AccelRat 2.5 Tournament Software

AccelRat software programs implement a tournament method that takes
advantage of the power and sophistication of the AGA rating engine to
match players based on playing strength, and to assign lightweight
handicaps as needed.

To request the latest AccelRat software, please contact

    Paul Matthews
    466 Foothill Road
    Bridgewater, NJ 08807
    U.S.A.
    70451.2663@compuserve.com

The AccelRat method is data driven.  A tournament director (TD) should
provide the following information.

- An ID (e.g., AGA membership number), full name, and seed tournament
  rating for each player.

- The tournament rules, either AGA rules or Ing rules.

- The number of tournament rounds.

Special case pairing and handicapping options are available, but the
standard method outlined below is suitable for most competitive
tournaments.

- The pairing algorithm gives preference in opponent selection to higher
  rated players.

- In general, the algorithm selects an opponent of similar strength.

- In early rounds, the algorithm avoids pairing players from the same
  club, or who are right next to each other in the rating order.  In
  particular, the pairing of the two top rated players is postponed
  to a later round.

- Handicaps are assigned, based on probability calculations, to give the
  stronger player some degree of difficulty (i.e., probability of losing).
  The target difficulty varies with the average rating of the two players:
  for stronger dan level players (e.g., 7 dan), the target difficulty is 0
  (zero), which means that even game komi is always assigned; for weaker
  kyu players (e.g., 7 kyu), the target difficulty is 50%, which means
  full handicaps; in between, there is a continuous ramp from 0 to 50%.
  Given a typical distribution of playing strengths, dan level games are
  played even, and handicaps start to appear at about the 3 kyu level when
  there are no opponents of similar strength.

- Ratings are updated after each round.  This allows a player to move to
  a higher board based on his/her wins.

- The ratings of very weak players (e.g., below 7 kyu) are considered
  less certain, which allows their ratings to adjust faster, and also
  reduces the number of stones in high handicap games.

The principal awards criterion is simply the number of wins.

A tiebreaking statistic is calculated for each player based on the
average difficulty of his/her games.  That statistic, called
"opposition," is expressed as the rating of a hypothetical opponent who
would be that difficult in a truly even game.  If all a player's
opponents were his/her strength, and all games were played with even
game komi, then the opposition value would be approximately the average
of the opponents' updated ratings.  The calculation of opposition also
takes into account handicaps, rating uncertainty, and the non-linear
relationship of rating difference to difficulty.  Intuitively, a player
who wins the same number of games against stronger opposition deserves a
break.  Players almost never have the same opposition value, and so
there is no need for a secondary tiebreaking statistic.

The recommended method for giving awards is straightforward.

- The player in the "top group" who wins the most games is the
  tournament champion.  You can identify the top group by looking at the
  opposition statistics:  the players who were paired together on the top
  boards form a cluster of high opposition values.  Those players met the
  strongest opponents, and are the natural championship candidates.

- After the championship, give the remainder of the awards based on the
  number of wins, regardless of playing strength.  You can give these
  awards in opposition order, if you wish.

Note that there is no separate "open" section for the top rated players.
An endemic problem with separate sections is that players near a section
cut are likely to get lousy pairings.  There is no need for such cuts.

In summary, the AccelRat tournament method is designed for smooth and
trouble free operation, and offers the following features.

- Predictable peer pairings.

- Lightweight handicapping that is generally accepted by all players.

- Robust tiebreaking when needed.

- No need for separate tournament sections.

- No complicated data or options to mess up.

- Completely scalable, from the smallest to the largest tournament.