Alternate pairing systems

Running Chess Tournaments
1

About three decades ago, Chicago organizer Richard Verber (now deceased) ran a monster (by 1970s standards) scholastic tournament – about 300 players, 1 section, 5 rounds.

His TD was Tim Redman, a straight-laced, by-the-book kind of TD (after all, he wrote one of the editions of the USCF rulebook).

Redman’s pairings were straight top-half-vs-bottom-half, right out of the rulebook. After 3 rounds, with only 2 rounds to go, there were still about two dozen players with perfect 3-0 scores. Most of these had ratings far from the top, and many would likely have finished 5-0 without ever facing serious competition.

Verber, an imaginative fellow and never a slave to the rulebook, hit upon a brilliant idea to force contenders to actually contend, and to reduce drastically the number of perfect scores.

Instead of simply pairing the bottom 3-0 against the top 2.5 (as would be standard when there is an odd player), Verber paired just a few of the 3-0 players (the highest-rated ones) against each other. He then paired the remaining 20 (or so) 3-0 players against high-rated players with 2.5, or even 2.0 (maybe even a few with 1.5).

The result? After round 4, there were only three perfect scores remaining. After round 5, there were none – the top two drew each other, while the third lost to a higher-rated player in a lower score group.

Comments?

Bill Smythe

2

The first question, higher rated and low rated players of the 1970’s and now in the early 21st century are so much different. Scholastic players do play a great deal more then they did in the 1970’s. The markerting of scholastic players have so much changed from the 1970’s to the present.

With the long time between one scholastic tournament to the next, and having a state title tournament for a scholastic event: when the parents are more willing to coach and train their child or children. Even when the parents are so willing to have adults train their child, with the coaching from top ranking players can make the child be able to play at higher performance ratings then their true USCF ratings. Lets take a example, if having a child and have a master teach my child 8 hours a week for a year. Very sure my childs scholastic rating should be much stronger then 400 after one year, in fact should be around a class C or better.

There are two schools of thought on this issue, have the child start to play open tournaments with all age groups – like the state championship. The childs rating would in time become class C or even higher. The other school, only have the child only play in scholastic tournaments just to keep a low rating. Have the child go to the scholastic tournament like the ‘Super Nationals’, with a 400 rating and have in fact be as strong as 1500 or better.

Having a tournament when the low rating player has 3.0 and the higher rated player has 1.5 – is not the same as it was in the 1970’s as stated above. Sure there are those 400 players that should have a 1500 rating without blinking a eye. Having them play someone with a score of 1.5 or 2.0 because they have a higher scholastic rating, and not someone with a 3.0 score group as themselves the times have changed.

There are a number of parents, that do want to keep their childs rating low. It would be like having a master being able to win the under 1200 section in a open tournament. If the tournament is 5 rounds, the master should be after the over all title with a perfect score: the under 1200 should be able to win the class prize with 1.5 or 2.0 out of 5.0 points. If having a 400 player thats true rating is around a class C player or better, should not have a problem to win that under 600 trophy.

Bills idea could have worked fine during the 1970’s, at present with scolastic players being so under value for the ones that do so for that trophy. Parents are more looking for the silly trophy then having their child play at there best. What would have happened too Bobby Fisher if he only was in scholastic tournaments – he would never became a Grand Master or even World Champion.

Parents are in so much in demand to have their child win a trophy at any cost, even to have their childs true rating be much lower with the understanding a lower rating is more able to win some kind of trophy. Bills idea is a great idea, the trophy hunters have killed his theory.

3

Bill-

Interesting idea. The current uscf national scholastic events use a similar method of pairing large sections. However it’s slightly different in that they do the “filtering” on the front end of the tourney.

It’s called accelerated pairing. Instead of pairing the top half with the bottom half the first round, they pair the top quarter with the second top quarter for the first two rounds. The third quarter is paired with the bottom quarter the first two rounds. Within these “2 seperate” brackets normal swiss rules apply.

After the first two rounds are over, the entire section (all 4 quarters) are paired together using standard swiss rules from the results of the student’s first two rounds.

What this effectively does is let the top players play each other like the event has already played two rounds (assuming the higher rated player always win). The top most players eliminate many perfect scores among themselves during these first two rounds. On round 3 and 4 the top players knock out the lower bracket’s perfect score top players. Then the final round or two the last of the perfect scores play each other.

My own son has played in the primary championship section of nationals for 2 years now. He started each time in the top quarter. His first two rounds would be real games from the get-go. Then round 3 was usually a “gimme”. Round 5 would be another battle. And finally, round 6 and 7 would determine the national champ.

While I agree this works very well to ensure that only 1 or 2 national champs will be crowned in each section which is good, it’s really lousy at determining who is 3rd-25th etc. But swiss has never been good at that.

just my $.02

4

Sounds like partial class pairings. But in the 4th round? It certainly fixed a problem, but I was under the assumption that partial class pairings would have to be done in the final round.

I’ve always wanted to do partial class in the final round where it would eliminate many ties. But not having advertised the possibility in pre-pub ads, I’ve never done it. I’m too lazy to look in the rulebook right now, but can a TD do partial class without advance notice? Is this a makor variation?

Any other ideas?

5

I’ve always been so turned off by the idea of class pairings, whether “partial” or otherwise, that I’ve never even tried to understand the idea thoroughly, let alone try to figure out whether it is a major or minor variation.

But, unless I’m mistaken, even with class pairings, you still pair players with equal scores whenever possible. Richard Verber’s idea was quite different – he paired higher-scoring, lower-rated players vs lower-scoring, higher-rated opponents. That way, you could eliminate perfect scores (the aforementioned “players”) without “using up” other perfect scores (the aforementioned “opponents”). E.g. if a 3-0 loses to a 2-1, rather than to another 3-0, you have killed one bird with zero stones, much better than killing one of two birds with one stone.

That idea is also present in the accelerated (quarter) pairings modification (28R), where in round 2 you pair top-half losers vs bottom-half winners. Verber simply took accelerated pairings to extremes, and it worked brilliantly.

Bill Smythe

6

Oops. Are you sure that’s what they did? Usually, “accelerated pairing” means to pair by quarters in round 1, but in round 2 top-half losers play bottom-half winners. (Top half winners play each other, as do bottom-half losers.)

That way, if the top-half losers beat the bottom-half winners, you end up with only 1/8 of the players, rather than 1/4, with a perfect score after 2 rounds. Your method, by contrast, would still pair winners vs winners, so you’d still end up with 1/4 perfect scores. Granted, you’ll have better matches in the early rounds, but now the later rounds may tend to feel like the earlier rounds do with “standard” pairings.

Bill Smythe

7

But now back to the original topic of alternate pairing systems.

Another alternate pairing system is the 1-vs-2 swiss. In this system, players with equal scores are paired whenever possible, just as in a standard swiss. But, within each score group, instead of top half vs bottom half, you pair 1 vs 2, then 3 vs 4, then 5 vs 6, etc.

In round 1, you simply pair 1 vs 2, and 3 vs 4, etc. You can alternate colors (if 1 gets white, then 3 gets black) just as in a standard swiss. And in subsequent rounds, you make minor transpositions to improve colors, just as in a standard swiss.

The 1-vs-2 pairing system is mentioned in the 5th edition of the rulebook, as Variation 29L1. For some reason the rulebook suggests it as an alternative to “using round robin table in small swiss”, but it also recommends it as a recreational possibility for club tournaments.

Unlike Richard Verber’s imaginative pairings, 1-vs-2 will not result in fewer perfect scores than standard pairings (except that there may be a few more draws). But it will result in more interesting match-ups in the early rounds.

Bill Smythe

8

Thinking of Bills idea when he started this talked of higher rated players with lower scores play weaker players with higher scores. Having scholastic players, the rating of any scholastic player under 900 is a little hard too tell who is stronger or weaker. Making a claim that a rating of 500 being stronger then a 300 rating: only looking at points for a cause and effect.

The other problem, his idea would be acceptable if it is between players within rating range of 2000 - 1200, as how many class A players or stronger withdraw after the first round: when they lose to someone having a much weaker rating then their own. The class A player is strong, and should be still in the run with the 1 point group even having zero points for the second round pairings.

The other problem with Bills idea, as he only debates it with scholastic players, the majority would be within the 1000 - 100 rating level. Not to make a bias, the tournament would need to be acceptable with any rating, not a set of ratings in a established group. Stronger players will be at a tournament, for some personal or health reasons – the player is not at the norm of their established ratings. This is the reason why the stronger player would withdraw after the first round loss. The bias of only having Bills system at a scholastic tournament, and not a tournament like the New York Masters – would be a clear bias. The rules are designed for all tournament players, as it would be a bias for having a rule for masters and a rule for scholastic players: example, different rules for touch move.

Will ask Bill this question: “the system you state for the scholastic players would you also use this same system at the US Open(?)”, this is my question? If you can use it for scholastic players only, and not for other tournaments, ‘do you see the bias’ of this system.

9

I agree that Verber pairings (high-scoring low-rated vs high-rated low-scoring) would not work as well in today’s scholastic environment as it did in the 1970s, when most scholastic tournament players were rated over 1000 and a few were rated almost 2000.

I also did not intend to convey the message that this idea should be used only in scholastic tournaments. The example I remember just happened to be a scholastic tournament. Indeed, nowadays a scholastic tournament might be the WORST place for such an experiment.

Actually, I’m not even sure I’m recommending the idea at all. I just threw it out as an interesting possibility to generate some discussion.

The idea would seem to work best when most ratings are 1000 or higher, and when there are a significant number of players in each rating class up through A or higher. It would not work too well in a class tournament where everybody is rated about the same – the end result would look a lot like a regular swiss or a 1-vs-2 swiss. It would also not work well if many of the ratings are unreliable, as they often are in present-day scholastic events.

In other words, the ideal tournament for Verber pairings might be – as Doug suggests – the U.S. Open!

Bill Smythe

10

Just out of curiosity, if someone wanted to use that system, would it have to be announced in the event publicity? If it does, would “1 vs 2 pairings used” be a sufficient announcement?

-ed g.

11

Yes, it needs to be public before the start of the round and any information with the tournament. If it in a flyer, the organizer needs to address the information as clear as it can be. Just having 100 words on the pairings format, still will have players still woundering what is going on.

Even established players have questions on the standard formating of pairings. With 1 vs 2 pairings, or any other pairing formating – you will have players asking questions. It is always best to be up front with the players, before having something new.

12

Yes, “1 vs. 2 pairings used” would be okay, although you might want to add the relevant USCF rule (Variation 29L1) in the ad.

13

It would certainly have to be announced in all pre-tournament publicity. Given that 1-vs-2 pairings are described in the rulebook, I would think “1 vs 2 pairings used” would be sufficient, although just to be sure, you might want to also cite the rule number or page number from the 5th edition rulebook.

Bill Smythe

14

What Bill said! :stuck_out_tongue:

15

And now, fellow chess players and organizers, we come to the REAL reason I started this “Alternate pairing systems” conversation.

I would like to combine Verber pairings (described in the first post) with 1-vs-2 pairings, and call the result “Modified Verber 1-vs-2 Pairings”.

The idea of Verber pairings is to pair lower-rated, higher-scoring players against higher-rated, lower-scoring opponents, in cases where the rating difference is sufficient to make it extremely likely that the higher-rated, lower-scoring player will win. In this way, you can use a player with a less than perfect score to knock off a player with a perfect score, so that neither of them will be perfect anymore. This should reduce the number of perfect scores MUCH faster than a standard swiss (or even a quarter-paired swiss).

Specifically, my proposal for combined Verber and 1-vs-2 pairings is as follows:

In round 1, pair 1 vs 2, and 3 vs 4, and 5 vs 6, etc. Alternate the colors, as usual – if 1 has white, then 3 should have black, 5 should have white, etc.

In subsequent rounds, add 300 points to each player’s rating for each win, and subtract 300 for each loss. Again pair the rounds 1 vs 2, but ignore the score, and use the adjusted ratings instead of the original ratings. Make slight transpositions for the usual reasons – to avoid pairing the same players twice, or to improve colors.

Note that pairings are NOT done by score, only by adjusted rating. There are no “score groups” – or, you might say there is just one big “score group” in each round.

This plan quantifies the Verber idea, and creates an algorithm for carrying it out. In round 1, the 1-vs-2 pairings will create lots of high-rated losers, and lots of low-rated winners, ripe for pairing against each other in subsequent rounds.

Below is a sample crosstable paired with this idea. You will note that, with 3 rounds and 16 players, there is only one perfect score, despite the total lack of draws, byes, forfeits, etc.

The columns in the crosstable are: Number, rating, result, adjusted rating, result, adjusted rating, result, colors.

01 2300 W02 2600 W03 2900 W05 wbw
02 2200 L01 1900 W09 2200 W03 bwb
03 2100 W04 2400 L01 2100 L02 bww
04 2000 L03 1700 W11 2000 W06 wbw
05 1900 W06 2200 W07 2500 L01 wbb
06 1800 L05 1500 W13 1800 L04 bwb
07 1700 W08 2000 L05 1700 W09 bwb
08 1600 L07 1300 W15 1600 W10 wbw
09 1500 W10 1800 L02 1500 L07 wbw
10 1400 L09 1100 W12 1400 L08 bwb
11 1300 W12 1600 L04 1300 W13 bwb
12 1200 L11 0900 L10 0600 W16 wbw
13 1100 W14 1400 L06 1100 L11 wbw
14 1000 L13 0700 W16 1000 W15 bwb
15 0900 W16 1200 L08 0900 L14 bww
16 0800 L15 0500 L14 0200 L12 wbb

Bill Smythe

16

Bill, with no upsets simple acceleration will also bring 16 players down to 1 perfect score in three rounds:
Rd 1) 1-4 beat 5-8 and 9-12 beat 13-16
Rd 2) 1-2 beat 3-4, 5-8 beat 9-12, 13-14 beat 15-16
Rd 3) 1 beats 2, 3-8 beat 9-14, 15 beats 16
1@3-0. 7@2-1, 7@1-2, 1@0-3

However, if you double the 16 players to 32 then you can see the difference. Simple acceleration of 32 players with no upsets allows 2 perfect scores (2@3-0, 14@2-1, 14@1-2, 2@0-3). The +300/-300 1vs2 method, however, gives:
01 2350 w2+(2650)b3+(2950)w5+ 3
02 2300 b1-(2000)w13+(2300)b9+ 2
03 2250 b4+(2550)w1-(2250)b8+ 2
04 2200 w3-(1900)b15+(2200)w6+ 2
05 2150 w6+(2450)b7+(2750)b1- 2
06 2100 b5-(1800)w17+(2100)b4- 1
07 2050 b8+(2350)w5-(2050)b12+ 2
08 2000 w7-(1700)b19+(2000)w3- 1
09 1950 w10+(2250)b11+(2550)w2- 2
10 1900 b9-(1600)w21+(1900)b11+ 2
11 1850 b12+(2150)w9-(1850)w10- 1
12 1800 w11-(1500)b23+(1800)w7- 1
13 1750 w14+(2050)b2-(1750)w15+ 2
14 1700 b13-(1400)w25+(1700)b16+ 2
15 1650 b16+(1950)w4-(1650)b13- 1
16 1600 w15-(1300)b27+(1600)w14- 1
17 1550 w18+(1850)b6-(1550)w19+ 2
18 1500 b17-(1200)w29+(1500)b20+ 2
19 1450 b20+(1750)w8-(1450)b17- 1
20 1400 w19-(1100)b31+(1400)w18- 1
21 1350 w22+(1650)b10-(1350)w23+ 2
22 1300 b21-(1000)w24+(1300)b25+ 2
23 1250 b24+(1550)w12-(1250)b21- 1
24 1200 w23-(900)b22-(600)w30+ 1
25 1150 w26+(1450)b14-(1150)w22- 1
26 1100 b25-(800)w28+(1100)b27+ 2
27 1050 b28+(1350)w16-(1050)w26- 1
28 1000 w27-(700)b26-(400)w32+ 1
29 950 w30+(1250)b18-(950)w31+ 2
30 900 b29-(600)w32+(900)b24- 1
31 850 b32+(1150)w20-(850)b29- 1
32 800 w31-(500)b30-(200)b28- 0

1@3-0, 15@2-1, 15@1-2, 1@0-3

That is reducing perfect scores one round faster than simple acceleration, and the number of draws is likely to be much higher than with simple acceleration. Even using sextile acceleration rather than quartile acceleration does not reduce perfect scores as fast. Note that in this example the lower rated 1-0 players paired with higher rated 0-1 players are facing a 550 point rating deficit while in simple acceleration that deficit would be only 400 points, and thus the chances of a round two upset allowing a lower half player to reach 2-0 are less than with acceleration.

One significant drawback is that the distribution of the twos and ones is nowhere near what standard or accelerated pairings would give, and thus the luck of the draw (the 2100 is paired up twice in three rounds and finishes with 1 point while the 950 is paired down twice and finishes with 2 points) has even more of a chance of affecting class prizes than simple acceleration does. Another drawback is that colors and prior opponent considerations becomes a problem quickly (note that there were multiple potential round 3 pairings depending on the determination to avoid color conflicts, and the ones I quickly chose may not be optimal).

Bill, if you run a tournament with this type of pairing, let us know how it turns out.