Swiss Pairing.

[Peter Rhodes]

Purely out of curiosity,

when Swiss Pairings take place, are the FIDE directives often followed (for non FIDE events) ?

If not, is there a standard set of procedures in place, or does ECF/BCF even recommend anything.

Has anyone ever experienced anomalies or unsatisfactory pairings ?


(for example a mum approaches the arbiter and asks that her two boys don't play each other etc).

[Ian Thompson]

If not, is there a standard set of procedures in place, or does ECF/BCF even recommend anything.

See http://chessarbiters.co.uk/pairings.aspx for the pairing rules that should be used in England, Scotland and Wales.

[Peter Rhodes]

That's a very handy link, thanks Ian.

[Sean Hewitt]

Like with everything else, the rest of the world does one thing whilst the British think that they know best and doing something different!

What do people think of the differences between the two (FIDE and British) pairing systems?

[Richard Bates]

Like with everything else, the rest of the world does one thing whilst the British think that they know best and doing something different!

What do people think of the differences between the two (FIDE and British) pairing systems?

Not perhaps directly addressing the question, but would it be fair to say that a fundamental difference between UK and abroad is the attitude to "human error".

Outside of the UK, tournaments are dominated by Computer pairing systems (presumably mainly for convenience) which tend to be rejected by UK arbiters on the grounds that Computer systems, whilst being generally sound, have a tendency to generate strange pairings on occasion. Of course the arguable flaw in this attitude is that UK pairings have a tendency to produce wrong pairings, as a result of simple human error. Essentially in the UK the object is (unachievable?) perfection.

[Ola Winfridsson]

In the past the computer pairings had a very strong tendency to throw colour anomalies such as 5 whites and 2 blacks, but I believe they've mostly been ironed out now that they've developed the pairing software further.

[Kevin Thurlow]

"(for example a mum approaches the arbiter and asks that her two boys don't play each other etc)."

Sometimes they're that polite!

The big difference between England and Rest of World is that RoW uses the pairing that the computer produces, whatever it is, whereas here we do tend to keep family members and club colleagues apart. There has also been a tendency to fiddle the draw to help some selected people get norms and ratings. This might be OK if it doesn't adversely affect others (who have after all also paid for the pleasure of being in the tournament).
I think that you should try to avoid club colleagues & family members etc. playing each other in the first round, and thereafter only do it if doesn't foul up anyone else, (e.g. someone else gets a 180 instead of a 140), or if you are dealing with a low score group. This also applies to sorting out colour sequence - If you have two people on 0.5/2 who have gone WW and two on 0/2 who have gone BB, the correct pairing is to pair the 0.5s and give one a third white, and to pair the 0s and give one a third black. Unless I see a good reason not to, then I would prefer to pair 0 v 0.5 in each case, so they get their colour sequences back on track. But if the players were on 2/2 I wouldn't do that (although Jack Speigel did.)

The best computer pairings I've seen come from software written by David Horton specifically for the Thanet Congress. The commercial packages I've seen do produce utter tripe at times. Tournament Director did one where there were players on 3, 2.5, 2.5, lots on 2 etc. and it gave 3 v 2 and 2.5 v 2.5 (with the wrong colour), whereas the obvious 3 v 2.5 and 2.5 v 2 worked!

The big advantage of the software (apart from the speed of doing the pairings) is that when people say they have the wrong opponent or colour, you can look at the screen and say, "Computer says 'No'." But doing the pairings manually can be fun...

[Sean Hewitt]

Interesting replies so far. I've always thought it odd that English Arbiters don't trust computers to get the draw right given some of the draws that I've see done in this country!

However, I was actually referring to the differences between the swiss pairing rules rather than the different processes employed for doing the draw.

Perhaps most people don't know (or care) what the differences are?!

[Roger de Coverly]

However, I was actually referring to the differences between the swiss pairing rules rather than the different processes employed for doing the draw.

A suggested axiom for computer pairings should be that the same data with different software should always produce the same results if the underlying pairing rules are the same. The British rules are full of "may" and "can" where the arbiter appears to have a choice so the consistency condition would be difficult to satisfy using British rules. Also it may require a form of AI to be built into the pairing program. That said, it does allow pairings to be "adjusted" as Kevin suggests without breaking the rules. It's amazing how the British rules always seem to give norm seekers the final round opponents they need.

[Paul McKeown]

Perhaps most people don't know (or care) what the differences are?!

To be honest, I think most people come to tournaments to play chess, rather than argue the toss with the arbiter about the Swiss Pairing System, and anyway the presumption must be that one shouldn't unnecessarily annoy the arbiter, who might, after all, have to make a decision about one of one's own games!

Barring the very occasional outright blunder, e.g. same opponent twice, four of the same colour in a sequence, etc., it is usually more profitable to preserve one's fighting reserves for the games themselves!

Regards,
Paul McKeown.

[Mick Norris]

Interesting replies so far. I've always thought it odd that English Arbiters don't trust computers to get the draw right given some of the draws that I've see done in this country!

Blimey Sean, and you've not even played in a tournament where I've been doing the pairing manually!

[Ola Winfridsson]

Sean, I believe this can be very arbitrary and differ from organizer to organizer (i.e. not necessarily a difference between the UK and RoW). As long as it's stated in the tournament invitation that a "modified" Swiss system will be used, you can pretty much do whatever you like.

For instance, accelerated pairings - subdividing the field into let say the top 40% and the bottom 60% - in the first couple of rounds before amalgamating the whole field again for round 3 is quite common in Swedish tournaments.

Another interesting rule that I've seen applied concerns grading prizes (in 7-round events): once a player has dropped 2½ points in the tournament he/she will only be paired with other players in the same grading prize bracket (within reason of course).

[Alex McFarlane]

Humans make mistakes – be they arbiters or programmers.

I would be very pleased to see a copy of the David Horton pairing program as I have an obvious interest in the matter.

At the top tournaments in Britain it is common to have computer and manual pairings working in tandem and differences are compared and judged on merit. Sometimes the computer pairing wins but more often it is the Arbiter (a ratio of 2:5 would seem right but I’ve never kept score).
Some programs are dreadful, some much better – probably better than the less experienced arbiter. The problem with most programs is their inability to break a pairing and will then do something plain daft rather than change the upfloat, for example.

Three comments about the FIDE pairing system: 1) It encourages the Swiss gambit (discussed elsewhere) where a high rated player can take a first round bye and then meet relatively easier opponents than his contemporaries as the upfloat is the top player and plays the bottom of the scoregroup above. 2) It means that if you have a sole leader that person always gets the hardest possible opponent whilst his/her closest rival due the same colour will get a relatively easy ride. 3) I believe that the pairing system was designed to be programmed and it therefore does not necessarily contain the best Swiss pairing ideas, merely those that are easier to program.

Norm chances are certainly looked at in Britain but no major ‘fiddles’ are carried out. Indeed, in the last round of the Scottish this year, a player was given the only player in his scoregroup who made an IM norm impossible for him. This was not changed because of the effect it would have had on other players in the group.

I had occasion to work with a foriegn arbiter who had only previously done computer pairings. He enjoyed the experience and made the very valid point that he would now be able to attempt to explain why a certain pairing had been arrived at by the computer. He had had little idea before.

Jack Rudd will confirm that when the computer comes up with a different pairing it is analysed to see why. Occasionally it is because it has done something clever, more often because it has done something idiotic. My worst experience with Tournament Director was when it paired a player against himself, insisting that the person he should have met be given a bye. To add insult to injury it then complained that the player was getting his third black in a row!!!

Swiss Master, to my mind the best of the commercially available packages, will be used at the Open and other events at the London Classic in December. Human pairings will not be used at that event.

Alex McFarlane
CAA Chief Arbiter

[Sean Hewitt]

For instance, accelerated pairings - subdividing the field into let say the top 40% and the bottom 60% - in the first couple of rounds before amalgamating the whole field again for round 3 is quite common in Swedish tournaments..

That's exactly what I've done here this weekend!

[IM Jack Rudd]

This topic seems to have gained in relevance after the comments to the Uxbridge thread. Specifically, there were justified questions about how good a pairing system is that allows a player seeded outside the top ten to win with the Swiss Gambit.

I think the question to bear in mind is "what is the fundamental aim of a Swiss pairing system?". And my answer would be that it's to approximate the effects of an all-play-all between the players concerned; in other words, to neutralize the "luck of the draw" factor as much as possible. Thus, the better final score correlates with TPR, the better the pairing system is likely to be.

I don't know what the best way of doing this is, and I suspect nobody else does either. But I'm pretty confident that the best way of finding the best possible pairing system is for there to be a large number of pairing systems in use, and for there to be plenty of analysis of the relative performances of these systems. That way, one can test the effects of altering various variables.