cannot get players of the rating of Rogers.



I should add something here: Not only do juniors find it an experience to play against someone quite strong, but many adults do too! For one thing, there are pretty much as many juniors as adults in Gold Coast tournaments anyway, so your figure of '98' is simply ludicrous. Added to that, those players who enjoy the easy first round and you actually get a lower % of people opposed to your "junk round".



Yes, the 98% was a provocative exaggeration.
There have been various posts earlier about survey's revealing player attitude to all of this. That would be one way of getting a better appreciation of the general consensus.




Show me a tournament played with acc. swiss then please, I'd like to thoroughly analyse the final standings for myself - I was merely going by the one tourn I had played in with the system and the clustering together was rampant in that instance.

http://www.boxhillchess.org.au/e2004/e0401fac/crosstable.htm


starter

Alan Shore
12-04-2004, 09:43 AM
Bruce
Here is the most recent
http://www.boxhillchess.org.au/e2004/e0401fac/crosstable.htm



Um... ok, a bit to say about this.

What I said about the warping of scores seems to be as I expected here: A 450 rated player getting a higher score than 1600 and 1700 rated players. An 1100 rated player being a full point higher than a 2000 and three 1900 players. Despite them being in different divisions, this is an enormous discrepancy.

You were fortunate Geoff Saw owned everyone in that tournament, otherwise it may have been a cluttered leaderboard. I think the most critical pairing came in the final round between Elbrish and Yu - this can't be for real?? You almost had a 1300 come outright second (even if under your system he's ineligible to win open prizes) but if he was rated just 7 points higher than he was, then what?!

I can see now, what you were saying about C quartile players and such, with your tournament being sliced in two. This has been tried before but not in the same style, with lower div. players not playing higher div. ones. The problems with that system are the players in the lower half of the top draw are given a really raw deal regarding any chance at a prize. Hopefully you allow for this problem with rating category prizes. I'll concede the acc. swiss is an okish alternative for your bi-divisionary tournament yet there's no way you could seriously consider implementing it into a tournament without such divisions.

ursogr8
12-04-2004, 01:46 PM
Um... ok, a bit to say about this.

What I said about the warping of scores seems to be as I expected here: A 450 rated player getting a higher score than 1600 and 1700 rated players. An 1100 rated player being a full point higher than a 2000 and three 1900 players. Despite them being in different divisions, this is an enormous discrepancy.


BD
But it is not a problem because the C (and D) blokes/gals are essentially in a different tournament. They cannot win B or A prizes, so outscoring and A or B is no drawback at all. (The warp is not a problem). However, the C and D players have enjoyed the opportunity to earn the right to play against the B and perhaps the A players and so be challenged. The C and D players can win good rating prizes in their Division (up to 60% of the OPEN prize). The C and D players have a fair result because they are winning a lot of games and earn their prize; not a lucky last round pairing.




You were fortunate Geoff Saw owned everyone in that tournament, otherwise it may have been a cluttered leaderboard.



I only offered to give you real data, not hypothetical data. Of course, if you want to pursue I will give you another tournament.




I think the most critical pairing came in the final round between Elbrish and Yu - this can't be for real?? You almost had a 1300 come outright second (even if under your system he's ineligible to win open prizes) but if he was rated just 7 points higher than he was, then what?!



This one is a full toss, hip-high, on the leg stump. If Elbrish was 9 points higher rated then he would have displaced Chong in Division A. Then Elbrish would have been in B Division and would have had quite a different opponent profile in the early rounds.
You see BD, we draw the line between B and C so that the top field is an even number, and with a bit of judicious thought. For example if Joe Superunderratedjunior was the top seed in C then we would likely draw the line a little lower to force him into B.






I can see now, what you were saying about C quartile players and such, with your tournament being sliced in two. This has been tried before but not in the same style, with lower div. players not playing higher div. ones. The problems with that system are the players in the lower half of the top draw are given a really raw deal regarding any chance at a prize. Hopefully you allow for this problem with rating category prizes.



Problem solved as you mused BD; we give them an excellent prize.




I'll concede the acc. swiss is an okish alternative for your bi-divisionary tournament yet there's no way you could seriously consider implementing it into a tournament without such divisions.



Have you got any clothes on BD?
You have conceded on every point and yet your conclusion is a conservatives gruff old "dammit, I am not conceding"


Hope you try again.
Do you want another cross-table?

starter

ursogr8
12-04-2004, 04:54 PM
Starter, I'm not an idiot, I can read that you said 800 points. At the time, my rating was horribly underrated (my point) in the low 1400's. Given this is a difference of 700-800, I'm sure you wouldn't be so picky to say 'oh but that's not above 800'

As for the whole 'travel' and 'only 1 or 2 games a day' I could make such a concession for circumstances like this. Specifically I do not want the accelerated Swiss in weekenders or one-day tournaments. For tournaments like Australian Juniors, I am more flexible on the subject.



I improved approximately 400 points, from low 1400's to 1829, from a performance rating of 2100+ in that tournament.



BD
In post #125 I asked how did you get to be rated at 14xx in the first place. This sounds like a fault that should be tracked down.
Would you like to respond.

Your wins in that tournament have caused a significant distortion to a few highly-rated players who value their ratings. They have been penalised significantly for the loss (to a 14xx player) when in fact the penalty should have been just for the loss to a 1850 player.

starter

Bill Gletsos
12-04-2004, 05:10 PM
BD
In post #125 I asked how did you get to be rated at 14xx in the first place. This sounds like a fault that should be tracked down.
Would you like to respond.

Your wins in that tournament have caused a significant distortion to a few highly-rated players who value their ratings. They have been penalised significantly for the loss (to a 14xx player) when in fact the penalty should have been just for the loss to a 1850 player.
There is nothing wrong with his rating. In fact he has shown no sign of being anything other than a 1500-1600 odd player in normal chess. As for whether he really is a much better rapid player is still to be determined as he has not played enough recent games to show if the result was anything more than a fluke.

His full stats are as follows:


Players Historical Rating Performance Normal
Period Rating Perf Score Games
Aug 2000 1374 0 7.5 16
Apr 2001 1470 1580 4.0 6
Aug 2001 1608 1747 7.0 12
Dec 2001 1611 1619 5.0 9
Apr 2002 1661 1769 10.5 13
Aug 2002 1626 1509 3.5 7
Mar 2003 1522 1162 3.0 6
Sep 2003 1507 1449 4.5 7
Nov 2003 1511 0 0.0 0
Mar 2004 1537 1295 3.5 6

Players Historical Rating Performance Rapid
Period Rating Perf Score Games
Aug 2000 1423 0 0.0 7
Aug 2001 1446 1500 6.0 10
Dec 2001 1458 1485 7.5 11
Mar 2004 1829 2188 4.5 7

Alan Shore
12-04-2004, 07:57 PM
Have you got any clothes on BD?
You have conceded on every point and yet your conclusion is a conservatives gruff old "dammit, I am not conceding"

You've got to me kidding me right? I'm being kind to you by saying I won't necessarily take such a hard line stance but I am not 'conceding' that you are right in al respects - in terms of a weekender, that has no divisions, I maintain your system is not the right one.





BD
In post #125 I asked how did you get to be rated at 14xx in the first place. This sounds like a fault that should be tracked down.
Would you like to respond.

Your wins in that tournament have caused a significant distortion to a few highly-rated players who value their ratings. They have been penalised significantly for the loss (to a 14xx player) when in fact the penalty should have been just for the loss to a 1850 player.

starter

As Bill showed, I was very inactive previously for rapid. I will assert I play rapid a lot stronger than standard but as you saw my standard rating was as high as (translated) 1731, a couple of years ago. An 1800-1900 performance might be called a fluke but 2188 is quite decisive.

Rincewind
12-04-2004, 08:05 PM
As Bill showed, I was very inactive previously for rapid. I will assert I play rapid a lot stronger than standard but as you saw my standard rating was as high as (translated) 1731, a couple of years ago. An 1800-1900 performance might be called a fluke but 2188 is quite decisive.

It makes sense to me that some people might play blitz weaker than normal chess due to slow calculations, nervious tension or whatever. However, it's hard to think of a reason why a 1500-1600 player would play rapid at 2100. There would seem to be too great a difference at the point in the scale where the difference is more to do with understanding, knowledge and technique, rather than raw calculation.

The best way to show it is not a fluke is to play more rapid and bring that RD down. :D

Garvinator
12-04-2004, 08:07 PM
The best way to show it is not a fluke is to play more rapid and bring that RD down. :D
or increase your normal rating ;)

Alan Shore
12-04-2004, 08:14 PM
Yes, the 98% was a provocative exaggeration.
There have been various posts earlier about survey's revealing player attitude to all of this. That would be one way of getting a better appreciation of the general consensus.


What do you suggest for a theoretical survey? My major point is the competition people enjoy against higher rated opponents which you've failed to adequately address. Since it's only been you against the current system the onus is on you to show me how many people are disgruntled with the current system. Good luck...

Alan Shore
12-04-2004, 08:22 PM
It makes sense to me that some people might play blitz weaker than normal chess due to slow calculations, nervious tension or whatever. However, it's hard to think of a reason why a 1500-1600 player would play rapid at 2100. There would seem to be too great a difference at the point in the scale where the difference is more to do with understanding, knowledge and technique, rather than raw calculation.

The best way to show it is not a fluke is to play more rapid and bring that RD down. :D

Barry, I have some problems with concentration in general, so I tend to lose my train of thought in longer games. I also play a lot of internet chess so I have had a large amount of practice in quicker games.

But anyway, sure, let's see how I go in the next rapid tourn (:

Garvinator
12-04-2004, 08:23 PM
What do you suggest for a theoretical survey? My major point is the competition people enjoy against higher rated opponents which you've failed to adequately address. Since it's only been you against the current system the onus is on you to show me how many people are disgruntled with the current system. Good luck...
:hmm:

Alan Shore
12-04-2004, 08:27 PM
:hmm:

You and Garvin then :p Or however many people you have onside that's ok. If you have sufficient numbers of loyal fans that will flock to your side then you can make a more determined case.

Rincewind
12-04-2004, 08:29 PM
or increase your normal rating ;)

In theory, increasing the nromal rating will not tell us anything about the fluke or otherwise of the 2100 result in the rapid. But I see your point. Have a more reliable standard and rapid rating will help to evaluate the difference between the two.

Bill Gletsos
12-04-2004, 09:17 PM
As Bill showed, I was very inactive previously for rapid. I will assert I play rapid a lot stronger than standard but as you saw my standard rating was as high as (translated) 1731, a couple of years ago. An 1800-1900 performance might be called a fluke but 2188 is quite decisive.
You goose.
Just because you got a 70 point uplift in 2004 does not mean that said uplift was valid back in April 2002.
In fact given that your rating peaked between Dec 2001 and August 2002 it is reasonable to suspect that the April 2002 rating was an anomaly.

As for 2188 being decisive, even if your rapid is stronger than your normal given the rapid 1829 is based on 7 games it is more likely the 2188 performance is a fluke.

Bill Gletsos
12-04-2004, 09:20 PM
It makes sense to me that some people might play blitz weaker than normal chess due to slow calculations, nervious tension or whatever. However, it's hard to think of a reason why a 1500-1600 player would play rapid at 2100. There would seem to be too great a difference at the point in the scale where the difference is more to do with understanding, knowledge and technique, rather than raw calculation.

The best way to show it is not a fluke is to play more rapid and bring that RD down. :D
Actually that 2188 performance is based on the Australian Allegro in SA.
I believe that it is played at 15 mins per game with lightning rules in effect.
Can anyone confirm the exact details.

ursogr8
12-04-2004, 09:54 PM
What do you suggest for a theoretical survey?


hi Bruce
I didn't say I was in favour of a survey. On the contrary, I am preferring to rely on the verbal feedback from our members, and the growth in our tournament entries as the evidence of whether players like the chess competitive, or junk.



My major point is the competition people enjoy against higher rated opponents which you've failed to adequately address. Since it's only been you against the current system the onus is on you to show me how many people are disgruntled with the current system. Good luck...


Yes BD, you have made it clear what you prefer. And Matt Sweeney is also on your side.

What does disappoint me is that I find your exceptional performances have been in Allegro tournaments, not in serious chess. As I am not in the business of trying to promote Allegros (but I am not against them either; I just think they are a snipers skill) then I think our debate is finished. Thanks for being engaged.

regards
starter

Alan Shore
13-04-2004, 12:56 AM
And rapid isn't considered chess? Poppycock. That's just pure snobbery. What a shame to end on that note.

Alan Shore
13-04-2004, 12:58 AM
Actually that 2188 performance is based on the Australian Allegro in SA.
I believe that it is played at 15 mins per game with lightning rules in effect.
Can anyone confirm the exact details.

It's a rapid tournament, we played rapid rules, not lightning. I'm befuddled as to why you even ask.

ursogr8
13-04-2004, 08:12 AM
And rapid isn't considered chess? Poppycock. That's just pure snobbery. What a shame to end on that note.

BD
I did not say that rapid was not chess. I just said that I don't go out of my way to organise ALLEGROS because I regard it as a snipers skill.
BTW, I am adding 'snob' to the names you have called me.
Misquoting me and calling me names does make me circumspect.

starter

Bill Gletsos
13-04-2004, 11:32 AM
It's a rapid tournament, we played rapid rules, not lightning. I'm befuddled as to why you even ask.
This clearly shows your lack of experience.
Allegro as a term had long existed before FIDE came up with the idea of having Rapid rules or even the long defunct "active" chess rules.

Allegro tournaments were of a 15 minute duration played under lightning rules.

Since the tournament was called the Australian Allegro it seemed a reasonable assumption they were playing under what was "Allegro" rules.

Alan Shore
13-04-2004, 12:08 PM
This clearly shows your lack of experience.
Allegro as a term had long existed before FIDE came up with the idea of having Rapid rules or even the long defunct "active" chess rules.

Allegro tournaments were of a 15 minute duration played under lightning rules.

Since the tournament was called the Australian Allegro it seemed a reasonable assumption they were playing under what was "Allegro" rules.

Well no.. illegal moves didn't lose and you could claim a draw under 10.2.

Alan Shore
13-04-2004, 12:12 PM
BD
I did not say that rapid was not chess. I just said that I don't go out of my way to organise ALLEGROS because I regard it as a snipers skill.
BTW, I am adding 'snob' to the names you have called me.
Misquoting me and calling me names does make me circumspect.

starter

If you say things like this:


...I just think they are a snipers skill) then I think our debate is finished.

Then you can only be referred to as such, discounting everything I said corresponding with rapid performances. So don't YOU try to turn it around.

ursogr8
13-04-2004, 12:43 PM
If you say things like this:



Then you can only be referred to as such, discounting everything I said corresponding with rapid performances. So don't YOU try to turn it around.

BD

In my post #179 the paragraph you took exception to was only three lines long. You could have quoted it in its entirety; but you chose to truncate it and change its sense. I don't intend to try to defend the truncated version.

Second, I didn't discount your arguments; what I said is that I didn't want to debate construction of Allegro tournaments. Personally I enjoy Allegros (and Lightning) and I enter these readily as tournaments. But I don't put my energies into designing and administering them. It requires different skills from normal chess.

starter

Alan Shore
13-04-2004, 12:57 PM
BD

In my post #179 the paragraph you took exception to was only three lines long. You could have quoted it in its entirety; but you chose to truncate it and change its sense. I don't intend to try to defend the truncated version.


I was merely pointing out the poignant part of the quote - in its entirety it's still the same message.



Second, I didn't discount your arguments; what I said is that I didn't want to debate construction of Allegro tournaments. Personally I enjoy Allegros (and Lightning) and I enter these readily as tournaments. But I don't put my energies into designing and administering them. It requires different skills from normal chess.

Chess is chess is chess and the construction of such tournaments are the same as the longer time controls as of now.

Yes it does require different skills, it requires you to think faster, know your position better at a glance without having aeons of time to calculate 10 moves in advance fritz-like variations and to utilise the resources you have, regarding time and position. To me it makes the game more exciting and more realistic in the sense of a battle. So anyone who devalues rapid chess against the interminable snail pace of a long time control hasn't got a clue.

ursogr8
13-04-2004, 01:24 PM
Chess is chess is chess and the construction of such tournaments are the same as the longer time controls as of now.



The construction of tournaments does differ according to the time-controls; contrary to what you assert. For example, in correspondence chess it is quite normal to see double round-robins; and I guess this is because players want to reduce the chance factors inherent in a SWISS (where you get to play a subset of players, and only one colour against each eventual opponent). But you would have to search hard and long to find a double round-robin Allegro event. The conclusion is inescapable; different constructions suit different time-controls, and suit different players.




Yes it does require different skills, it requires you to think faster, know your position better at a glance without having aeons of time to calculate 10 moves in advance fritz-like variations and to utilise the resources you have, regarding time and position. To me it makes the game more exciting and more realistic in the sense of a battle.


No argument from me on this one, Lightning certainly raises the adrenalin level.



So anyone who devalues rapid chess against the interminable snail pace of a long time control hasn't got a clue.


You have changed from discounts to devalues. Who did you have in mind when you wrote these words? All that I am saying personally is that I put my energies into organising longer time-controls. I have not discounted nor devalued. In fact, one of my fondest memories was beating jammo (not the banned one) in two successive weeks with the same move in a Lightning game. Given the difference in our ratings it is just a sniper's win; cherished nonetheless.


starter

Alan Shore
13-04-2004, 01:42 PM
The construction of tournaments does differ according to the time-controls; contrary to what you assert. For example, in correspondence chess it is quite normal to see double round-robins; and I guess this is because players want to reduce the chance factors inherent in a SWISS (where you get to play a subset of players, and only one colour against each eventual opponent). But you would have to search hard and long to find a double round-robin Allegro event. The conclusion is inescapable; different constructions suit different time-controls, and suit different players.

Now this is an extremely fastidious case of nitpicking, using correspondence chess as an example... in any case it's more often than not the number of players that determines a tournament to be changed to RR or DRR. The only instance I can relate to where time controls would consititute a DRR would be 1 0 tourns on the net.





You have changed from discounts to devalues. Who did you have in mind when you wrote these words? All that I am saying personally is that I put my energies into organising longer time-controls. I have not discounted nor devalued. In fact, one of my fondest memories was beating jammo (not the banned one) in two successive weeks with the same move in a Lightning game. Given the difference in our ratings it is just a sniper's win; cherished nonetheless.

Read what I said.. devaluing rapid games, and discounting my arguments because of them. Your argument was of the form:

1. Bruce used rapid games as part of his argument
2. Rapid games are snipers shots
_________________________
3. Therefore, I discount his evidence involving rapid games.


The devaluing is your second premise, the discounting is your conclusion.


P.S. Congrats on beating Jammo, nicely done.

ursogr8
13-04-2004, 04:28 PM
Now this is an extremely fastidious case of nitpicking, using correspondence chess as an example...



My example may have been inconvenient to you in that it disproved your contention that chess is chess is chess. As you have reverted to name-calling (not me this time, just my example) then I presume you have run out of arguments to back up your contention that the speed of play does not influence tournament design.




in any case it's more often than not the number of players that determines a tournament to be changed to RR or DRR.


In my experience, talking to correspondence players, they only admit enough players so that they can play a DRR. If they have more, then they construct another DRR. In other words, the value they place, on games and the tournament, is so high that they want to ensure equality of opportunity. This point drives the tournament design, not the number of players available.




Read what I said.. devaluing rapid games, and discounting my arguments because of them. Your argument was of the form:

1. Bruce used rapid games as part of his argument
2. Rapid games are snipers shots
_________________________
3. Therefore, I discount his evidence involving rapid games.


The devaluing is your second premise, the discounting is your conclusion.



You can re-word my intent as often as you like but I clearly am not discounting nor devaluing. If you want d-words…I am disengaging because I am disinterested in ALLEGRO tournament design.

The thread topic is “How competitive do you want it to be”. I would have thought the just completed Canberra tournament shows players are attracted when there is more than 1 Division. Three divisions in Canberra’s case. (Not forgetting the attractive prizes). As a counter, can you point to any Australian Monster single-division SWISS that is anywhere near as attractive?

starter

ursogr8
22-04-2004, 02:30 PM
Rounds 1 of the current Box Hill Championship has a calculated mean absolute deviation of rating differences, of >>>>> 394 .

Unrated-opponents games excluded.

The pairings and results can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round1.htm

starter

Bill Gletsos
22-04-2004, 02:39 PM
I still think you should calculate the Standard Deviation.

After all which is better.

1) A mean of 250 and an SD of 100
2) A mean of 275 and an SD of 75
3) A mean of 300 and an SD OF 50
4) A mean of 300 and an SD OF 25

ursogr8
22-04-2004, 02:45 PM
I still think you should calculate the Standard Deviation.

After all which is better.

1) A mean of 250 and an SD of 100
2) A mean of 275 and an SD of 75
3) A mean of 300 and an SD OF 50
4) A mean of 300 and an SD OF 25

Yes Bill. You said this once before. And I responded with my counter view. As you did not reply I presumed I had won the argument. I sent the results to the Moderator to have my BB rating recalculated and looked for another opponent.

But, as you still have time on your clock, we will resume, and I will calculate the SD, for the hang of it.

starter

Bill Gletsos
22-04-2004, 03:01 PM
Bill
OK, I go back to this post of yours and I say I agree (for the second time), it is a good thought of yours.
Now what do you actually do with the observation. You seem to have ruled out use as a marketing phrase.

Just how are you going to get the pairings with a 500:50 profile instead of 450:150?

starter
I was not suggesting you use it as a means of pairing necessarily.

You have been simply posting figures and suggesting that the smaller the number the better.
I was suggesting that a larger number might be better if it has a smaller SD.

ursogr8
22-04-2004, 08:55 PM
I still think you should calculate the Standard Deviation.

After all which is better.

1) A mean of 250 and an SD of 100
2) A mean of 275 and an SD of 75
3) A mean of 300 and an SD OF 50
4) A mean of 300 and an SD OF 25

hi Bill
Yeh, I remember this post of yours now.
First reaction by me is that you had played an ordinary move and then you had walked away from the board in disgust. That is why I thought you had given up.

For the sake of argument let us presume there is a chess player who is discerning enough to be able to judge between your 1) to 4). And further assume that if he was discerning enough to know the difference, that he is also sensitive enough to change his entry decision into a tournament based on which of 1) to 4) was being offered by the tournament organiser. And also let us presume that the organiser is able to engineer the field to meet any of 1) to 4).

I contend that a player might make his decision on the basis of
> enter if MAD + SD < 310 (say)
> not enter if MAD + SD > 310.

Under your 1) to 4), he enters 1) but not 2), 3), and 4). Now notice this is the same as if he simply sought out the lowest MAD; which is what I am proposing.

So your criteria for entry yielded (in this case) the same as mine. BUT, mine was simpler. Therefore it is clear what we should be tracking.

Thus, I withdraw my offer to calculate the SD because , well, just because...


starter

Kevin Bonham
22-04-2004, 09:28 PM
I was suggesting that a larger number might be better if it has a smaller SD.

For the fairness of the competition overall, yes.

For "competitiveness", not so clear cut. A higher SD for a high rating difference means a more competitive round on average than a lower SD. For instance if the RD on every board is 400 there will be a lot less upsets than if half the RDs are 200 and half are 600.

Similarly, assuming the players care about the question at all, what would bother them more - the possibility of a massive mismatch or the certainty of a merely large one?

ursogr8
26-04-2004, 09:49 AM
Rounds 1 of the current Box Hill Championship has a calculated mean absolute deviation of rating differences, of >>>>> 394 .

Unrated-opponents games excluded.

The pairings and results can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round1.htm

starter

In round 2 the competitive metric tightened to 280 .

(And Bill, the SD is 74)

starter

ursogr8
30-04-2004, 08:49 AM
Rounds 1, 2, and 3 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>> 394, 280, 240 .

Unrated-opponents games excluded.

The pairings for round 3 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round3.htm

starter

ursogr8
07-05-2004, 11:20 AM
Rounds 1, 2, 3, 4 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>>

394, 280, 240, 204 .

Unrated-opponents games excluded.

The pairings for round 4 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round4.htm

starter

Alan Shore
07-05-2004, 05:22 PM
Whee.

ursogr8
14-05-2004, 08:07 AM
Rounds 1, 2, 3, 4, 5 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>>

394, 280, 240, 204, 242 .

Unrated-opponents games excluded.

The pairings for round 5 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round5.htm

starter

ursogr8
18-05-2004, 08:29 AM
Rounds 1, 2, 3, 4, 5, 6 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>>

394, 280, 240, 204, 242, 248 .

Unrated-opponents games excluded.

The pairings for round 6 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round6.htm
starter

ursogr8
28-05-2004, 08:14 AM
Rounds 1, 2, 3, 4, 5, 6, 7 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>>

394, 280, 240, 204, 242, 248, 301 .

Unrated-opponents games excluded.

The pairings for round 7 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round7.htm

starter

ursogr8
01-06-2004, 12:34 PM
why can't a new player's starting rating be the average rating of all established rated australian players of the same age or age group including those on the master list.



or given he always excludes unrated games in his calculations would this ruin starter's junk round eliminator differential metric?


eclectic

hi ecelctic

If two paired players were unrated then your suggestion would see them both given a number like 1408.
The differential of the ratings is 1408-1408=0.
The change in the competitiveness metric would be from (for example)
40,000/90=444
to
40,000/92=434

No. The metric is not ruined.

starter

ursogr8
04-06-2004, 08:15 AM
Rounds 1, 2, 3, 4, 5, 6, 7, 8 of the current Box Hill Championship have a calculated mean absolute deviation of rating differences, of >>>>>

394, 280, 240, 204, 242, 248, 301, 277 .

Unrated-opponents games excluded.

The pairings for round 8 can be seen at
http://www.boxhillchess.org.au/e2004/e0404fcc/round8.htm

starter

Alan Shore
04-06-2004, 10:52 AM
Do you play to be a winner, or a competitor?

Since the goal in chess is to win, a winner!

ursogr8
04-06-2004, 12:07 PM
Since the goal in chess is to win, a winner!

Aha BD.

Is the next logical conclusion then 'should such a player prefer to play in a 7 round SWISS where he/she had a good chance of winning all 7 rounds; or a 7 round SWISS where the first round or two are junk rounds with little chance of winning (of course if he/she is in the top 10% of the field then question I pose is not relevant).

starter

Alan Shore
04-06-2004, 09:26 PM
Aha BD.

Is the next logical conclusion then 'should such a player prefer to play in a 7 round SWISS where he/she had a good chance of winning all 7 rounds; or a 7 round SWISS where the first round or two are junk rounds with little chance of winning (of course if he/she is in the top 10% of the field then question I pose is not relevant).

starter

You should be able to infer from the past what my answer to that is!
Then again I don't even know if I'm that fussed any more :rolleyes:

ursogr8
15-06-2004, 02:57 PM
The 2004 Victorian OPEN
Competitive index calculations

Round 1
48 pairings
43 pairings involving two rated players.

Index calculated at 627.
Accepted junk round thresh-hold is 450.

Conclusion
Round 1 was predicted to be a junk round. :(

Actual data
Only one upset against the pairings

starter

ursogr8
15-06-2004, 05:11 PM
The 2004 Victorian OPEN
Competitive index calculations

Round 1 and 2
48 pairings
43 pairings involving two rated players.

Index calculated at 627, 387.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor

starter

Alan Shore
15-06-2004, 05:18 PM
Actual data
Only one upset against the pairings

starter

Maybe you should ask that one player if he would rather have acc. swiss :p

Congrats on the tourn though starter, looked like a success.

ursogr8
15-06-2004, 05:26 PM
Maybe you should ask that one player if he would rather have acc. swiss :p

Congrats on the tourn though starter, looked like a success.

BD
1 I am not in favour of accelerated SWISSES. And neither were the four people with a vote on the organising committee.
2 The one player who lost was very upset that there were two late entrants paired after the initial pairings at 11.15 am, and before the start of round 1 at noon.
3 Thanks for the :clap:
4 Had to be a success otherwise Amiel would have the wrong view of Melbourne. :rolleyes:


starter

ursogr8
16-06-2004, 08:08 AM
The 2004 Victorian OPEN
Competitive index calculations

Round 1, 2, and 3
48 pairings
43 pairings involving two rated players.

Index calculated at 627, 387, 381.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)
Round 3 was predicted to be a competitive round. :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor
Round 3 > Nine upsets against the ratings predictor

starter

ursogr8
16-06-2004, 12:23 PM
The 2004 Victorian OPEN
Competitive index calculations

Round 1, 2, 3 and 4
48 pairings
42 pairings involving two rated players.

Index calculated at 627, 387, 381, 384.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)
Round 3 was predicted to be a competitive round. :)
Round 4 was predicted to be a competitive round. :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor
Round 3 > Nine upsets against the ratings predictor
Round 4 > Eight upsets against the ratings predictor

starter

ursogr8
16-06-2004, 01:11 PM
The 2004 Victorian OPEN
Competitive index calculations

Round 1, 2, 3, 4 and 5
48 pairings
42 pairings involving two rated players.

Index calculated at 627, 387, 381, 384, 352.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)
Round 3 was predicted to be a competitive round. :)
Round 4 was predicted to be a competitive round. :)
Round 5 was predicted to be a competitive round. :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor
Round 3 > Nine upsets against the ratings predictor
Round 4 > Eight upsets against the ratings predictor
Round 5 > Eleven upsets against the ratings predictor

starter

Ian Rout
16-06-2004, 01:45 PM
What is an "upset"? The rating system exepcts you to score about 1/7 against a player 300 higher. Does scoring that one point constitute an upset of itself, or only if you do it a second time in seven games?

Beating a player 50 points higher would not normally be called an upset, but beating someone 500 higher probably is. Where is the crossover - is it higher for a draw, or the same point with a draw being a smaller upset. Or is there a gradation with a result being a "surprise" rather than an "upset" for a smaller rating difference.

Any thoughts?

ursogr8
16-06-2004, 02:11 PM
What is an "upset"? The rating system exepcts you to score about 1/7 against a player 300 higher. Does scoring that one point constitute an upset of itself, or only if you do it a second time in seven games?

Beating a player 50 points higher would not normally be called an upset, but beating someone 500 higher probably is. Where is the crossover - is it higher for a draw, or the same point with a draw being a smaller upset. Or is there a gradation with a result being a "surprise" rather than an "upset" for a smaller rating difference.

Any thoughts?

Ian
Thanks for the response. :)

I am calling an upset if the lower ranked player wins.
Thus, draws, and games where one is unrated, are excluded.

Just as we bench-marked the definition of a junk round as >450 MAD (after discussion amongst interested parties), so we also need to agree the benchmark for an upset. Or a surprise.
Happy to explore options that you seriously want to put forward. But remember, I am interested in measuring competitiveness more than outliers.

regards
starter

Alan Shore
16-06-2004, 04:47 PM
An upset is when your opponent gets upset.

........

Actually I found it strange you didn't include draws, you think a 1000 rated player drawing with a 2200 doesn't constitute an upset?

Kevin Bonham
16-06-2004, 04:53 PM
Just as we bench-marked the definition of a junk round as >450 MAD (after discussion amongst interested parties), so we also need to agree the benchmark for an upset.

Who's "we"? I don't remember signing that protocol.

*runs away*

Kevin Bonham
16-06-2004, 05:03 PM
Actually I found it strange you didn't include draws, you think a 1000 rated player drawing with a 2200 doesn't constitute an upset?

I agree. Even a draw with a 1000 point differential virtually never happens compared to a win for the lower rated player with a 100 point differential. I think of it more as points of upsets in a given round, so an upset draw counts as half a point of upsets.

Depending on the strength of the players there may be a point where a draw becomes the expected result between them and really isn't an "upset" anymore.

One problem with all this is that if you set limits "below X points difference isn't an upset" then you need to remove those pairings incapable of generating upsets from consideration.

It would be interesting to know, on average, at what rating difference A is so much stronger than B that there is a stat. sig. difference in playing strength between them. I think Elo originally set the class interval at 200 points but my feeling is that for players rated above 1500 it is actually a lot lower than that. At the top end of the tree it may be as low as 50 pts.

Rincewind
16-06-2004, 09:27 PM
Can I suggest the following definition for an upset?

Beating someone rated more than 100 point higher, Drawing with someone more than 400 points higher.

Draws make the calculations tricky but the expected result in both cases is pretty close to 35% and 10% which are both 15% below the median between a loss and the result in question. Don't know if this is statistically valid but it sounds about right. :)

The statistical way to go would be to consider the actual less expected for all games round by round. A calculation of the mean and varience might be interesting following this method.

ursogr8
16-06-2004, 10:26 PM
An upset is when your opponent gets upset.




A nice thought, but hard to turn into a metric.






Actually I found it strange you didn't include draws, you think a 1000 rated player drawing with a 2200 doesn't constitute an upset?


I chose a provocative simple definition to leave the better ideas to other posters as this thread was not attracting its due attention.
Your suggestion has merit.



I agree. Even a draw with a 1000 point differential virtually never happens compared to a win for the lower rated player with a 100 point differential. I think of it more as points of upsets in a given round, so an upset draw counts as half a point of upsets.


The definition of an upset is certainly negotiable.





Depending on the strength of the players there may be a point where a draw becomes the expected result between them and really isn't an "upset" anymore.

One problem with all this is that if you set limits "below X points difference isn't an upset" then you need to remove those pairings incapable of generating upsets from consideration.



And this makes the metric cumbersome to calculate.
Hence my simple choice.





It would be interesting to know, on average, at what rating difference A is so much stronger than B that there is a stat. sig. difference in playing strength between them. I think Elo originally set the class interval at 200 points but my feeling is that for players rated above 1500 it is actually a lot lower than that. At the top end of the tree it may be as low as 50 pts.



So, KB, have a go at defining and upset.




Who's "we"? I don't remember signing that protocol.

*runs away*


Well, you didn’t disagree with the proposed 450 benchmark. And I have rarely known you not have a (good) opinion on mostmatters.



Can I suggest the following definition for an upset?

Beating someone rated more than 100 point higher, Drawing with someone more than 400 points higher.

Draws make the calculations tricky but the expected result in both cases is pretty close to 35% and 10% which are both 15% below the median between a loss and the result in question. Don't know if this is statistically valid but it sounds about right. :)

The statistical way to go would be to consider the actual less expected for all games round by round. A calculation of the mean and varience might be interesting following this method.


Baz
Thanks for becoming engaged.
I will pursue your suggestion after I have finished calculating with the currently adopted definition of an upset.

starter

ursogr8
16-06-2004, 10:54 PM
The statistical way to go would be to consider the actual less expected for all games round by round. A calculation of the mean and varience might be interesting following this method.

Baz

Can you explain what you mean by 'actual less expected'.
Is it for a game?
Is it for a round?

starter

Rincewind
16-06-2004, 11:26 PM
Can you explain what you mean by 'actual less expected'.
Is it for a game?
Is it for a round?

The calculation is game-by game.

You have the actual result: 1 for White win, 0.5 for draw and 0 for Black win
The expected is (10^(R1/400))/(10^(R1/400)+10^(R2/400))

Where R1 is white's rating and R2 is black's rating.

Then you calculate the mean and variance round by round to see if there is a correlation with mean rating difference. I don't know what to expect, I guess any interaction might be interesting.

ursogr8
17-06-2004, 08:10 AM
The calculation is game-by game.

You have the actual result: 1 for White win, 0.5 for draw and 0 for Black win
The expected is (10^(R1/400))/(10^(R1/400)+10^(R2/400))

Where R1 is white's rating and R2 is black's rating.

Then you calculate the mean and variance round by round to see if there is a correlation with mean rating difference. I don't know what to expect, I guess any interaction might be interesting.

Baz
I am closer to understanding what you mean now.
In essence you are introducing a graded measure of the 'upset' rather than my 1,0 approach...I think.

Now, yours takes a lot more to calculate than mine.
So, first, as an 'economic mathematician' I have to ask myself will your metric show a different mean from mine. If yours has the same expectation/mean as my approach then it will be very debatable whether the extra effort is justified.

But just before I start, let me validate my understanding of your formula. Could you give me the outcome for
R1 = 1000 R2 = 1500
and for
R1 = 1600 R2 = 2100
and for
R1 = 2100 R2 =1600.

BTW, since you are being so fine-tuned why did you not introduce the accepted % advantage for white factor into the calculation? While Bill leaves it out of the rating calculation (for reasonable reasons), it surely would have been consistent with your objective?

starter

ursogr8
17-06-2004, 08:44 AM
The 2004 Victorian OPEN
Competitive index calculations

Round 1, 2, 3, 4, 5 and 6
48 pairings
42 pairings involving two rated players.

Index calculated at 627, 387, 381, 384, 352, 328.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)
Round 3 was predicted to be a competitive round. :)
Round 4 was predicted to be a competitive round. :)
Round 5 was predicted to be a competitive round. :)
Round 6 was predicted to be a competitive round. :) :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor
Round 3 > Nine upsets against the ratings predictor
Round 4 > Eight upsets against the ratings predictor
Round 5 > Eleven upsets against the ratings predictor
Round 6 > Nine upsets against the ratings predictor

starter

ursogr8
17-06-2004, 09:20 AM
The 2004 Victorian OPEN
Competitive index calculations

Round 1, 2, 3, 4, 5, 6 and 7
48 pairings
42 pairings involving two rated players.

Index calculated at 627, 387, 381, 384, 352, 328, 341.
Accepted junk round thresh-hold is 450.

Conclusions
Round 1 was predicted to be a junk round. :(
Round 2 was predicted to be a competitive round. :)
Round 3 was predicted to be a competitive round. :)
Round 4 was predicted to be a competitive round. :)
Round 5 was predicted to be a competitive round. :) :)
Round 6 was predicted to be a competitive round. :) :)
Round 7 was predicted to be a competitive round. :) :)

Actual data
Round 1 > Only one upset against the ratings predictor
Round 2 > Six upsets against the ratings predictor
Round 3 > Nine upsets against the ratings predictor
Round 4 > Eight upsets against the ratings predictor
Round 5 > Eleven upsets against the ratings predictor
Round 6 > Nine upsets against the ratings predictor
Round 7 > Four upsets against the ratings predictor

starter

Rincewind
17-06-2004, 11:11 AM
I am closer to understanding what you mean now.
In essence you are introducing a graded measure of the 'upset' rather than my 1,0 approach...I think.

Just looking at every result as an upset to some degree and looking at the spread of the upsets. In theory most upsets would be minor and the major ones should be more rare. The degree of variation might be interesting.


Now, yours takes a lot more to calculate than mine.
So, first, as an 'economic mathematician' I have to ask myself will your metric show a different mean from mine. If yours has the same expectation/mean as my approach then it will be very debatable whether the extra effort is justified.

It might take more time but should not take that much more. If you want to mail me a standard SP crosstable it usually does take too long to use Excel to extract all the relevant info, cross match to opponent's ratings and then perform the calc. You need to do this with your method anyway since you calculate the mean rating difference to classify the round and look at all the results to identify upsets.


But just before I start, let me validate my understanding of your formula. Could you give me the outcome for
R1 = 1000 R2 = 1500
and for
R1 = 1600 R2 = 2100
and for
R1 = 2100 R2 =1600.

There is a |500| rating difference in all three examples so perhaps you could have picked better examples. Yours are as follows.

R1=1000; R2=1500 => E=0.0532402152020224
R1=1600; R2=2100 => E=0.0532402152020224
R1=2100; R2=1600 => E=0.946759784797978

Some other variations are

R1=1100; R2=1500 => E=0.0909090909090908
R1=1200; R2=1500 => E=0.150979557211323
R1=1300; R2=1500 => E=0.240253073352042
R1=1400; R2=1500 => E=0.359935000197115
R1=1500; R2=1500 => E=0.5
R1=1600; R2=1500 => E=0.640064999802885
R1=1700; R2=1500 => E=0.759746926647958
R1=1800; R2=1500 => E=0.849020442788677
R1=1900; R2=1500 => E=0.909090909090909

NB E(R1,R2) = 1 - E(R2,R1) and 0 < E < 1 for all real R1 and R2


BTW, since you are being so fine-tuned why did you not introduce the accepted % advantage for white factor into the calculation? While Bill leaves it out of the rating calculation (for reasonable reasons), it surely would have been consistent with your objective?

The same reasons apply here for formula is the basic logisitic ratings calc (Elo).

Rincewind
17-06-2004, 11:47 AM
BTW if the orginal form I game is complicated the standard math form would be

E = 1/(1+EXP(-(R1-R2)/173.7177927613))

Where EXP(x) is the standard natural exponentiation function (e^x)

173.717... = 400 / LN(10)

LN(x) is the natural logarithm function (ie the inverse of EXP).

This is the form that a mathematician might like, but most rating articles I've seen quote the form I posted originally.

ursogr8
17-06-2004, 12:40 PM
It might take more time but should not take that much more. If you want to mail me a standard SP crosstable it usually does take too long to use Excel to extract all the relevant info, cross match to opponent's ratings and then perform the calc. You need to do this with your method anyway since you calculate the mean rating difference to classify the round and look at all the results to identify upsets.



.

Yes, I yield on this point.
I was misjudged complexity-of-the-formula with the effort-to-extract. I now agree they require much the same effort.

Not sure I can handle the complexity and export though. Will think about doing so. As well as an 'economical mathematician' I am a 'lazy retired mathematician'. :uhoh:

ursogr8
17-06-2004, 12:45 PM
The same reasons apply here for formula is the basic logisitic ratings calc (Elo).

Baz

I invite you to reconsider. If it worth introducing the granularity that is different from my 1,0 approach then it is surely worth going the whole hog and getting the white-black distortion factored into the calculations.
While I agree that the size of the %-distortion may be debatable, that debate is no reason not to work for its inclusion.

starter

ursogr8
17-06-2004, 12:46 PM
BTW if the orginal form I game is complicated the standard math form would be

E = 1/(1+EXP(-(R1-R2)/173.7177927613))

Where EXP(x) is the standard natural exponentiation function (e^x)

173.717... = 400 / LN(10)

LN(x) is the natural logarithm function (ie the inverse of EXP).

This is the form that a mathematician might like, but most rating articles I've seen quote the form I posted originally.

I think I will park this for later consideration.

ursogr8
17-06-2004, 01:03 PM
There is a |500| rating difference in all three examples so perhaps you could have picked better examples. Yours are as follows.

R1=1000; R2=1500 => E=0.0532402152020224
R1=1600; R2=2100 => E=0.0532402152020224
R1=2100; R2=1600 => E=0.946759784797978



Baz

Actually, I only needed one example.


But just in case I was misreading the formula I constructed a litmus test for my likely misreading. So, the |500| rating difference was deliberate and appropriate.

starter

ursogr8
17-06-2004, 01:31 PM
It might take more time but should not take that much more. If you want to mail me a standard SP crosstable it usually does take too long to use Excel to extract all the relevant info, cross match to opponent's ratings and then perform the calc

Baz

I work from the data at
http://www.boxhillchess.org.au/vicchess/04VicOpenres.htm
which has been exported from SP. After import (copy and paste actually) to EXCEL I have to go Data, text to columns, change the row height, and then format to numbers instead of text in the ratings column.

Bill Gletsos
17-06-2004, 01:42 PM
The calculation is game-by game.

You have the actual result: 1 for White win, 0.5 for draw and 0 for Black win
The expected is (10^(R1/400))/(10^(R1/400)+10^(R2/400))

Where R1 is white's rating and R2 is black's rating.

An easier formula for people to calculate might be:
1/(1+ 10^(-D/400))
where D = the difference in rating

So to get Whites expected score use D = R1-R2
and to get Blacks use D = R2-R1

Rincewind
17-06-2004, 03:41 PM
I work from the data at
http://www.boxhillchess.org.au/vicchess/04VicOpenres.htm
which has been exported from SP. After import (copy and paste actually) to EXCEL I have to go Data, text to columns, change the row height, and then format to numbers instead of text in the ratings column.

Very interesting. I do this too but get slightly different answers to you for some rounds.

For example the mean rating difference I see is

1 => 627.0930233 match
2 => 387.5813953 close enough
3 => 369.9268293 11 point discrepancy
4 => 392.452381 8 point discrepancy
5 => 352.9069767 close enough
6 => 328.8571429 close enough
7 => 333.3953488 8 point discrepancy

The number of upsets using your original definition by my calculations are

1 => 1 match
2 => 6 match
3 => 10 discrepancy
4 => 8 match
5 => 12 discrepancy
6 => 10 discrepancy
7 => 4 match

So we seem to be treating the data slightly differently. My ten upsets from Round 3 were

Lin-Campara
Lycett-Morris
Jia-Braham
Belletty-Stanisheff
Voon-Maguire
Fletcher-Kable
Elliott-Frangakis
Lugo-Hopf
Fraser-Hamilton
Paterson-Lee

Can you compare this to your list? I expect you have nine of these and I have an extra one for some reason.

Rincewind
17-06-2004, 07:46 PM
I invite you to reconsider. If it worth introducing the granularity that is different from my 1,0 approach then it is surely worth going the whole hog and getting the white-black distortion factored into the calculations.
While I agree that the size of the %-distortion may be debatable, that debate is no reason not to work for its inclusion.

The problem is there is no universally accepted way I know of to incorporate the colour allocation. Also it doesn't matter as it will affect the mean. So if there is no colour advantage we would expect a Actual - Expected to have a mean of 0. Colour bias would move the Actual - Expected +ve if white had advantage and -ve if Black had the advantage.

The other practical issue is you cannot normally determine the colour allocatoin from the SP crosstable for a Swiss event. That being the case I even have to modify my method as we can't use white's perspective if we don;t know who white is.

Therefore what I propose is an analysis of |Actual - Expected|. This can be done by taking every players results (therefore double counting every game). When doing this I believe one would have to normalise the standard deviation by dividing it by sqrt(2). Can anyone confirm or deny that this step is required?

ursogr8
18-06-2004, 08:12 AM
Very interesting. I do this too but get slightly different answers to you for some rounds.



Can you compare this to your list? I expect you have nine of these and I have an extra one for some reason.

Baz

Hmm
Took me a while to work out the discrepancy.
I was using a slightly different path and introduced a criteria BECAUSE I WAS TOO CLOSE TO THE DATA. (The oldest statistician's error in the book).

So now I agree with your upset count.
Can you now apply your formula. Or Bill's version. Or your formula with a colour mod.

starter

ursogr8
18-06-2004, 08:32 AM
The other practical issue is you cannot normally determine the colour allocatoin from the SP crosstable for a Swiss event. That being the case I even have to modify my method as we can't use white's perspective if we don;t know who white is.




Baz

If we worked from the original SP cross-table, instead of Phillip's web-posted cross-table on the BH web-site, the colour allocation is observable (I think) because the print comes out in bold for white. I say, I think' because I don't have SP and I don't have a printed table available; just working from memory.
If the W was embolded, and if you had the original SP file, would EXCEL be able to recognize (filter on) this when you do your calculation?

starter

ursogr8
18-06-2004, 10:06 AM
The problem is there is no universally accepted way I know of to incorporate the colour allocation.


Baz
Not sure what you mean.
When we previously explored % bias from having the advantage of the white pieces (and it would have been on a ratings thread probably) there was a number quoted; 57.5%, or something like that.
Now go back to the sample I gave you earlier >
R1=1000; R2=1500 => E=0.0532402152020224

I think you are asking how to fold the 57.5% into the formula that calculated the E=0.0532402152020224?

And then you are perhaps asking ...'is the 57.5% really applicable to all rating differences, or is it a function of the rating difference'.

Dunno, mate. This is one reason I went with my 1,0 approach that you are granulating/pulverising/refining. Over to you.




Also it doesn't matter as it will affect the mean.



And this is your comment that I really don’t understand. I think I might wait until you have addressed the above and I might see more.


starter

Rincewind
18-06-2004, 01:31 PM
If we worked from the original SP cross-table, instead of Phillip's web-posted cross-table on the BH web-site, the colour allocation is observable (I think) because the print comes out in bold for white. I say, I think' because I don't have SP and I don't have a printed table available; just working from memory.
If the W was embolded, and if you had the original SP file, would EXCEL be able to recognize (filter on) this when you do your calculation?

I;ve got a better idea. Just use the pairing lists that appear below the crosstable on the website. :owned:

ursogr8
18-06-2004, 01:43 PM
I;ve got a better idea. Just use the pairing lists that appear below the crosstable on the website. :owned:

Barry Cox
That is just what I said in post #234.

Re-set switches and let us go back to square 1.

Can you calculate and include the white % bias?

starter

Rincewind
18-06-2004, 01:51 PM
And this is your comment that I really don’t understand. I think I might wait until you have addressed the above and I might see more.

I've analysed the data taking colour into account and you will see an interesting trend in the colour advantage round by round. It might be coincidence or have something to do with particular individuals alternating colours across rounds but it is intersting nonetheless.

Anyway here are my results...


No Mean StdDev Up RDiff
Round 1 43 -0.0177 0.1751 1 627
Round 2 43 0.1071 0.3381 6 388
Round 3 41 0.0048 0.4179 10 370
Round 4 42 0.1466 0.3756 8 392
Round 5 43 0.0618 0.4553 12 353
Round 6 42 0.1412 0.3845 10 329
Round 7 43 -0.0524 0.3277 4 333

No = Number of games with a result and both players rated
Mean = mean of Average - Expected from White's POV
StdDev = Standard Deviation of the same
Up = Upsets (Number of games where Abs(Act - Exp) > 0.5
RDiff = Mean Rating Difference

One intersting feature is the increase in the StdDev column for rounds 2-7 while for Round 1 this was much lower. This is what I would expect. The more competitive the round the higher the deviation as the results would be spread more widely from the expected.

Unexpected interesting feature is the Mean value oscillating from round to round. A number near zero for the first round is to be expected as the advantage of colour would not show itself in the junk round. However, reason black did so well in Rounds 3 and 7 is less clear. Round 5 seems to be an odd man out, not falling neatly into either category.

The overall mean for the tournament is a 0.0556 advantage to White.

Kevin Bonham
18-06-2004, 03:07 PM
Unexpected interesting feature is the Mean value oscillating from round to round. A number near zero for the first round is to be expected as the advantage of colour would not show itself in the junk round. However, reason black did so well in Rounds 3 and 7 is less clear. Round 5 seems to be an odd man out, not falling neatly into either category.

The overall mean for the tournament is a 0.0556 advantage to White.

I would be interested to see if this effect showed up in other large draws as well, and if it did so with the same colour pattern or not. If another data set shows the same thing I'll post some speculations.

Rincewind
18-06-2004, 03:55 PM
I would be interested to see if this effect showed up in other large draws as well, and if it did so with the same colour pattern or not. If another data set shows the same thing I'll post some speculations.

I'm in the process of requesting the data from the NSW Open. I'll publish the results here if and when I do. I suspect it is just an anomoly but more data is needed, even to draw any conclusion from the near zero result in round 1.

It might also be interesting to do some analysis based on the actual values of both rating to see whether White is (in general) more or less of an advantage for 1000, 1500, 2000 or 2500 rated players.

Garvinator
18-06-2004, 04:24 PM
I'm in the process of requesting the data from the NSW Open. I'll publish the results here if and when I do. I suspect it is just an anomoly but more data is needed, even to draw any conclusion from the near zero result in round 1.

It might also be interesting to do some analysis based on the actual values of both rating to see whether White is (in general) more or less of an advantage for 1000, 1500, 2000 or 2500 rated players.
what information do you require barry, i have sp files for the nsw open, womens and u1600 tournaments.

Rincewind
18-06-2004, 04:41 PM
what information do you require barry, i have sp files for the nsw open, womens and u1600 tournaments.

Garvin, Could you please send me either the SP files or else the Round by Round pairings with results as text for the Open and u1600. Before anyone asks, I'm not sexist but the Women's was a round robin and I'm interested in Swiss Paired tournies.

Bill, please ignore my PM. ;)

Garvinator
18-06-2004, 04:44 PM
Garvin, Could you please send me either the SP files or else the Round by Round pairings with results as text for the Open and u1600. Before anyone asks, I'm not sexist but the Women's was a round robin and I'm interested in Swiss Paired tournies.

Bill, please ignore my PM. ;)
is this thread of any use to you? : http://chesschat.org/showthread.php?t=666&page=3&pp=15

scroll down to post 42

Rincewind
18-06-2004, 04:51 PM
is this thread of any use to you? : http://chesschat.org/showthread.php?t=666&page=3&pp=15

scroll down to post 42

No, don't think so. Looks like only the crosstables have been posted.

Garvinator
18-06-2004, 04:54 PM
No, don't think so. Looks like only the crosstables have been posted.
have received your email barry and replied with zipped sp files