“Accelerated Pairings”
<The following document was submitted to the IWF Council at the 2004 Rome AGM as a proposal for use in the World Individual Championships>

If there are more players in a tournament than can technically handle by Swiss Chess pairings (typical of large DBM events), "accelerated pairings” is a widely established method used in international chess tournaments. With Accelerated Pairings, it is possible to handle 2ⁿ⁺¹ player tournaments with only "n" rounds. If the initial seedings are reliable, only one player will complete the tournament with clear wins and therefore be a clear winner of the tournament or championship.

How does it work?

Players are seeded from highest to lowest based on their ratings. In the first round, players seeded in the top quarter of the draw are paired against those in the second quarter and those in the third quarter paired against those in the fourth quarter. The next round winners in the top half of the tournament played each other as did losers in the bottom half. But non winners in the top half played winners in the bottom half. Thus it is possible to have a first round winner play a first round loser which at first looks rather strange. The idea however is to try to avoid anyone in the bottom half of the draw being amongst the tournament leaders after the second round.

Field Trials

We were challenged to apply Accelerated Pairings at a national competition level (in Australia) and used it for our national Championship “Cancon” held in January, plus two additional state championships last year (Victorian & NSW). All these tournaments had player numbers well in excess of the 2ⁿ limit for standard Swiss Chess. The Cancon results can be viewed at;
http://www.nwa.org.au/dbx/Cancon/2004/2004results.html

What’s effect did Accelerated Pairing have? For Cancon, the eventual winner was put up against the best players the tournament had to offer. The winner Bram Jakins went through 3 out of the 5 top ranked players in the event. Similar results were achieved at the state championships.

Software developed by David (Glicko) Young was used to process the Accelerated draw.

Comparison To Current IWF Practice

The current IWF method of allowing the top 30-50% seeded players to be paired with the lower ranked players (round one only) does nothing to eliminate the luck of the draw for opponents. Previous IWF championship winners have frequently played only a small percentage of other top ranked players to achieve the title. While this is not a reflection on their ability, the tournament has simply thrown them an easier string of opponents to get there result.

In many respects, we have to determine “are we seeking a champion?”

Followup Reading;

If your interested in reading further articles on Accelerated Pairings, please try the suggested sites or simply do a Google search on “Accelerated Pairings.” Accelerated Pairings are officially detailed in the US and Canadian Chess Federation handbooks.

http://www.bcf.org.uk/national/2002/mcfautumn02.htm
http://www.ap76.com/LV/gotournaments/gotournaments.htm
http://scichess.org/

 

<Basic Programming Outline developed by David Young for software used at Cancon.>

ACCELERATED PAIRINGS:

The following method results in only about half as many players having a perfect score as under the basic Swiss system and increases the opportunity of a single clear winner. It has the effect of adding an additional round to a tournament with no bias towards the top seeded players. Accelerated Pairings only affects the first two rounds, after which standard Swiss system kicks in using each player’s accumulated score.

In the first round, after the bye, if any, is issued, the players are arranged in the order of their ratings from top to bottom. The field is then divided into four groups (A, B, C, D) starting with the top rank players in “A” progressing down to the lowest ranked players in “D”. These groups should contain equal numbers of players. If not, ensure that groups A & B are equal and the same for C & D. Preferably place the extra players into groups A & B [eg. 26 players = A(7), B(7), C(6), D(6)].

The first-round pairing* of players is then conducted as follows

• Section (1); players in group A are paired against the players in group B.
• Section (2); players in group C are paired against the players in group D.

For the second-round, the players are regrouped accordingly;

• Group W: winners and players who drew from section (1),
• Group X: remaining players from section (1),
• Group Y: winners and players who drew from section (2),
• Group Z: remaining players from section (2).

Then pair off players as follows:

• Group W players are paired off against each other with, as much as possible to improve equity, group B members being matched against each other.
• Group X players are paired with Group Y players.
• Group Z players are paired off against each other with, as much as possible to improve equity, group D members being matched against each other.

* the actual method used for pairing within or between these groups can be any detailed in sections #### (ie. Random, Chronological, Historical and allowing for exclusions based on geography, club or consanguinity ).

For the third and all subsequent rounds of the tournament, pairings are processed using standard Swiss Chess system based on accumulated total scores from previous rounds,  applying whichever pairing* method has been selected.

“Drawn” games are defined as those which achieve a 5-5 or 6-4 result (or equivalent using alternate scoring systems, eg. BHGS)

Floaters; prior to pairing players between groups in the second round, there will inevitably be a mismatch in numbers. If so, use the following for assigning opponents to balance the group numbers; (a) remove players who drew in the first round are pair them off with each other [assuming they haven’t previously met]; or (b) select the lowest seeded player within the group and pair off with the top seeded player from the next lowest group.

Bye Player; For the first round, select a player who would initially be grouped in the lower half of the field and re-insert back into Group Y for the second round. For the second round, select an appropriate player from Group Z. Byes are only issued if the overall number of players is odd.