A different pairing method

Running Chess Tournaments
21

In an effort to reduce the mismatches I’ve tried acceleration in our club championship in the past with a number of players that didn’t require it to reduce perfect scores (such as accelerating 5 rounds with 24 people). What I found is that the overall level of opposition is significantly different for people at the very top or the very bottom, but the broad middle (over 90% of the players) had different opponents but approximately the same general level of opposition. It’s been a while, but when I ran a mock-up with the higher-rated player always winning I found that in general there were more close matches with standard pairings than with acceleration when the players are less than 2 to the Nth power where N is the number of rounds.
Note that after two rounds there will theoretically be the top 1/8 at 2-0, the bottom 1/8 at 0-2 and the middle 3/4 at 1-1 (instead of 1/4, 1/4 and 1/2 respectively). That means that in a 32 player field (4 round event) you are supposed to have round three pairings with 5-17, 6-18, 16-28 instead of 9-17, 10-18, 16-24. The middle groups stay larger and have a wider rating range under acceleration as opposed to standard, with the result that the rating differences in those round three and later games are greater.

Because I haven’t used the OP’s pairing method it looks complicated (as opposed to the “easily” explainable Swiss pairings that I’m adept with), but it may meet his goals just fine.

22

Looking at it further, with only one round of accelerated pairings, and no upsets, everyone would play in the second round the people they would have played in the first.

In a 16-player field, however, TWO accelerated rounds would still only drop the largest disparity from eight steps to six.

23

Doing only a single round of acceleration would generally be an error. Acceleration should assume at least two rounds (and usually exactly two rounds).

24

It’s significantly easier than standard Swiss pairings. (I would still call it Swiss paired, just not standard Swiss). There’s no need to decide odd men. There’s also no color balancing, although it’s a bit unfair to tout that as an advantage because in reality I’m just ignoring it.

I actually have a simulation engine for pairing methods and a standard model for running different methods, so I can simulate 1000 tourneys within given parameters and measure such stats as average rating difference in games. I will eventually run this through it, and see how it compares.

25

I think it was Voltaire, but it might have been Humpty Dumpty, who cautioned “Before you would debate with me, please define your terms”.

Many players seem to think that “accelerated pairings” are some kind of magic potion, or perhaps a magic push-button on the computer, that somehow automatically creates closer pairings and/or reduces perfect scores. Since TDs don’t need to know anything about pairings anymore – they just let SwisSys or WinTD do it – they don’t need to know what accelerated pairings are, nor why they work, nor even whether they work.

But let’s look at the rulebook for a description of accelerated pairings. For simplicity I’ll overlook the possibility of draws.

In round 1, instead of pairing top half vs bottom half, one pairs first quarter vs second, and third vs fourth.

In round 2, instead of pairing players with equal scores, top-half losers are paired against bottom-half winners. (Top-half winners still play each other, as do bottom-half losers.)

It takes both of these concepts together to achieve the desired result of fewer perfect scores after two rounds.

Because the top-half losers (generally players in the 2nd quarter) will mostly defeat the bottom-half winners (generally players in the 3rd quarter), neither group will end up 2-0. And half of the top-half winners will defeat the other half, resulting in only 1/8 of the field, instead of 1/4, standing at 2-0 after two rounds.

And do accelerated pairings reduce the number of large mismatches? Not significantly. They just re-arrange them. For the vast majority in the middle 6/8 of the field, round 3 will feel pretty much like round 1 in a “regular” tournament.

So how can it make sense to talk about accelerated pairings “for a round” or two? Accelerated pairings are inherently a two-round phenomenon. If you use accelerated pairings “for a round”, what does that even mean? Do you use the round 1 method (quarters) or the round 2 method (unequal scores)? Unfortunately, neither will work without the other.

Aha, you figured that out.

No, it would have them in round 3 instead.

Bill Smythe

26

Because writing out pairings and standings by hand is a pain, and printing them out by computer is pretty and professional-looking? Plus what Eastside said about pairing cards and results submissions.

27

That’s my biggest reason. If I didn’t use a computer, people would have to read my handwriting.

28

My experience with accelerated pairings for rounds one and two is that is moves most of the usual round one pairings to round 3. I think there are some advantages to this in a five round tournament with three rounds on Saturday but it is startling to see essentailly round one pairings magically appear in round 3.

29

Pretty much true except for the top and bottom eighths. So if you want a blow-out round to end the first day and get most of the players home earlier that is a possible reason to accelerate (though the top eighth will have difficult pairings and probably not as much sleep as those just behind them).

Note that if there is a 32-player 5-round event (which doesn’t need to be accelerated) and you opt to accelerate for closer pairings then the broad middle sees little difference. If the ratings are perfectly predictable, the higher-rated wins every game, and you don’t do shifts for color (to make the matchups easier to calculate) then some of the differences between standard and accelerated are (with standard listed first):
1 plays 17 instead of 4 (unlisted opponents are the same for both methods)
5 plays 9, 11 and 21 instead of 8, 12 and 17
8 plays 3, 14 and 24 instead of 5, 15 and 20
9 plays 5, 15 and 25 instead of 10, 16 and 21
11 plays 5 and 27 instead of 4 and 23
13 plays 7 and 29 instead of 6 and 25
16 plays 10 and 32 instead of 9 and 28

So accelerating for better matchups will only affect a few of the players and leave the bulk of them wondering why you bothered. Also, if you accelerate and have a number of round two upsets (with the lower rated 1-0 beating the higher rated 0-1) then you risk finishing with multiple perfect scores when standard pairings would have prevented that. An accelerated 32-player event could have twelve 2-0 players with three 4-0 players going into the final round (5) instead of standard pairings having at most eight 2-0 players with two 4-0 players going into the final round.

30

Alas, a snowstorm severely impacted attendance for my tournament yesterday and, as a result, I was unable to try out my pairing style on live human beings.

Back to the programming, and get some simulations running Real Soon Now.

31

As I said before, before you worry about the programming, just do the pairings by hand. It may turn out you need to make some changes in your ideas, which would affect the programming.

Bill Smythe