Thursday 22 January 2009

Upside down pairings

A number of years ago I was directing a tournament when a New Zealand player (I cannot remember who) asked me, "In which rounds do you reverse the pairings?" "Reverse pairings? What are those?" I replied. He explained to me that in the swiss pairing rules used in NZ, some rounds played 1vn, 2vn-1 etc in each scoring group, rather than top half v bottom half as is used in most common systems. "Not in Australia" I remarked, and that answered his question.
While reversed pairings aren't in the Lim or Dutch pairings rules approved by FIDE, they are used in the Dubov and Burstein systems (which was the system used in last years Olympiad). The intent of these systems is to try and 'even' the field that players face. In part this counteracts the effect that seeding can have on pairings (ie in a tournament with no upsets players seeded close together can play fields if markedly differing strengths). Of course having pairings always reversed can also have its problems, as the New Zealand team found in Dresden.
However the idea of using reversed pairings does have some merit. For example in a theoretical 32 player tournament where round 1 is paired normally and round 2 is reversed (and the higher seeded player always wins), the total of the seeding numbers of opponents for every player is 33!
Having spent a couple of hours scribbling figures on my notepad I've looked at a couple of different approaches to reverse pairings. On approach I wouldn't recommend is using reverse pairings for every round. This has the same problems that the traditional systems have, just visited upon different players (ie in a normal 32 player swiss, seed 16 has it easy and 17 has it tough, in a reversed system this is also reversed).
However a mixture of normal and reversed pairings (as the New Zealand system uses) isn't bad. Pair round 1 normally, reverse rounds 2&3 (so players don't play white in every reversed round) and the revert to normal pairings for round 4 (and repeat as required), keeps the strength of the fields much closer together. Without boring you with tables of numbers the seeds 1&2 (both on 4/4) played opponents with a seeding total of 45 and 43 respectively. For the players on 2/4 (12 in all) there is only a range of 58-74 in seeding totals (ie the toughest field anyone plays is seeded 14.5 on average, the weakest 18.5). There is a bigger difference in the 1/4 and 3/4 score groups (9.75 to 18 for the 3/4) although this reflects the fact that one player went WWWL to get to 3 points while another went WLWW.
Nonetheless this is more of an academic exercise, rather than a serious proposal to change the current pairing systems. If there is one thing chessplayers crave, it is familiarity, and to offer change is to invite confusion. I'll probably put this idea to a practical test at Street Chess (or a junior coaching day) but it probably won't go much beyond that.


Garvin said...

Shaun, will you be giving this pairing system a few practical pairing tries at Street Chess?

If yes, would you be willing to share restuls afterwards?

Shaun Press said...

I've still looking for a suitable tournament to run it on. Anything with a field of less than 20 players probably won't reveal much, as the score groups become to small to matter. But when I do I'm happy to share the results.