Team Tournament Pairings

Running Chess Tournaments
21

They would probably end up in 5th place based on game points. If the 396 player was treated as being only 400 points lower than board three, then this team could not play in the USAT, assuming they still use this rule.

Will they still have cake and all of the unusual prizes at the USAT-East this year? I have friends who still talk about the year the tournament had to deal with a blizzard like the recent Nemo storm. They got home two days after the tournament ended because the Pa Turnpike was closed.

22

Something like this has happened before. GM Anatoly Karpov, GM Ron Henley, IM Irina Krush and young Albert Pinnella won the 1998 USATE.

Of course, all these hypothetical teams involve massive differences in rating between first and last board. The easy way to avoid these hypotheticals (or, for that matter, the Karpov superteam) is to put a cap on how much that difference can be. The USATE used to have a cap of 1,000 points on that difference. I believe it was repealed in 1998, specifically to allow Karpov’s ultra-loaded team to play. I’m not sure if it has been renewed in any form since then.

23

Forgive me, but this still doesn’t seem like a very logical way to pair large score groups. For example, in a 100-team tournament, after 3 rounds, you would expect maybe 20-30% of the field to have 2 match points, so 20-30 teams. If there were no seeding at all, then I could see completely re-randomizing the score group. However, the pods being used indicate that there is believed to be some difference, or hierarchy, between teams.

I apologize if there’s something I don’t understand about this system. I readily admit I’ve never attempted to rank or pair players or teams in this way. I have to admit, I’m glad for that. :slight_smile:

24

Consider how tennis handles its brackets. Despite having a ranking system designed rather specifically to give a relatively current rank ordering of players (unlike USCF ratings, which can be rather stale), in making the brackets, seeds 1-2 are effectively identical as are the groups of 3-4, 5-8, 9-16, 17-32. Everybody else, from 33 down to the lowest qualifier are randomized into the brackets. The seeding committee thus doesn’t, by its decisions alone, make the brackets.

That’s not all that different from how the IHSA seeding process works. As with tennis, it’s much easier to rank order the top players (teams) who get deep into tournaments and often end up playing each other than to do the same to the large number of teams who lose early and play in the bigger 1-2 and 2-1 scoregroups, or who play in weaker tournaments or weaker conferences.

Note that I’m not saying the seedings should be ignored. IHSA rank orders 1-16, and randomizes only within groups 17-24, 25-32, etc. (with larger groups for the bottom half). That’s the current practice—I believe that at one point there was randomization within 1-7 and 8-14.

What I’m saying is that if you are randomizing in a group because the information is weak about the specific order within that group, then you should probably re-randomize (within each of those groups) in subsequent rounds as that’s the standard way of handling players with identical ratings in Swiss pairings.

25

Just to clarify, if 1-32 are 2-0, 33-96 are 1-1 and 97-128 are 0-2 then Tom is saying that the randomization is within the
17-24 group (all playing 1-8 but random as far as which one)
25-32 group (all playing 9-16 but random as far as which one)
33-40 group (all playing 65-72 but random as far as which one)
41-48 group (all playing 73-80 but random as far as which one)
49-56 group (all playing 81-88 but random as far as which one)
57-64 group (all playing 89-96 but random as far as which one)
97-104 group (all playing 113-120 but random as far as which one)
105-112 group (all playing 121-128 but random as far as which one)

It won’t be quite that neat because there will be upsets and the top 8 1-1 teams will not all be in the same pod (maybe the first 2 are in one pod, the next 5 are in the next, the next 7 are in the next, etc., where the randomization of the middle pod could put some teams at the bottom of the top half of the score group and others at the top of the bottom half of the score group).

26

There was most definitely randomization in the top pod(s) back when I played/coached, and that was the crux of my beef with the directors. I am glad that the top teams are no longer randomized. I can certainly see the case for randomizing the bottom 80% or so of an IHSA field, as those teams are not all that distinguishable from one another, especially in terms of their potential effect on the top three places.

I completely agree with this. As can be gleaned from the above, my problem was that IHSA was randomizing teams where there was plenty of known information. I didn’t think that was logically defensible.

27

The IHSA team tournament is first seeded by representatives of the leagues one week before the tournament. If there are over 128 teams, the top 64 teams are broken into 8 groups of 8. For the top two groups the teams are rank ordered based on the group consensus. The remaining groups are set by consensus and then teams within the group are randomized. The teams below 64 are split into 4 equal groups and randomized within each group. The teams are assigned a rating based on their seeding (using negative numbers so team number 1 has a rating of -1 team 2 is -2) By the current rules teams do not play other teams in the same league provided they have less than 4 points. Above 4 points that restriction is off. We also set the restrictions for alternation to 8 rating points to reflect one group and for equalization to 16 to reflect two groups.

28

I see one issue with the assigned ratings and alternation/equalization limits. In the top two groups those limits for alternation allow a 1 vs 9 swap, but not a 1 vs 10 swap, which seems fine because there is a determined difference between 9 and 10.
However, a third group vs fourth group swap is more problematic in that 20 vs 28 can swap but not 20 vs 29 even though 29 might actually be stronger than 28. The 24th strongest team assigned number 17 would be limited in alternations to number 25 (possibly the strongest of that set of eight) and could be alternated with number 9, while the 17th strongest team assigned 24 could alternate with any of the next eight including the weakest and could not have an alternation swap with a team stronger than number 16.

A possible solution is to keep the sequence but change the numbers for the assigned ratings. Have the top eight get -1 through -8, the second eight get -141 through -148, the next six sets of eight go from -201 through -208 to -701 through -708, and the remainder split into four groups from -801 through -840 to -1101 through -1140. Then you can set the alternation/equalization limits to 140/240 and keep the same restrictions on the top two groups while allowing freer alternations within the lower, randomized groups.

29

The idea of adjusting the ratings is interesting. The first 16 teams we seed exactly but in the other groups they are grouped and then randomized. In the lowest groups you have over 16 in a group randomized so switching among any of those would seem to be just fine. The trouble with your system is that 8 is then dramatically different from 9. So I could not switch 8 and 9 for equalization even though they are one spot apart. It would just take some more thinking to determine what the best rating system would be to allow for the goal. However realistically there is not a problem with this since colors do not usually work out very badly.

30

In WinTD if players have identical ratings does it re-randomzie the order in each round?

31

The 2008 USATE was won gy GGGg, comprising GM Eugene Perelshteyn, GM Zviad Izoria, GM Roman Dzindzichasvili, and Stephen Fanning, age 5, rated 178. http://www.uschess.org/content/view/8205/436/

GGGg was subsequently encouraged not to accept the berth to the USAT playoff, after other regional winners squealed. http://www.kenilworthchessclub.org/kenilworthian/2008/02/usate-2008-gggg-cannot-play-in-finals.html

Sorry for going all Louis Blair on everyone with the links.

32

8 would be -8 and 9 would be -141. The difference is only 133, which is within the 140 points for alternation. 1 (-1) and 9 (-141) could still switch but 1 (-1) and 10 (-142) would not (retaining the current result with your 8 point limit swap for (-1) and (-10).

33

Yes. These were the original round two pairings. Glenbrook South got the booby prize with the even smallest even number in the 17-24 score group.

Rd Bd Scr White Scr Black 02 1. 1.0 Chicago (Whitney Young) (4.0,-1) 0.0 Glenview (Glenbrook Sout (3.0,-18) 02 2. 0.0 Bloomington (H.S.) (3.0,-19) 1.0 Chicago (Northside) (3.0,-2) 02 3. 1.0 Wheaton (W.-Warrenville S (3.0,-3) 0.0 Palatine (Fremd) (3.0,-22) 02 4. 0.0 Glen Ellyn (Glenbard Sou (3.0,-21) 1.0 Urbana (University) (4.0,-4) 02 5. 1.0 Evanston (Twp.) (4.0,-5) 0.0 Naperville (Central) (3.0,-20) 02 6. 0.0 Mt. Prospect (Prospect) (3.0,-23) 1.0 Hinsdale (Central) (4.0,-6) 02 7. 1.0 Naperville (North) (4.0,-7) 0.0 Downers Grove (North) (3.0,-24) 02 8. 0.0 Wheaton (North) (3.0,-25) 1.0 Winnetka (New Trier) (3.0,-8) 02 9. 1.0 Lisle (Benet Academy) (3.0,-9) 0.0 Carol Stream (Glenbard N (2.0,-26) 02 10. 0.0 Glen Ellyn (Glenbard Wes (3.0,-27) 1.0 Lincolnshire (Stevenson) (4.0,-10) 02 11. 0.0 Skokie (Niles North) (3.0,-11) 1.0 Oak Lawn (Richards) (3.0,-28) 02 12. 0.0 Chicago (Lane) (3.0,-31) 1.0 Aurora (Waubonsie Valley (3.0,-12) 02 13. 0.0 Naperville (Neuqua Valle (3.0,-13) 1.0 Barrington (2.0,-47) 02 14. 0.0 Hoffman Estates (H.S.) (2.0,-62) 1.0 Orland Park (Sandburg) (3.0,-14) 02 15. 0.0 St. Charles (East) (3.0,-64) 1.0 Aurora (IL Math & Sci) (2.0,-16) 02 16. 1.0 Riverside (R.-Brookfield (3.0,-17) 0.0 Oak Lawn (Community) (2.0,-61)

With everyone in 17-24 at (-20), similarly for the others, these are the pairings generated:

Rd Bd Scr White Scr Black 02 1. xxx Chicago (Whitney Young) (1.0,-1) xxx Naperville (Central) (1.0,-20) 02 2. xxx Riverside (R.-Brookfield (1.0,-20) xxx Chicago (Northside) (1.0,-2) 02 3. xxx Wheaton (W.-Warrenville S (1.0,-3) xxx Palatine (Fremd) (1.0,-20) 02 4. xxx Glen Ellyn (Glenbard Sou (1.0,-20) xxx Urbana (University) (1.0,-4) 02 5. xxx Evanston (Twp.) (1.0,-5) xxx Downers Grove (North) (1.0,-20) 02 6. xxx Bloomington (H.S.) (1.0,-20) xxx Hinsdale (Central) (1.0,-6) 02 7. xxx Naperville (North) (1.0,-7) xxx Glenview (Glenbrook Sout (1.0,-20) 02 8. xxx Chicago (Lane) (1.0,-28) xxx Winnetka (New Trier) (1.0,-8) 02 9. xxx Lisle (Benet Academy) (1.0,-9) xxx Oak Lawn (Richards) (1.0,-28) 02 10. xxx Wheaton (North) (1.0,-28) xxx Lincolnshire (Stevenson) (1.0,-10) 02 11. xxx Skokie (Niles North) (1.0,-11) xxx Carol Stream (Glenbard N (1.0,-28) 02 12. xxx Glen Ellyn (Glenbard Wes (1.0,-28) xxx Aurora (Waubonsie Valley (1.0,-12) 02 13. xxx Naperville (Neuqua Valle (1.0,-13) xxx Barrington (1.0,-44) 02 14. xxx Hoffman Estates (H.S.) (1.0,-60) xxx Orland Park (Sandburg) (1.0,-14) 02 15. xxx St. Charles (East) (1.0,-60) xxx Aurora (IL Math & Sci) (1.0,-16) 02 16. xxx Mt. Prospect (Prospect) (1.0,-20) xxx Oak Lawn (Community) (1.0,-60)

All the -20’s are “identical” (other than due color and conference) for pairing purposes and end up (semi-)randomly assigned to slots where they would go if they were assigned specific seeds. The “semi-” is that if you erase the pairings and redo them, you’ll get exactly the same result.

34

Maybe the first round can be randomized but after that the player who has the higher performance rating after each round could be seeded higher as we are trying to differentiate based on skill and this is probably the best thing we have to go on. One potential problem is that if both players win in round 1, whoever was randomly seeded higher will have the higher performance rating. Still, I think differentiating based on performance rating is better than just doing it randomly.

35

Doesn’t this conflict with the 80/200 point rule?

Alex Relyea

36

The tournament in question isn’t USCF-rated (Illinois State HS Championship) and doesn’t have ratings, just seedings, so the performance rating doesn’t really exist.

In a ratings-based Swiss, in practice, the only situation where you have enough people with identical “ratings” to need to worry about how to handle them in subsequent rounds are the unrateds. After all, if you have just two players with the same rating, they will be on different color schedules, and that will control, and even if you have four or five players sitting on xx00 floors, they have to be in the same score group, and be on the same color schedule before they are interchangeable for pairing purposes.

At any rate, the suggestion was to rejiggle the order within a group of players with identical ratings based upon the performance rating. That doesn’t affect their rating for pairing purposes, or their placement relative to others, just their placement within their own group.

37

WINTDOAN, do you think differentiating based on performance rating is a good idea?

38

No. Since it only matters when the individuals (or teams) involved are in the same score group, a higher “ranking” means that you play a tougher opponent. If, for instance, both lost the previous round, why should the one who played the harder opponent now again have to play the harder opponent (and have that probably continue all the way through the tournament).

39

Yes, I brought that point up in my post but I think there are some instances where it could be done where this isn’t an issue, such as when one unrated has demonstrated they are better than another unrated, even though they are currently at the same score.

40

So it’s usually a bad idea, but there might be some situations (which probably would be hard to determine in advance) where it might be OK. Not exactly an operational pairing method now, is it?