A different pairing method

Running Chess Tournaments
1

I haven’t been satisfied with standard Swiss pairings for my tournaments. I’ve decided on something that I think will be better. I want to run it past people, say why I like it, and see if anyone has either a better recommendation, or spots a severe flaw that I have overlooked.

First, let me say something about my tournaments. They are small. In the past, they have been anywhere from 12 to 20 players. Second, there is no cash prize. Third, the range of that set of players runs in rating from around 300 to around 2000. At the last tournament it was about 1/3 under 900, and 1/3 under 1500, with always at least one, and up to four, unrateds. The after tournament rating of the unrateds has ranged from 100 to 1895. They all play in one section, because breaking up such a small group does not work well. It’s a five round Swiss tournament. When we get to around 16, I could break it up, but I never know if I’ll get that many in advance, and I always feel very bad for that 1200 rated player who would get stuck at the top of the bottom section, and have no more than one worthwhile game all day.

On the other hand, the first round of such a tournament is practically a waste of time, and the kids with 600 ratings aren’t really keen about playing 1400 rated adults. What to do.

My first thought was to just use 1-2, 3-4, pairings, but that puts the top two players against each in the first round. So, I’m going to modify those pairings a bit. If there are N rounds remaining, the top N players cannot play against each other, and the bottom N players cannot play each other. Otherwise, it’s 1-2,3-4 pairings. If there are less than 2*N players, it gets a little screwy, because the protected zones at the top and bottom overlap. For a 24 player, five round, tournament, that means the first round is

1 - 6
2 - 7
3 - 8
4- 9
5 -10 (top five players can’t play each other)
11 -12
13- 14
15 - 20(bottom five can’t play)
16 - 21
17- 22
18- 23
19- 24

For a 14 player tournament, it looks like this:

1 6
2 7
3 10
4 11
5 12
8 13
9 14

In the second round, only four players are protected against playing each other. The ranking of the players will be based on points first, then rating, just like a standard Swiss

I did some simulation, and the average gap between players is smaller, and the extreme gaps are usually avoided. Of course, if there are dramatic upsets, that can lead to dramatic differences in ratngs in a game, but that only happens when players are playing well outside their performance levels.

If all goes according to plan, the top player will play players 6,5,4,3, and 2, in order.

The advantage is the elimination of the huge rating difference games from the first round of the Swiss.

The disadvantage is that the top player has a much rougher day than if he played a standard Swiss. No relaxing during rounds 1 and 2, and whoever he meets at the end of the day will have played a noticably easier schedule than he. If we were allocating cash prizes among high ranked players, that would be a problem, but we aren’t.

Color equalization would not be part of deciding which opponents to play, but it will be used to decide who plays white or black in a given game.

Thoughts? Feedback? Have I overlooked an obvious flaw?

2

Last Fall we ran what we called The Untournament with the following design.

The Untournament – USCF rated Game 90

Swiss Pairings will not be used.

Pairings will pair those closest in ratings while trying

to maintain color allocation.

The winner will be the person that gains the most rating points.

This was run over the course of 4 Mondays. We even ignored score groups when pairing. Initially it looked like we basically were going to end up with Quadish divisons, but due to different people showing up that didn’t quite happen. There were no prizes and the declared winner was actually the individual that gained the most USCF rating points. This actually ended up favoring the lower rated players.

Anyway I would think the key thing here is to not be afraid of trying something different every now and then.

3

Interesting, I assume that means you didn’t determine the winner until the event was rated by the USCF. (Did you wait until after the first rerate?)

4

The biggest rating gain was 76 points vs 2nd at 33 points.

5

I like most of your ideas, at least for the type of tournaments you are running (small events with no prizes).

Martin Morrison, in the first two editions of the rulebook, would still call your tournament a Swiss, since players with similar scores are still playing each other. He would not, however, call it a “ratings-controlled” Swiss. Actually, though, it is ratings-controlled, since the ratings (within equal score groups) still control the pairings, albeit not in the conventional manner.

I agree. With 16-20 players, pairings in round 5 become pretty screwy anyway. Leave it as one section. (With only three rounds, you could divide it into sections of about 8 players.)

In other words, you are placing the top five players on boards 1-5, and the bottom five on boards N-4 through N, and assigning the remaining players using as close to 1-vs-2 pairings as possible. (N is the number of pairings, not the number of players. With 24 players there are 12 pairings.)

Same observation – even though, now boards 1-5 overlap boards N-4 through N.

Things won’t go according to plan. They never do. But that’s still OK.

So it would seem, yes.

From the top player’s standpoint, I would think that would be an advantage, not a disadvantage. Without prizes, the top player’s only reason to live is the prospect of being paired against challenging opponents.

This is the only part I disagree with, and only partially. I agree that color alternation should not be taken into consideration in small tournaments. (Use an alternation limit of 0 instead of 80.) I have argued this point before, in several older threads. If you have some bad alternations in round 3, then good pairings (and good colors) will become easier in the subsequent rounds.

However, it would seem reasonable, in rounds 2 and 4, to equalize colors whenever practical. Also, in round 1 you should still alternate down the boards, as usual.

Fear not – try out your idea in your next small, no-prize club tournament. Just tell your players in advance. There’s nothing like a real-life test to work out any kinks.

Bill Smythe

6

Exactly my thought. If there were prizes on the line, then the top player might feel cheated if he fell to a lower rated player who played a much easier schedule. I don’t think this style would work well for any tournament with anything more than a nominal prize. As it is, the effect of this pairing system will probably be that he doesn’t have to play an 800 point rating difference in round 1. It’s not uncommon at my tournaments for most of the round 1 games to be finished within 30 minutes.

I’ll be fiddling with color allocation rules a bit, to see if I can come up with a good method. I’ll probably just stick with the alternating scheme for round 1. One factor will be whether I can get this programmed in the next few weeks and have a computer do it. If not, the system is simple enough that it can be implemented pretty easily with pairing cards.

Thanks for the feedback.

7
8

[quote=“Meadmaker”]
I haven’t been satisfied with standard Swiss pairings for my tournaments. I’ve decided on something that I think will be better. I want to run it past people, say why I like it, and see if anyone has either a better recommendation, or spots a severe flaw that I have overlooked.

First, let me say something about my tournaments. They are small. In the past, they have been anywhere from 12 to 20 players. Second, there is no cash prize. Third, the range of that set of players runs in rating from around 300 to around 2000. At the last tournament it was about 1/3 under 900, and 1/3 under 1500, with always at least one, and up to four, unrateds. The after tournament rating of the unrateds has ranged from 100 to 1895. They all play in one section, because breaking up such a small group does not work well. It’s a five round Swiss tournament. When we get to around 16, I could break it up, but I never know if I’ll get that many in advance, and I always feel very bad for that 1200 rated player who would get stuck at the top of the bottom section, and have no more than one worthwhile game all day.

On the other hand, the first round of such a tournament is practically a waste of time, and the kids with 600 ratings aren’t really keen about playing 1400 rated adults. What to do.

My first thought was to just use 1-2, 3-4, pairings, but that puts the top two players against each in the first round. So, I’m going to modify those pairings a bit. If there are N rounds remaining, the top N players cannot play against each other, and the bottom N players cannot play each other. Otherwise, it’s 1-2,3-4 pairings. If there are less than 2*N players, it gets a little screwy, because the protected zones at the top and bottom overlap. For a 24 player, five round, tournament, that means the first round is

1 - 6
2 - 7
3 - 8
4- 9
5 -10 (top five players can’t play each other)
11 -12
13- 14
15 - 20(bottom five can’t play)
16 - 21
17- 22
18- 23
19- 24

For a 14 player tournament, it looks like this:

1 6
2 7
3 10
4 11
5 12
8 13
9 14

In the second round, only four players are protected against playing each other. The ranking of the players will be based on points first, then rating, just like a standard Swiss

I did some simulation, and the average gap between players is smaller, and the extreme gaps are usually avoided. Of course, if there are dramatic upsets, that can lead to dramatic differences in ratngs in a game, but that only happens when players are playing well outside their performance levels.

If all goes according to plan, the top player will play players 6,5,4,3, and 2, in order.

The advantage is the elimination of the huge rating difference games from the first round of the Swiss.

The disadvantage is that the top player has a much rougher day than if he played a standard Swiss. No relaxing during rounds 1 and 2, and whoever he meets at the end of the day will have played a noticably easier schedule than he. If we were allocating cash prizes among high ranked players, that would be a problem, but we aren’t.

Color equalization would not be part of deciding which opponents to play, but it will be used to decide who plays white or black in a given game.

Thoughts? Feedback? Have I overlooked an obvious flaw?[/quote

I would prefer to decrease the number of rounds to three or 4 as opposed to
future pairing of established players, especially, scholastic, with a huge rating
disparity. Further, I have in many such tournaments left the number of sections,
as well as number of rounds to depend on the number and ability level of those
playing.

It is not at all uncommon, say, for a new player to emerge from a U300 section
with a 700+ rating. To then say, that this player cannot play say in a U600
section his next tournament would be ridiculous as there has been no indication
that as he/she has not defeated anyone at this level, that in fact that the 700+ rating
has much meaning. In this, simply pair as UNR, or assign a rating for pairing
purposes. (that is if current ratings are used at all, or if the 700+ rating appears
on the monthly supplement.)

I think you have hit on the heart of the matter–as tournament directors/organizers,
a key concern has to be the enjoyment of the kids attending, and 300 vs 1000+
pairings rarely serve any useful purpose–boredom for one, humiliation for the other.

Say I have 10 kids, 6 900+, the rest U500. Run a quad for the U500, and a 4 round
Swiss for the 900+. This way, the newbies, and novice are not destroyed, have a
shot at some initial success-so important for both Mommy and the player, and
GREATLY increases the odds of the said player returning. And it gives the 900+
players a greater chance to grow in their skills.

9

You might be interested in checking out the McMahon system. Basically it’s a Swiss in which stronger players start the tournament with more than 0 points.

10

You can also use a version of 52 pickup. Before the first round throw all the pairing cards in the air. Pair the closest cards or those cards who touch. The first card of the pair you pick up is white. In subsequent rounds, take each score group and throw those up in the air and pair as above. Or you could pay no attention to score and throw all of the cards up in the air again. Of course, this may ignore rating and color allocation. But if the players are willing to participate in a tournament with a lot of randomness, this is another method.

11

Only about 1/3 of my players are kids, so I don’t want to taylor the format for them too much. I’m not all that picky about them coming back if the truth be told. Some do. Most don’t. All of my “regulars” are teens and adults, mostly in the 1000-1500 range, with a few higher.

The problem with any variation on splitting sections is that the person at the top spot of the bottom section gets really stuck with boredom. He usually gets compensation in the form of better prize chances, but that won’t work at my events. At my events, the play has to be its own reward.

I have thought about cutting to four rounds, but then if I have more than 16 players, it doesn’t work well. A lot of people like quads, but for me, I want more than three games if I’m going to wreck my whole day anyway, and I think quads are mostly catering to competition for prizes, anyway. So many tradeoffs. What to do?

This seems like a nice compromise that still has a “tournament” feel to it, with winners and losers, but avoids the rated Swiss time wasting competitions.

12

PLEASE don’t worry about programming it just yet. Any new idea ought to be tested “by hand” at first.

For the life of me, I cannot figure out why any TD would use a computer in a tournament with fewer than 20 players, even with standard Swiss pairings.

In general, I think it’s a poor idea to have the number of rounds, number of sections, etc depend on the size and/or strength of the tournament. You can get a reputation this way (not a good one). Announce what you’re going to do in advance of the tournament, and run the tournament in the manner announced if at all possible.

Bill Smythe

13

There are good reasons to do this. However, most of them don’t really relate to actually making pairings. The following is a small sample of what I mean.

It makes submitting the event for rating very simple at its conclusion. Exporting a final wall chart and/or prize report to HTML for web display is quite trivial as well. It’s much easier to keep and review past tournament files if they’re just on a directory, rather than stuffed in a file drawer. If you have a frequent player database, setting up a tournament on the computer would be much easier than doing all the data entry by hand. Finally, I imagine many directors don’t have pairing cards on hand anymore. :slight_smile:

I will note that it is standard practice for me to check the top several boards in any tournament I’m running with a pairing program. So, checking the program’s work in a 20-player event should not be too taxing.

I completely agree with this.

(EDIT: fixed spelling issue. It seems speech-to-text still requires proof reading. :laughing:)

14

I have long thought that a pairing variant could be the “inverse Swiss” – in esence doing the score group seeding similar to brackets. So the first round of a 12 player event would have 1-12 and 6-7 rather than 1-7 and 6-12. There are certainly some arguments against this approach…but the middle games are more interesting early and it may do a better job of dividing the pool. For the right kind of club pool it may be interesting.

McMahon system was mentioned. I did a post awhile ago on various systems including McMahon. I’ll see if I can find and link it.

Update: Here is the link: viewtopic.php?p=245805#p245805

also see
uschess.org/forums/viewtopic … on#p237291

15

One of the advantages of having a “different” pairing method is that no one knows what is going on. That leads to dozens of questions about the pairing method and people leaning over your shoulder while you are doing pairings. After you put the pairings up late you then get to field more questions, especially about when the next round is going to start since this one was late. At least you won’t have to worry about the consistency of the tiebreaks when the event is finished.

16

Wouldn’t accelerated pairings for a round or two achieve an acceptable result?

17

The primary (possibly only) purpose of accelerated pairings is to reduce the number of perfect scores by getting the high rated players (the ones most likely to have perfect scores) to play each other sooner.

A number of experienced TDs have noted that with non-accelerated pairings the first round is often a turkey shoot, but with accelerated pairings it is the 3rd round (if it is a 2 round acceleration) that is likely to be the turkey shoot.

It has also seemed to me that tiebreaks for class prizes are negatively impacted by accelerated pairings, especially for players in the the middle of the pack overall, but I don’t know how to test that feeling.

18

I know that is the purpose of accelerated pairings, but it seems that using them would achieve the desired goal of not having huge disparities in rating for the first rounds.

19

Sometimes it seems like accelerated pairings just move the disparities into the first round after acceleration, and sometimes results in a few even bigger disparities.

I played in a 200+ player event with two rounds of acceleration once, and scored 2-0 (both moderate upsets, as I recall.)

So, in round 3 I was up on about board 4 against the reigning Nebraska champion, a master rated over 800 points higher than me. I was hosed by move 10.

Assuming ‘normal’ results, had the event not used accelerated pairings, a 1400 player like me with a 2-0 score would probably have gotten a low Expert in round 3, about a 300 point ratings difference compared to the pairing I got, after pairings against approximately A players in rounds 1 and 2.

20

Perhaps, but a 1400 player would not have gotten to 2-0 w/o first beating folks 800 points above him in the first or second round.

The question here is, would such disparities happen in a much smaller tournament using accelerated pairings?