PDA

View Full Version : Accelerated vs Abominated Swiss Pairings :)



Kevin Bonham
05-02-2010, 12:08 AM
A common form of accelerated Swiss pairing tournament seen in Australia goes like this:

1. Put the top half of players in one group and the rest in another.
2. Add one notional point to the top half for the purposes of doing the draw for the first two rounds only. Thus the top-halfers are paired as if they are on 1/0, and a top-halfer is paired as if on 2, 1.5 or 1 out of 1 for round 2 depending on whether they won, drew or lost round 1.
3. Remove the notional point after round 2, so that for round 3 those who have had two wins are in the same scoregroup, no matter who they defeated or where they came from in the draw.

This system is known to Wikipedia as "Accelerated Pairings" because Wikipedia's chess sections are for the most part written by Americans. The correct name of this system (and its variants) shall henceforth be Abominated Pairings and any event in which it occurs shall be an Abominated Swiss, especially if it is directed by an IA, since one would hope IAs would know much better.

A much better system of Accelerated Pairings that has been used in the UK (although I am unsure what the current situation is with the BCF acceleration rules) is discussed in Reuben's books (at least the first edition thereof). In comparison to Abominated Pairings:

1. The first round is effectively the same (top quarter vs second quarter, third quarter vs fourth quarter).

2. There is no notional added point. Instead, those bottom half players who are on perfect scores after a given round are paired against top half players who are not on perfect scores (listed in rating order irrespective of score provided they have not dropped >1 point), while top half players on perfect scores play each other. This is the only form of acceleration that exists (although for very long tournaments it is possible to apply similar principles to scoregroups just below 100% too) .

3. There is no fixed number of rounds after which the acceleration in step two is removed. Instead, it is removed once there are no bottom-half players left on a perfect score. Sometimes this is the case after round two, more often after round three, in rare cases it may take until round four to do the job. However, once the number of bottom-halfers on a perfect score is small, the difference between the draw and a normal Swiss draw for that round is not great, as only a few pairings are affected.

4. Just in case some upstart from the bottom half (perhaps an unrated or inaccurately rated player) keeps winning and winning and winning, the acceleration is never applied to the last two rounds.

To consider which of the systems does the job better we need to consider the objectives of acceleration. The first objective of acceleration (and the reason such systems were invented) is to cull the perfect score group as quickly as possible and thus increase the chance of an event with many players but few rounds having an outright winner rather than co-winners each on perfect scores. The second objective is to eliminate the vast mismatches seen especially in round 1 of a normal Swiss.

The Abominated Swiss is much less effective than the Accelerated Swiss at culling the 100% scoregroup, for two reasons:

Firstly, in the Abominated Swiss the bottom half winners in round 1 are then paired only against the top-half losers (and, if insufficient top-half losers, with each other, which is an even bigger problem). But allowing them to also be paired against the top-half players on 0.5 means that the bottom-half winners play a stronger list of players and therefore fewer of them are likely to make it to 2/2.

Secondly and more importantly, when the Abominated Swiss removes the acceleration after round 2, opportunities to cull the bottom-half 2/2s by feeding them to top-half 1.5/2s and 1/2s are lost. For instance, in one such event I looked at tonight, the acceleration was removed after round 2 with 14 players on 2/2 of which 5 were from the bottom half. After this round there were still five players on 3 (and it would have been 6 but one of the bottom-halfers very surprisingly drew across a c. 900 point mismatch). Under a proper accelerated system, the five bottom-halfers on 2/2 would be fed to the top end of the top-halfers on 1.5 and 1, most likely none of them winning. The remaining 2/2s could play each other (except for one floating down) and most likely the top score group would be culled to 2 or 3 players.

That said, all five leaders were removed from the 100% scoregroup in round 4 anyway, which just goes to show why acceleration to cull the lead group is frequently unnecessary anyway. There are widespread delusions that if your number of rounds is n and you have more than 2^n players then you need acceleration to get the 100% scoregroup culled by the event's end. In fact, there are often draws on the top boards.

Also the inadequate culling of the 2/2 scoregroup in the Abominated Swiss produces grotesquely unfair discrepancies in opponent strength, with some of the round 3 100%-group pairings very competitive games and others effective one-point byes.

As for the avoidance of mismatches, the Abominated Swiss does give much closer round 2 pairings among the .5/1 top-halfers, and also slightly closer round 2 pairings in the notional 1/1 scoregroup (top half losers vs bottom-half winners mostly). But because it is less efficient at culling the 100% scoregroup, it often leads to some massive mismatches on top boards in round 3 (c. 1000 points not at all uncommon) so the gain in mismatch avoidance compared to true acceleration is questionable.

It can also be argued that pairing a top-halfer who drew in round 1 with another in the same boat is perverse. The second-quarter player who drew in round 1 with a top-quarter player is paired with another top-quarter player. This is a tough round 2 pairing for both and it will most likely result in either: (i) the second-quarter player going to 0.5/2 while other second-quarter players who lost round 1 have easy wins and go to 1/2 or (ii) the top-quarter player drawing again and now being 1/2, and most likely getting a very big mismatch in round 3.

So why is the Abominated Swiss so common? Because it's a solution to a problem that is simple, elegant and wrong. Computer programmers who are not familiar with the intricacies of pairings debates (there being no specified FIDE acceleration system) find it easy to write programming modules that just add a point to the score then take it off after a certain number of rounds. SP offers only this as an acceleration pairing option, and I am not sure if any other program offers anything but this "bonus point" system.

It is worth noting, however, that in SP it is possible to change the round at which the accelerated pairings are removed through the tournament, and thus to run a hybrid Accelerated/Abominated system in which you continue the bonus point until the 100% scoregroup contains only top-half players. This is extremely easy to do, eliminates a lot of round 3 hassles with Abominated pairings, and it is therefore surprising that it is not done more often.

I realise that true acceleration is somewhat impractical if you have to do it by hand. But if those using SP-acceleration would at least keep the acceleration on in round 3 if there are still bottom half players on 2/2 then that would be significant progress.

This can be done in SP as follows: In Options: Tournament: Accelerated Pairings initially set "Stop acceleration AFTER round" to 2.

The instant a bottom-halfer on 1/1 wins round 2, change the 2 to a 3.

Is that too difficult? :lol:

Adamski
05-02-2010, 11:43 AM
A very well-argued case, Kevin. I agree with you. The current accelerated system you described is indeed an abomination. The second system sounds much better.

SHump
05-02-2010, 12:57 PM
Yes, and it sounds easy to implement, even with SP. If I ever get into that situation (so long as the players know in advance what acceleration will mean, as to opposed to abomination) and it is deemed appropriate that it be done (not that this is a foregone conclusion by any means), then it would be interesting to put it into practice, and to get feedback on it.

FM_Bill
08-02-2010, 07:35 PM
I am opposed to the first form of accelerated pairings.

I have been in a tournament when a 1400 player was leading with 4/4
while Ian Rogers only had 3.5/4. its essentially a handicap system. All sorts of silly things happen.

Neil Davis once made the point that if you get a notional point you should
get to keep it at the end of the tourney.

Haven't read through the second type yet.

William AS
08-02-2010, 08:16 PM
A very well-argued case, Kevin. I agree with you. The current accelerated system you described is indeed an abomination. The second system sounds much better.
Hear Hear :clap: :clap:

I also think that acceleration is often used when it is not needed and in those circumstances can actually increase the possibility of ties for first place.

Danko
01-03-2010, 05:08 PM
I am opposed to the first form of accelerated pairings.

I have been in a tournament when a 1400 player was leading with 4/4
while Ian Rogers only had 3.5/4. its essentially a handicap system. All sorts of silly things happen.

Neil Davis once made the point that if you get a notional point you should
get to keep it at the end of the tourney.

Haven't read through the second type yet.
I'm running a low level unofficial tournament at an elementary school and am sorting through all the nuances and variations in pairing theory. The kids range from barely knowing the rules of play to about 520 ELO. In order to have a more enjoyable experience for all, I tried to avoid the unpleasant extreme pairing mismatches and decided to use accelerated pairings. It seems most logical to not remove the acceleration point in this setting. Do you have any more information about what Neil Davis said about it and whether anyone uses this method?

Kevin Bonham
01-03-2010, 05:24 PM
It seems most logical to not remove the acceleration point in this setting.

The main risk there is that if someone who you seed in the bottom half is actually very good and just keeps on winning, they could reach a perfect final score without playing the tournament leaders. That is why it is normal to at worst remove the acceleration two rounds from the end. However if you have lots of rounds for the number of players then that situation gets sorted out.

Danko
04-03-2010, 10:27 PM
... they could reach a perfect final score without playing the tournament leaders.
But that's why it seems that keeping the point would be necessary, since even with an undefeated record, a player from the lower half would still be a point lower than a person with a perfect record from the upper half. I think that seems fair. A player from the lower half could still come in first, but it would require doing a point better than someone facing tougher competition in the upper half.

However, after having applied this idea to a school tournament that has now been through 4 rounds, I do have the uncomfortable situation of having to try to explain to the kids the logic behind giving the strongest players a bonus point. If we hadn't done the acceleration at all, the first round would have been such huge mismatches in our club that it would have been a waste of time to even play it. Even though we applied it to the first round, the mismatches were big enough (since there is such a wide range of ability in our club) that the results came out predictably with only one exception at the very bottom level. If acceleration were removed after round one, it turned out that the matchups were just about exactly the same blowout pairings as if we were playing that unaccelerated first round that I thought we had avoided. And so the acceleration bonus point was kept. Beyond that round, removing the point didn't seem fair to the players that had had to face tougher competition for the first two rounds compared with those that had an easier time in the lower half. But perhaps this was my mistake as a first time tournament organizer.

Since this was just an unofficial school tournament it gave room for some tinkering to try to make it a better experience for the kids, without hugely mismatched games and postponing the climactic pairing of the top two til the final round. Unfortunately it has been much harder than I expected and there doesn't seem to be a perfect solution to achieve the ideals without the downsides. Maybe avoiding a Swiss System would be better next time and going with an imperfect but easier to understand single elimination system.

Kevin Bonham
04-03-2010, 10:58 PM
But that's why it seems that keeping the point would be necessary, since even with an undefeated record, a player from the lower half would still be a point lower than a person with a perfect record from the upper half. I think that seems fair. A player from the lower half could still come in first, but it would require doing a point better than someone facing tougher competition in the upper half.

Yes - and that might be out of their hands, because they can't control it if a top-half player keeps winning and never plays them. Then the bottom-half player might complain that they had been denied their shot at the tournament win through no fault of their own just because their rating was poor. Admittedly, it is a pretty unusual scenario, since typically either the top half will sort themselves out, or any overperforming bottom-halfer will get sorted out somewhere.


However, after having applied this idea to a school tournament that has now been through 4 rounds, I do have the uncomfortable situation of having to try to explain to the kids the logic behind giving the strongest players a bonus point.

I sympathise with this because sometimes I've found that even very good ideas are a hard sell to children if there is anything about those ideas that smacks of potential subjectivity or special treatment. It's good that so many children have "got" the concept of fairness on such a basic level but they don't have the experience base to understand about cases where there can be good reasons for treating different skill levels differently.


If acceleration were removed after round one, it turned out that the matchups were just about exactly the same blowout pairings as if we were playing that unaccelerated first round that I thought we had avoided.

Yes. Under normal circumstances removing acceleration after round one is almost totally pointless since it just switches rounds 1 and 2. Acceleration only does things if you keep it on for at least two rounds.


Beyond that round, removing the point didn't seem fair to the players that had had to face tougher competition for the first two rounds compared with those that had an easier time in the lower half.

This is one of the negatives of an accelerated system - in the early rounds score and performance don't have much to do with each other in the middle of the pack. Typically a player from the second 1/8th of the field will play a player from the fourth 1/8th and a player from the first 1/8th, scoring 1/2. A player from the seventh 1/8th of the field will play a player from the fifth 1/8th and a player from the eighth 1/8th, also scoring 1/2. Thus the differences in strength of field met in the first two rounds in the same scoregroup are much greater than in a single swiss.

However, there is some compensation, because when the acceleration is taken off those who've had a hard run to a midfield score should get a very easy opponent while those who have had an easy run should get a hard opponent.


Unfortunately it has been much harder than I expected and there doesn't seem to be a perfect solution to achieve the ideals without the downsides. Maybe avoiding a Swiss System would be better next time and going with an imperfect but easier to understand single elimination system.

Divisions are one alternative but then you have to explain how some players are in one division and some in another. Eliminations have the disadvantage that not everybody plays all games.

Some of these problems don't have perfect solutions and everything you do has disadvantages of some kind. Are the stronger players typically much older in your case or is it an even-age group?

Danko
07-03-2010, 03:50 AM
Are the stronger players typically much older in your case or is it an even-age group?
Our club is made up of 24 kindergarten thru fifth graders. There is a full range of ability from just starting out to fairly well-ranked tournament players. Half the club is made up of third graders who also span the whole range of abilities. By my intraclub rating assessments, the number 1 and 3 players are third graders and the 2, 4, and 5 players are fifth graders. We have a first grader at about 8th. I think that the wide-ranging abilities seem to need a different arrangement for a tournament than would a group of players of similar ability. In a big tournament they are simply separated as different divisions, but within a small club it didn't seem right. The system used last year divided by grade level, but I thought that since the levels didn't match the grades, that would not allow kids of similar abilities (irrespective of grade level) to face each other.

Now that we are through the fourth round with a couple surprises, there is a chance we will have a three way tie for first place after the final round. So now I have to consider the tiebreak options. In the event of a tie, I hope we can just call them all co-champions and make them all happy, but I'm afraid the school will probably want a single winner to place on a plaque. If we have to have one single winner, I'm inclined to have a round robin playoff between the last three, rather than leaving it up to tiebreak points which seem to leave too much to luck.

Kevin Bonham
07-03-2010, 10:59 PM
If we have to have one single winner, I'm inclined to have a round robin playoff between the last three, rather than leaving it up to tiebreak points which seem to leave too much to luck.

And which leave someone feeling left out too. In the Aus Juniors when there is a more than two way tie, tiebreaks are often used to cut the tied players to two, then they play a match, but at inexperienced junior level it is much harder for players to understand that sort of thing.

Garvinator
29-01-2011, 03:32 AM
Been looking further at accelerated pairings for the tournament this weekend.

Something I have picked up which is a concern is to do with floats. With those bottom halfers on 1/1 being fed to top halfers on less than 100% in round two, this increases all the number of players who will have upfloated or downfloated in round two, meaning that for round three, it will be harder to make decent pairings.

This could be even worse in round 4 if the acceleration is removed after round three because almost all players will have a strong colour preference ie plus 1 or minus 1.

This could mean an arbiter needs to go quite a way down a score group to find a pairing that meets the dutch pairing rules.

Perhaps in acceleration tournaments, the floating rules should be ignored :hmm:

Kevin Bonham
29-01-2011, 12:54 PM
Perhaps in acceleration tournaments, the floating rules should be ignored :hmm:

Yes, for an Acclerated Swiss, a score difference that is forced by the acceleration should not be regarded as a float. (There may still be some floats that are not forced by the acceleration.) However, I'm not sure how computer programs that employ "acceleration" go with this. Even if you do accelerated pairings manually and then go back to computer pairings once the acceleration is taken off, it may still regard point-mismatches from previous rounds as floats. To avoid this it may be necessary to pair the whole accelerated tournament by hand, which is OK if you have only one round a day but a bit of a pain otherwise.

I have just tested SP's "acceleration" and found that it classifies a top halfer on zero playing a bottom halfer on one in round 2 as a float. This is incorrect.

Garvinator
29-01-2011, 03:09 PM
I think the issue of identifying if a person is a genuine floater or an accelerated floater could be one of the main issues for pairing programs in doing accelerated pairings.

In round two there are couple of 0.5 v 0 pairings, where the 0.5's drew against each other in round one (top half pairing).

So in my opinion the two 0.5 ers are downfloaters as they playing someone on 0 who are also top halfers.

Probably the best way to describe what to do is:

1) While acceleration is being applied or is a factor, if two players are playing each other and they are from opposite half of the draw, then floats are not applied.
2) If they are from the same half of the draw, then float rules do apply.

Garvinator
29-01-2011, 03:16 PM
I am not sure if it is even possible in any pairing program to manually change someone's float status. That would be a handy feature for acceleration.

Arbiter works who were the genuine floaters for a particular round, adjusts the float status of the required players and then pairs for next round.

Kevin Bonham
29-01-2011, 03:19 PM
In round two there are couple of 0.5 v 0 pairings, where the 0.5's drew against each other in round one (top half pairing).

So in my opinion the two 0.5 ers are downfloaters as they playing someone on 0 who are also top halfers.

Correct.

Garvinator
29-01-2011, 05:50 PM
Acceleration for round three was not as bad as I thought. I could see cases for why sp paired as being played, so left the draw as generated with limited time to make changes.

Now having had a few minutes to do some checking, I only get a couple of differences.

The two bottom half players are on boards 2 and 3 because of draws and losses in the top half. I would have paired 3 and 9, but as most of the other pairings are still the same, it does not appear to make much difference.

Back to normal pairings tomorrow for rounds 4-7.

Kevin Bonham
15-03-2011, 10:31 PM
Garvinator mentioned this thread in the context of comments about the pairings for the 2011 Begonia where the same simplified two round system that I disparaged above was again used.

Some impacts in this year's event included:

* When the "acceleration" was removed in round 3 this created a scoregroup of 20 players of whom five were bottom-halfers. Of the five pairings where both players were top halfers, three were drawn, but all the mismatches were won by the stronger player. This meant that after round 3 the acceleration had culled the lead group to seven players, but had proper acceleration been used (including not removing the acceleration after round 2) there would have been eight all-top-half pairings involving at least one player on 2, most likely culling the lead group to something like 3-4 players instead of seven. Most likely that would have sped up matches between the winner and defeated strong players bouncing back.

* Also the top-halfers left on 3/3 included the 6th through 10th seeded of the 2/2 top-halfers. With true acceleration three of those players would be in the bottom half of their scoregroup and likely only about one would get through. Instead the tournament had a round 4 perfect score group consisting largely of players who weren't there on either merit or rating as demonstrated in the tournament. This further slows up getting matches that strongly test the winner.

* The completely absurd pairing Solomon vs Gray (though those two had one very interesting game in the past), in which a top-halfer on 1 was paired with a downfloating bottom-halfer on 1.5. Under true acceleration this would only have happened if no bottom-halfers were on 2 after round 2 - unlikely given the size of field.

* Very strong 2300s players who had dropped points wasted kicking 1500s who had already dropped points down to 1/3. Ideally those 2300s should be playing the 1500s who haven't dropped points (thus making sure those players get kicked out of the perfect score group, unless someone is performing way off the scale in one direction or another) while the 2100s on 2 are mostly getting sorted out of the 100% group.

Oepty
16-03-2011, 12:09 AM
One thing that appears to have gone wrong with the accelerated pairings is the inclusion of Peter Miitel and Enoch Fan in the top group for pairings in Rd2. Both these players were in the bottom section in round 1 and beat lower rated players. I realise that things were complicated with the late addition of the Hamilton - Voon pairing in the first round, who both should have been in the top section but I don't know why the other 2 were included. Also Max Mollard who was held to a draw by Garvin was included with the top half players who drew although Garvin was not.
The order of the pairing of the bottom half of the top group also looks strange.
Scott

Garvinator
16-03-2011, 06:36 AM
One thing that appears to have gone wrong with the accelerated pairings is the inclusion of Peter Miitel and Enoch Fan in the top group for pairings in Rd2. Both these players were in the bottom section in round 1 and beat lower rated players. I realise that things were complicated with the late addition of the Hamilton - Voon pairing in the first round, who both should have been in the top section but I don't know why the other 2 were included. Also Max Mollard who was held to a draw by Garvin was included with the top half players who drew although Garvin was not.
The order of the pairing of the bottom half of the top group also looks strange.I think this is all caused by what number was chosen as the cut off point for top half/bottom half. Before the tournament that number was set at 64.

After the addition of Hamilton and Voon, that number should have changed to 66 so that all players around the top half/bottom half cut off stay on the same side of the acceleration cycle.

By not making this change, it would push the original seed 63 and 64 into the bottom half as they would become seeds 65 and 66.

Why this change was not made I have no idea. Easy to make an oversight like this when attempting to use a new pairing program you do not trust (Tornelo).

Garrett
16-03-2011, 07:32 AM
I agree that accelerration should be applied until all lower half players have dropped at least half a point.

Oepty
16-03-2011, 09:38 AM
I think this is all caused by what number was chosen as the cut off point for top half/bottom half. Before the tournament that number was set at 64.

After the addition of Hamilton and Voon, that number should have changed to 66 so that all players around the top half/bottom half cut off stay on the same side of the acceleration cycle.

By not making this change, it would push the original seed 63 and 64 into the bottom half as they would become seeds 65 and 66.

Why this change was not made I have no idea. Easy to make an oversight like this when attempting to use a new pairing program you do not trust (Tornelo).

Garvin, I think your numbers are slightly wrong, I think it was 62, top 31 boards or half the field as initially paired, which was also all those over 1600. There was then 2 more pairings made. Robert Bailey played on board 27 despite being lower rated, I guess he was a replacement for someone who didn't turn up or something like that. Also the wrong James Watson was put in the draw so he wrongly played in the top part as well. This means the number should not have changed with Watson and Bailey just being replaced by Hamilton and Voon in the top section.
Scott

Bill Gletsos
16-03-2011, 10:47 AM
According to the SP files I have the cutoff was 64.

Notice in round 1 that Justin Tan played Nigel Barrow.
He is ranked 8th in round 1 yet he really should have been ranked 18th.
As such he played out of order on board 8 in round 1 as they apparently used the NSW Justin Tans rating.

It appears they made a similar error with James Watson using the NSW James Watson's rating.
As such James was ranked 19th in round 1 instead of 106th.

Oepty
16-03-2011, 01:00 PM
According to the SP files I have the cutoff was 63.

Notice in round 1 that Justin Tan played Nigel Barrow.
He is ranked 8th in round 1 yet he really should have been ranked 18th.
As such he played out of order on board 8 in round 1 as they apparently used the NSW Justin Tans rating.

It appears they made a similar error with James Watson using the NSW James Watson's rating.
As such James was ranked 19th in round 1 instead of 106th.

It was clearly 62 for the first round before the Hamilton - Voon pairing was added. I also by counting players in the second round draw get to 64 players being paired as though they have been in the top section and Kordahi who was in the top section for the first round who had a second round bye. Board 41 looks a bit odd, I am guessing Ghobrial was initially given a 1 point bye in round 1.
I am certainly no expert on accelerated pairings and maybe I have something wrong but I don't think 63 was the number used for either round unless Tornelo or something else made errors. As Tornelo was supposed to be doing the pairings I wonder how relevant the number 63 in SP really is.
Scott

Bill Gletsos
16-03-2011, 04:49 PM
I am certainly no expert on accelerated pairings and maybe I have something wrong but I don't think 63 was the number used for either round unless Tornelo or something else made errors. As Tornelo was supposed to be doing the pairings I wonder how relevant the number 63 in SP really is.I made a typing error.
SP has it set to 64
Sorry for the confusion.

Oepty
16-03-2011, 05:18 PM
I made a typing error.
SP has it set to 64
Sorry for the confusion.

That makes more sense, thanks Bill.
Scott

Pepechuy
04-04-2011, 08:04 AM
The problem with not removing the bounts point is that some players can perceive favoritism.
For the situation you describe (24 players, wide difference in playing abilities, it seems it was 5 rounds but you do not say it explicitly), I think it would be best to use the Amalfi system.

Sergio Pagano
24-08-2011, 12:14 AM
The problem with not removing the bounts point is that some players can perceive favoritism.
For the situation you describe (24 players, wide difference in playing abilities, it seems it was 5 rounds but you do not say it explicitly), I think it would be best to use the Amalfi system.

I agree. The Amalfi system seems working fine in situations like the described one. See for instance this tournament held in Italy last year (5 rounds, 20 players rating interval 2258 - 1371): http://www.torneionline.com/tornei_d.php?codice=1003063A&tipo=c&ord=c&sen=a

Kevin Bonham
05-07-2015, 04:40 PM
Typical example - actually worse than typical because it's a double-point version - of an Abominated Swiss draw in the current South African Open (saw Nigel Short complaining about it on Twitter) :

http://chess-results.com/tnr172929.aspx?lan=1&art=2&rd=4&turdet=YES&flag=30&wi=984

Top half has been accelerated by two add-on points for the first three rounds and then the acceleration has been removed. The round three draw contained 17 bottom-halfers who had 2/2 and 16 top-halfers who had 0/2. With the stronger (and perhaps in cases underrated) end of the bottom-halfers thus playing the weaker (and perhaps in cases overrated) end of the top-halfers, the rating differences in the 16 bottom-vs-top halfer games in round 3 weren't great, and five of the bottom halfers won. The 17th bottom halfer was downfloated to play another bottom-halfer on 1.5 and also won.

Result: six bottom-half players go into the 3/3 scoregroup and with the acceleration removed they are playing GMs and IMs for no reason and most likely making the IMs' GM norm chances much more difficult (if there are norms in the event; I don't know if there are or not). Also there are two all-bottom-half 2.5/3 pairings that may well produce winners who then have to play GMs in round 5. The top two GMs in the 3/3 scoregroup do play top-halfers, but low rated ones who have had stellar results in the first three rounds and under a better acceleration system would play the third or fourth players from the top group.

Kevin Bonham
08-09-2016, 07:05 PM
Copied from FIDE Congress thread:

There was an excellent presentation on Accelerated Pairings by IA Otto Milvang (Norway). He reported on testing of various possible computerised acceleration systems using simulations based on actual tournament fields according to how well they scored on maintaining the relationship between rating and score, reducing average rating differences and mismatches and increasing norm chances. He described most existing acceleration systems as based on "faith and guesses" rather than empirical evidence. In testing, the following system, which he called the Baku System, performed extremely well for events of 9+ rounds:

* Give top-half players one notional bonus point for rounds 1-3
* Reduce the notional bonus to half a point for 4-5
* Remove the notional bonus point for remaining rounds.

He also found that one notional bonus point for rounds 1-2 with half a bonus point for 3 worked quite well. This is compared to the round 1-2 only system as seen in a certain Australian event, which performed so badly in terms of the round 3 comedown that it was not even taken seriously for further testing.

His report is at http://pairings.fide.com/meeting-minutes/2016.html (links from point 2).

Garvinator
09-09-2016, 12:09 AM
In testing, the following system, which he called the Baku System, performed extremely well for events of 9+ rounds:

* Give top-half players one notional bonus point for rounds 1-3
* Reduce the notional bonus to half a point for 4-5
* Remove the notional bonus point for remaining rounds.

He also found that one notional bonus point for rounds 1-2 with half a bonus point for 3 worked quite well.
I will admit I have not read the report as I have just a couple of questions at this stage. Was there any discussion or study of the effect of:

Retaining the notional one or two bonus points for the top half players until all top half players have dropped at least half a point, or until only one player is still on a perfect score?

Does the effect of the Baku System change depending on the rating range of the field? And yes I am thinking of the monster swiss weekenders we have here in Australia, where players are rated from 2500 to unrated and totalling sometimes 80 or so players.

Kevin Bonham
09-09-2016, 01:10 AM
Retaining the notional one or two bonus points for the top half players until all top half players have dropped at least half a point, or until only one player is still on a perfect score?

As far as I know, no contingent system of this kind was modelled.


Does the effect of the Baku System change depending on the rating range of the field? And yes I am thinking of the monster swiss weekenders we have here in Australia, where players are rated from 2500 to unrated and totalling sometimes 80 or so players.

I asked this question too. The test tournaments include similar events so it does work for monster swisses like ours. It might not be suitable for 7 rounds though.

Chigoresov
17-09-2016, 01:39 PM
So acceleration for 7 Round tournaments is dubious and for 5 Round tournaments is just silly, right?

jammo
17-09-2016, 04:57 PM
So acceleration for 7 Round tournaments is dubious and for 5 Round tournaments is just silly, right?

Looks like a bad case of asking for the result that you want. So far as I can see the conclusion of the report relates to events of 9 rounds or more but does not draw a conclusion about events with fewer rounds.

Otto
17-09-2016, 05:37 PM
Hi!
I had the presentation in Baku, and will answer questions about the Baku acceleration and other questions.

In a previous post it is described a system “Retaining the notional one or two bonus points for the top half players until all top half players have dropped at least half a point, or until only one player is still on a perfect score”. Does it exist a more detailed description of the method, or is it actually two methods, one with one points added, and another with two points added. I will run a test on the method.

The Baku acceleration was tested against a large number of different tournament types, and performed better for all tournaments we tested. The largest test set was 432 players with rating from 2700 to unrated.

Since the aim for this work was to have an accelerated system that could be used in title tournaments, the system was not tested on 7 rounds tournaments.
From my experience I will guess that
Give top-half players one notional bonus point for rounds 1-2
Reduce the notional bonus to half a point for 3-4
Remove the notional bonus point for remaining rounds.
will work OK, but it still to be proved.

For 5 rounds I doubt any method will work good. A scheme with 1+1+1/2 may work, but only two rounds without acceleration are too short to smooth the effect of the acceleration points. Anyway it may be interesting to test it out.

Kevin Bonham
17-09-2016, 05:54 PM
So acceleration for 7 Round tournaments is dubious and for 5 Round tournaments is just silly, right?

Any form of acceleration for a 5 round event with a very large number of entrants should reduce the risk of having multiple winners who have not played each other. But for the "abominated" system the chance is too high that the winner will have got there with the assistance of a mismatch win in round 3, which not only gives them an easy point but also gives them an advantage in freshness for round 4. It is better even to have multiple winners who do not play each other than to allow this risk.

The use of the reduced half-point benefit for round 3 for a 5-rounder is worth trying but even then I think there could be issues. There is still some chance that a bottom-halfer gets to 3/3 playing only other bottom-halfers along the way. For instance in the 2010 Ballarat Begonia (an extreme case) nine of the 22 players on 2/2 were from the bottom half. Only four of the ten players on 1.5 were from the top half, and one of these would have had to play a floater from among the 13 top-halfers on 2. So the notional 2-point group for pairing round 3 would have three top half players and the nine bottom-half players, forcing at least three bottom-half vs bottom-half pairings in the 2/2 group.

There also seems to be a big problem with half-point add-ons using Swiss Perfect. It seems to only accept whole point add-ons, and to process half-point add-ons as zero. It would be useful to know how many other pairing programs only accepted full point add-ons.

Kevin Bonham
17-09-2016, 06:09 PM
Hi!
I had the presentation in Baku, and will answer questions about the Baku acceleration and other questions.

In a previous post it is described a system “Retaining the notional one or two bonus points for the top half players until all top half players have dropped at least half a point, or until only one player is still on a perfect score”. Does it exist a more detailed description of the method, or is it actually two methods, one with one points added, and another with two points added. I will run a test on the method.

I think Garvinator actually meant to write "“Retaining the notional one or two bonus points for the top half players until all bottom half players have dropped at least half a point, or until only one player is still on a perfect score”. This captures some of the principle of Stewart Reuben's system from his Arbiter's Handbook first edition while being easy to implement with a computer. (By the way Stewart actually told me that he started his version because he at first misunderstood the concept of the Haley (bonus point) system!)

If I am right then I have used what Garvinator is suggesting in some local tournaments, but the fields are so small that it is rare to have to keep the acceleration on for round 3.

I suggest testing it with one bonus point rather than two retained for the top-half players, and with the bonus point taken off once any of the following conditions are reached:

* There are zero bottom-half players on 100%
* There is only one player on 100%
* There are three rounds to go

Otto
01-10-2016, 08:19 PM
I did the analyses with both 1 and 2 accelerate points.
The first test is performance. With 108 linear distributed players we may expect a player to have a better score in the tournament if (and only if) he has a higher rating. Acceleration with 1 point performs good, but the method with two points has a big jump on the border between accelerated players and unaccelerated players.
3244
The second test is a huge tournament with 432 players with range from 2700 to unrated. The rating distribution will reflect a real tournament, with most of the players in the range 1800-2200 (mixed Gaussian distribution). We will look at the percentage of games played with rating difference > level. The x-axis shows the level, and y-axis the percentage. The acceleration with 1 point performs OK, and the acceleration with 2 point behaves bad.
3245
A round per round analyses shows that acceleration with 1 point is all over good. It seems that acceleration will last for three or four rounds, and then the average rating difference increase in round 5. Acceleration with 2 points will last for four or five rounds, and round 6 will have a mean rating difference on 400 rating points. Not good at all.
3246
The last analysis is a probability for achieving a GM norm. It’s based on distribution om GMs, IMs and FM’s in a tournament, where we analyze rank x, and calculate the probability for he/she will get a GM norm.
3247
Conclusions:
Australian acceleration with 1 acceleration point performs OK. The disadvantage is the big mean difference in rating in round 5.
Australian acceleration with 2 acceleration points should not be used. It is not fear, and it performs bad.

Kevin Bonham
01-10-2016, 08:48 PM
Thanks very much for posting this, Otto.

For comparison people should see the chart on page 4 of Otto's report at http://pairings.fide.com/images/stories/downloads/2016-annex2-acceleratedpairing-v3.pdf

The red line "Mean Haley" is the system used at Begonia and other such events, where the acceleration is taken off for round 3 no matter what is happening.

Under "Mean Haley" the average rating difference is well over 400 points in round 3. In "Australian 1" there is a peak after round 5, making it less good than the Baku system in terms of avoiding mismatches, but it is nowhere near as bad as if acceleration is always taken off after round 3.

pappubahry
14-07-2018, 04:39 PM
I've done a similar sort of analysis to Otto but looking at small, Western-Australia-size fields (20-40 players) for 6-round weekenders. A long discussion and some interactive tables (so many numbers...) is here (https://www.pappubahry.com/misc/accelerated_swiss/).

My main conclusions are that, if all players are correctly seeded, then one point for the first three rounds followed by half a point for round four (almost Baku) is the best of the non-permanent acceleration methods that I tested -- it reduces the mean rating difference across all games by the most, and generally introduced the least amount of inequality of the mismatches faced by the stronger players.

But all forms of acceleration do introduce that inequality. And furthermore, the longer the acceleration lasts, the worse it is for handling a strong player incorrectly seeded into the bottom half of the field -- that player gets shielded from the top players until late in the tournament, leading to an abnormally high percentage of tournament wins or top-3 finishes.

Even though the mismatch inequality provides an advantage to the strong players who get those extra mismatches (and such victories might therefore be unsatisfying, at least to those who missed out!), overall each of the top players won a similar percentage of the time regardless of the acceleration method. So I would say there is still arguably a place for acceleration in "low-stakes" events where you want somewhat more evenly-matched games.

Kevin Bonham
14-07-2018, 06:38 PM
^^^

Excellent work.

These two conclusions are rather bleak for acceleration as a fair system, especially taken in tandem:


Using any sort of (non-permanent) acceleration leads to a more unequal distribution of mismatches, so that some top players get an easier path to a high score than others. The least unequal tested was (again) 1-1-1-0.5-0-0.

If a strong player is mistakenly seeded into the bottom half of the field, the 1-1-1-0.5-0-0 method was the worst-performing method, in the sense that the under-rated strong player becomes much more likely to win or finish in the top three than a correctly-seeded player of the same strength. This is because the wrongly-seeded player gets easier games, being shielded by the acceleration from facing the top players until very late in the tournament.

The Reuben-book method (Australian 1 in Otto's tests) can run into a similar problem to the second if the under-rated strong player is actually extremely strong (the player keeps winning against weaker opponents than they should be getting, and the acceleration doesn't end until they stop), hence the supreme importance of accurate ratings.


If all players are correctly seeded, then the distribution of winners is usually* approximately the same under any of the acceleration systems tested. i.e., a top seed who wins 40% of simulated Swisses without acceleration wins about 40% of simulated Swisses with acceleration, whatever the method.

This is interesting. I had previously suspected otherwise based on a trivially small number of simulations so it is good to know that that at least is not a problem.

Garvinator
21-09-2021, 08:44 PM
In this thread, has there been much, or any analysis, on the Fide Baku acceleration system for 80 to 100 players with 6 rounds where the top players are likely to be rated around 2400 and there will be a long tail of unrated or very low rated players?

This does not seem to be covered, but is rather relevant to Qld weekenders.