Pairing theory

    Keywords: Theory, Tournament

Pairing Theory is more deeply investigation for tournament directors about who to pair and why.

Most on this page comes from [ext] http://www.math.toronto.edu/jjchew/software/tsh/doc/pairing.html but I plan to give to add more go related information. Feel free to add to it.

Some points (like monagony for example ) seem to be not worth the space because they are always followed in go tournaments but science is sometimes just stating the obvious.

Table of contents

Goals of pairing

The formal main goal of a tournament is to find the best player of the tournament but it is also to give all (lower ranked) players interesting games.[1]

This gives the following list of properties which unfortunately can not all be satisfied in any particular tournament.

For finding the best player (tournament winner)

  • Aristomachy: Top players should be paired with each other, especially toward the end of the tournament. King-of-the-Hill pairings do this best.
  • Fairness: The final ranking of players should be equivalent to a ranking according to their level of performance in the tournament.
  • Inclusivity: A player in the topgroup should not needlessly be excluded from contention for a prize by being paired with someone who is out of contention.
  • Monotony: A lower-ranked player should not be paired so as to make it more likely that he will become the overall winner than that a higher-ranked player will.

For interesting tournaments

  • Exagony: Players who can play eachother at their local go club schould not be paired against eachother if it is not nessecary for other reasons.
  • Suspense: The outcome of the tournament should be determined as late as possible.[2]
  • Monagony: Players should play each other as few times as possible. Repeat pairings can prevent other players from catching up to the repeaters, and do not accurately measure the repeating players’ ability to defeat a wider field. Players also do not in general like playing each other in consecutive rounds.

Other desirable properties

  • Implementability: A pairing system must be so that if the computer fails and the pairing has to be done by hand it is reasonably possible.
  • Incentivization: A player should not be placed in a position where tieing or losing is strategically preferably to winning. For example, a player who has clinched a place in a two-player final should not be paired with a possibly weak player whose victory would send him to the finals.

[1] tapir: Is it? There are tournaments which are basically too short to do this - e.g. 4 round McMahon but still are enjoyed by participants, where the main aim seems to be to have a nice event on a single day. Or imagine full handicap tournaments with a pairing that encourages handicap games where the handicap pretty much randomizes the outcome. Also the point about desirable monotony isn't clear to me :)

[2] tapir: This principle should not be limited to pairing only, but to choice of scoring system and tie breaker as well.


This is a copy of the living page "Pairing theory" at Sensei's Library.
(OC) 2011 the Authors, published under the OpenContent License V1.0.
[Welcome to Sensei's Library!]
StartingPoints
ReferenceSection
About