Accelerated Pairings

Running Chess Tournaments
1

Several of the championship sections at SuperNationals used a test version of WinTD which had an improved algorithm for handling accelerated pairings—WinTD 4.11 (and earlier) didn’t deal well with the large number of top half non-winners in a section like the HS Championship, where first round draws are fairly common.

This is a description of how this works—this similar to what’s in the rule book, but fixes an obvious error in the description there (referring to “non-losers” where it should mean “winners”) and clarifies (and simplifies) some other steps. Note that this is not the “added score” method, which is quite simple, but not as effective at reducing perfect scores since it “wastes” top half draws by making them a separate virtual 1.5 score group.

Divide the players by rating into groups A (“top half”) and B (“bottom half”). The number in A should be rounded up to the next even number larger than 50% of the players, so with 121, A will have 62 and B 59.

First Round:

Pair groups A and B separately by standard methods. The first round pairings will be similar to what you would get in the second round by standard methods if all the higher rated players won their first round games.

Second Round:

The pairings are done in three batches.

a. First, take the winners in the group A first round games. If there is an odd number, one will float down into the second batch. That would usually be the lowest rated player, but you can switch the downfloater subject to standard equalization limits to improve colors in this batch. Pair this batch by conventional means.

b. There are two sets of players involved in this batch: the non-winners from group A and the winners (not the non-losers) from group B (plus the downfloater if any from a.) In practice, the number of A non-winners will be larger (possibly much larger) than the B winners due to draws in either group. If there aren’t enough A non-winners, you may need to upfloat one (at most) B non-winner to fill out the non-winners. Pair the non-winners in rating order (without regard to their first round result) with the winners in rating order. (I didn’t include the 100 point rating adjustment, which seems to have little point and only makes the resulting pairings even more confusing). This can be adjusted subject to the standard equalization limits to improve colors—changes may include using moving up slightly lower-rated players from the non-winner side. The goal here is to match all B winners with non-winners with (much) higher ratings.

c. At this point, all winners will have been paired. What remains are the losses and draws from A which were not needed to find opponents for the B winners, and all the losses and draws from B. These are paired by conventional means. (This is simpler than what is described in the rule book).

Accelerating by 1/6’s, if someone actually finds a practical use for it, is similar. Now the first step is to divide into 1/3’s (groups A, B and C), with A and B having an even number of players. In the first round, pair A, B and C separately by standard methods. In round two,
again pair the A winners against each other in the first batch. In the second batch, pair A and B non-winners in rating order against B and C winners. Any non-winners not paired against B-C winners are paired by standard methods.

2

Where is the documentation for accelerating by 1/8ths?

3

In an empty mayonnaise jar at Funk and Wagnalls.

4

Thank you, O Great Carsoni.

5

I like it! It seems much cleaner than the rulebook method. There will always be plenty of non-winners from group A to clobber the winners from group B. In fact, the idea behind accelerated pairings should work better than before, because (I assume) the highest-rated group A non-winners would be the ones paired against the group B winners, reducing the chances of a backfire.

Question: Does “by conventional means” in batch c. mean that, unlike batch b., the scores (loss or draw) are to be taken into account? i.e. are draw-ers paired against other draw-ers, and losers against other losers?

Bill Smythe

6

Picturing Tom Doan as Karnack…

(and remembering Carson with an envelop held to his temple, saying “Sis-Boom-Bah”…then opening the envelop and ‘reading’ “What is the sound of an exploding sheep?”)

7

Correct. Draws with draws, losses with losses. Each of those score groups will include some low rated players from A and then all the B’s.

8

That one was one of his best. Normally what was funny was his reaction to the audience when a Karnack joke fell flat. :laughing:

9

That’s “C-a-r-n-a-c”, as in “Carnac the Magnificent”. :slight_smile: (Yes, I’m old enough to have seen Johnny Carson doing this gag - just barely.)

10

Oops, you’re correct. I got confused with the temple of Karnak in Thebes (and added a typo to boot.)

11

May a queasy camel freshen your bath. :laughing:

12

Back to the original topic – sort of –

My problem with accelerated pairings has always been its effect on the great mass of 1-pointers (6/8 of the players) after two rounds. For this vast majority, the round 3 pairings feel like round 1 pairings in a “normal” tournament.

To prevent this round 1 “feel” in round 3, in an event of at least five rounds, the idea of accelerated pairings could be continued in rounds 3 and 4, as follows. For simplicity let’s assume there are no draws:

  • In round 3, pair the 2-pointers normally (top half vs bottom half). Likewise, pair the 0-pointers normally.
  • Pair the 1-pointers by quarters (top quarter vs 2nd, and 3rd vs 4th).
  • In round 4, pair the 3-pointers normally, and likewise the 0-pointers.
  • Pair top-half 2-pointers against each other, and likewise bottom-half 1-pointers.
  • Pair bottom-half 2-pointers against top-half 1-pointers.

Bill Smythe

13

That would be an interesting variation. Maybe a TD could be found to try it. One obvious comment that will get made is the bottom 2-1 player of the top half complaining that he has to face another 2-1 in the penultimate round while the top 2-1 player of the bottom half gets to face a 1-2 player. In a 64 player event with the first three rounds having all decisive games and no upsets (and no color swaps) that would be 2-1 player 20 vs 2-1 player 12 and very next 2-1 player 33 vs 1-2 player 21, with a good chance of having player 20 sitting one point behind player 33 and only one round to go to try to get back into a tie. Standard (2-round) acceleration would have had 2-1 players 5-32 paired amongst themselves and 1-2 players 33-60 paired amongst themselves.