PDA

View Full Version : Hyper accelerated pairings.



Davidflude
29-01-2006, 11:44 AM
This is a suggestion for a radically different form of Swiss Pairings. This system might be worth trying in an Allegro tournamet.


According to the Fide rules

Basic Principles of Swiss System Tournaments
The basic principles of a Swiss System tournament are:
1.The number of rounds to be played is declared beforehand.
2.Two players may play each other only once.
3.Players are paired with others of the same score, or nearest score.
4.When possible, a player is given the white pieces as many times as he is given the black pieces.
5.When possible, a player is given the colour other than that he was given the previous round.
6.The final ranking order is determined by the aggregate of points won: 1 point for a win, 0.5 point for a draw and 0 point for a loss. A player whose opponent fails to appear for a scheduled game receives one point.
Listing the Players in ranking order
My proposal does not alter how this is done.
Basis of pairings
One player is paired at a time
First round Pairings
The colour allocation for the top rated player is determined randomely. Next a dice is thrown to determine the number of the top players opponent. The eligible opponent = number on dice (d) plus Player number (n).
Pairing now proceeds straight down the list pairing each player against opponents d+n. When there are less then n players to be paired then each player is paired against his nearest opponent and the bye is allocated to the lowest rated player. Colours of the higher rated player are alternated board by board.
Pairings for subsequent rounds
Pair players one by one starting from the top according to score and rating. The top rated plays the closest player to him that he has not already played such that colour discrepancies are minimised. (I am not going to spell this out in detail).
Advantages
There will be far fewer pairings between players with big rating differences
The strong players are going to have to play strong opponents right from the start.
The yo yo affect where players close to the middle play grandmasters and beginners in alternate rounds will be greatly reduced.
There should be far more draws which will help to lead to a clear tournament winner.
Players who play well above their rating will gain heaps of rating points, players who play well below their rating will take a bath.

noswonky
29-01-2006, 03:06 PM
Sounds interesting.

It would seem that under your system, everyone gets to play mainly against opponents of similar strength. This would mean that the lower rated players would tend to finish with higher scores, and the stronger players would finish with lower scores than in a conventional Swiss. Am I right?

So it's a bit like a handicap system as it make it harder for the strong players and easier for the weak players. Do you think the lower rated players might have chances to win a tournament under this system?

Davidflude
29-01-2006, 03:40 PM
After two or three rounds those lower rated players who are performing well would get paired up to the top. Then they would have to show what they can do. Under the present system lower rated players keep getting paired up even with patchy results.

Frank Walker
29-01-2006, 03:58 PM
Hey how about we roll a dice to see who wins too?

Rincewind
29-01-2006, 04:36 PM
I think the system is a bad one if you want to reward the players based on their objective chess strength. Current 7 round tournaments have the problem with players in the middle and (to some extent) those towards the top having played a significantly different strength of field leading to a rating group lottery. This system will make that problem more pronounced. To me it sounds like a "reasonable" way to run a handicap tournament as the sweet runs will be distributed by rating, however with all handicaping you want to make sure your ratings are accurate. The underrated player (by your system) will be at a huge advantage compared to normal swiss pairings.


The yo yo affect where players close to the middle play grandmasters and beginners in alternate rounds will be greatly reduced.

I take it you are starting with the premise that the so-called "yo-yo effect" is a bad thing. I suspect some lower rated players often enter tournaments with a high average field with the idea that losses will not hurt them and one or two big scalps will see them make a huge rating profit from the tournament. They are attracted by yo-yo effect.


There should be far more draws which will help to lead to a clear tournament winner.

It's not clear that their will be more draws just because more games are between similar rated players. And even if this were true, it is not obvious that this would lead to a greater likelihood of a clear tournament winner. Perhaps you would like to expand on both of these points.


Players who play well above their rating will gain heaps of rating points, players who play well below their rating will take a bath.

What you say is true of all tournaments and I can't see why it would be a particular feature of the pairing scheme you outline. In fact the low rated player would have trouble getting paired up to high rater opposition, so while they might make some rating points by beating their peers, big scalps would be harder to come by.


Hey how about we roll a dice to see who wins too?

Good call Frank, but grammatically speaking you are both a pair of goose. ;)

shaun
29-01-2006, 05:31 PM
Looks like a fun way to run a handicap tournament, especially where the emphasis is not on finding a winner, but on giving everyone a number of good games. I'll give it a try at one of the ACTJCL develolment squad days. In honour of its inventor I'll call it the "Flavoured Dude Pairing System".

PHAT
30-01-2006, 08:52 AM
The colour allocation for the top rated player is determined randomely. Next a dice is thrown to determine the number of the top players opponent. The eligible opponent = number on dice (d) plus Player number (n).
Pairing now proceeds straight down the list pairing each player against opponents d+n. When there are less then n players to be paired then each player is paired against his nearest opponent and the bye is allocated to the lowest rated player. Colours of the higher rated player are alternated board by board.
Pairings for subsequent rounds
Pair players one by one starting from the top according to score and rating.

WOW! I had a similar idea some time back. The difference was there was no dice as there is/was a proscriptive progression to the pairing.

Pairings formula

1 Vs P+n+1-R

Where:
P = the player position to be paired
n = the number of rounds in the event
R = the number of the round

Thus, for the first round of a seven round event
1 Vs 8
2 Vs 9 ... 7 Vs 14
Then reverse the colours for the next group
22 Vs 15
23 Vs 16 ...

Round 2
7 Vs 1
8 Vs 2 ...
12 Vs 6
Then
19 Vs 13 ...

Following this algorithm the last round would be
1 Vs 2
3 Vs 4
5 Vs 6 ...

For players already paired the opponent should be the next eligable player up/down, alternately for each time they are floated.



Under this system,
1. The winning player has been playing only the top 8 players.
2. The best two performing players should meet in the last round, reguardless of the size of the tournament or the number of rounds.
3. Players generally, have opponents who are very close to their own current form in every round.

An advantage over Barry's system is that it has no random aspect that could see the best two players miss each other, and fluke dice rolls pairing the top seeds early in the tournament.

Comments please.

eclectic
30-01-2006, 09:17 AM
i reckon all this is treading on starter's "metrically competitive-index patented?" :rolleyes: "junk-round eliminator" :eh: "box-hill brewed" :eek: "dual-divisionalised" :hmm: "superaccelerated swiss system" :whistle:

any comments please ? ;)

PHAT
30-01-2006, 10:08 AM
i reckon all this is treading on starter's "metrically competitive-index patented?" :rolleyes: "junk-round eliminator" :eh: "box-hill brewed" :eek: "dual-divisionalised" :hmm: "superaccelerated swiss system" :whistle:

any comments please ? ;)

Agreed.

You may find it amusing that I don't actually like my own system, because it takes away many chances to claim a scalp. However, I do not try to come up with ideas that are good for me, but good for chess. I am not important - the group is important. :)

Kevin Bonham
30-01-2006, 04:48 PM
Comments please.

One obvious issue is that if there are no upsets, seed 2 plays all bar two of their games against players defeated by seed 1 in the previous round. Chances are that these players will be less motivated and seed 2 will get easier games against the same players than seed 1 and hence have an advantage when playing seed 1 in the final round.

Another is what happens if there are draws. Say seed 1 draws in round 1 but otherwise all games go according to seedings. In a normal Swiss, seed 1 will be back to top board in three or four rounds. In your system, seed 1 plays seed 2 in round 2, and seed 3 in round 3, seed 5 in round 4 (4 lost to 2 in round 3), and so on.

Systems that aim to force two equal leaders to meet in the last round are silly anyway. Typically the players will promptly split the point. It is better for the leaders to meet a few rounds out from the end.

I suggest anyone wishing to promote an alternative Swiss draw provide a full tournament simulation showing how it would work in practice. Get a list of players from a real event, use a formula for win/draw/loss likelihood per rating difference (or perhaps even per ratings as well) and do some random simulations. Post them as evidence that your system actually does what you claim it to. No radical reworking of the system can be taken seriously without a test. It is all too easy to come up with theories based on the top seeds always winning that fall over if there are draws or upsets.

Spiny Norman
31-01-2006, 06:09 AM
i reckon all this is treading on starter's "metrically competitive-index patented?" :rolleyes: "junk-round eliminator" :eh: "box-hill brewed" :eek: "dual-divisionalised" :hmm: "superaccelerated swiss system" :whistle:

any comments please ? ;)
I went and added "Competitive-index-powered" before I re-read and noticed the "patented" bit. :doh:

Desmond
19-07-2006, 08:57 PM
I suggest anyone wishing to promote an alternative Swiss draw provide a full tournament simulation showing how it would work in practice. Get a list of players from a real event, use a formula for win/draw/loss likelihood per rating difference (or perhaps even per ratings as well) and do some random simulations. Post them as evidence that your system actually does what you claim it to. No radical reworking of the system can be taken seriously without a test. It is all too easy to come up with theories based on the top seeds always winning that fall over if there are draws or upsets.
Has anyone taken Kevin's suggestion and done a simulation? I would be interested to see it, just wish that I was more tournamentically minded. (I always find it's good to learn a new word every day. If you must invent that word, so much the better. ;) )

Garvinator
19-07-2006, 10:36 PM
Has anyone taken Kevin's suggestion and done a simulation? I would be interested to see it, just wish that I was more tournamentically minded. (I always find it's good to learn a new word every day. If you must invent that word, so much the better. ;) )
Are you asking if anyone has done a simulation for accelerated pairings? Not quite sure what you are after?

Desmond
20-07-2006, 08:24 AM
Are you asking if anyone has done a simulation for accelerated pairings? Not quite sure what you are after?
Yes, taking an actual tournament that was paired using Swiss, and comparing what would likely have happened if it was paired using accelerated pairings instead. Pretty much what Kevin said :P

In particular, I'm interested which pairing system would produce a clear winner of the tournament in a smaller number of rounds.