Simple looking pairing question

Running Chess Tournaments
1

Going into the last round of a 4 rd. Swiss, these were the standings:

          1    2   3    4 (score after 3)

1.  1587  D4   W5  D2       2.0
                           
2.  1532  W7   W8  D1       2.5
                           
3.  1523  L8   W7  W4  -H-  2.0
                           
4.  1445  D1   W6  L3       1.5
                           
5.  1353  -H-  L1  L8       0.5
                           
6.  1345  -H-  L4  W7       1.5
                           
7.  1193  L2   L3  L6  BYE  0.0
                           
8.   985  W3   L2  W5       2.0

The player in clear first, #2 with 2.5, could not play any of the 2.0 players, so he had to face one of the two 1.5’s. It is obvious based on the rules, and also my gut, that the correct pairing based on ratings is 2-4, because it gives the leader the strongest possible competition. Several of the players had already made that same assumption. The problem was that #2 and #4 were both BWB. The first try at pairing with WinTD gave 2-6, to avoid assigning #4 BWBB. After poking around the ‘preferences’, I saw that ‘limit drop/raise swap’ was not in effect, and tried it again with that box checked. Sure enough, that run gave 2-4. Since the pairing software couldn’t dictate which way was correct, I consulted the rulebook.

Rule 29E5a. of the 5th ed.,‘The 80-point rule’ allows transpositions and interchanges up to a limit of 80 points to optimize due colors, but 29E5b on the next page, ‘The 200-point rule’, allows such switches up to 200 points to avoid giving a player one color two more times than the other (BWBB). Since #4 and #6 were less than 200 points apart, I figured that the rule forced me to pair 2-6, because pairing 2-4 (forcing BWBB on #4) would only have been allowable if #4 and #6 had been more than 200 points apart.

Question #1: Am I interpereting that correctly?

That’s the simple question. What happened next is fuzzier: After seeing preliminary pairings, player #4 objected to the 2-6 pairing, which gave him W against easier opponent #5! (The match-up 1-8 stayed consistant either way, so it was either 2-6 and 4-5, or 2-4 and 6-5). Player #4 thought that he should face #2 based on rating, and when I explained to him that the only reason it was not that way was to avoid giving him too many blacks, he stated that he did not mind getting BBWB.

So then I reasoned through it this way: The better pairings by rating were 2-4 and 6-5, and the ONLY person with a legitimate objection to that would be #4 because he would get BWBB. In fact, #4 could force me to give him BWBW by demanding 2-6 and 4-5, but he wasn’t doing that, and even wanted the opposite. Also, it seemed that the large score gap (2.5 vs 1.5) made optimizing pairings by ratings instead of by color even more important, but of course that is not a specific rule. I finally went with 2.BWBW-4.BWBB and 6-5.

Question #2: Did my decision make sense, was it fair, and also was it legal?

2

I don’t think you’re required to change pairings to improve colors. It’s allowed to do either if they’re less than 200 points apart, but not allowed if they’re more. In any event, allowing the player to pick his pairing is a very bad idea.

Alex Relyea

3

That happened to me once at a tournament in Rhode Island. In the last round, the top group could either be paired as 1-2, 3-4 with good colors or 1-4, 2-3 with bad colors. So the TD asked us to vote on it. We all voted for the pairing where we didn’t have to play John Curdo. Except for Curdo, who voted for the pairing where he had White.

4

So, who wound up playing Curdo, and as what color?

5

I got White against Curdo and lost. That happened a lot in those days. If you’re interested, the game is in one of his books.

6

Back to the original question, I would have gone with 2 vs 6. The rulebook seems to recommend such a transposition. But since it does not require it, I suppose you could go with 2 vs 4. The fact that 4 himself preferred this could, I suppose, give you additional ammunition, though it is a slippery slope to take player preferences into account when making pairings.

As for all the other players who “expected” 2 vs 4, it is a common phenomenon for players (and even some TDs) not to look at the whole picture when guessing what the pairings are going to be. Don’t let them sway you.

By the way, you were lucky you had any pairings at all in round 4. You could easily have fallen into the 6-player trap, mentioned in other threads, where no pairings exist at all.

Bill Smythe

7

I don’t know if your pairing was legal or not, but it is certainly “unfair” insofar as you allowed one player’s preferences to affect the pairings in the last round.

As noted, the “textbook” seems to call for 2 v 6 for better color distribution, and allowing 4 to lobby for and gain a 2 v 4, even if apparently unfavorable to him, gives him an advantage over the other players who do not attempt to influence the pairings.

8

Yes, I understand, and agree with you completely. In fact, in preperation for this event, I reviewed the posts in the forums concerning RR table-guided Swiss. I even had that table worked out and set down on paper between rds. 1 & 2, but then something interferred: player #6 was a late entry. I had already used the recommented system for a 7 player section, where player #5, with a pre-requested half-point bye in rd. 1, had the ‘ghost’ opponent that first round. Then #6 showed up. I saw that I could have simply continued the way it was going, inserting #6 in that ‘ghost’ spot. The reason that I didn’t was because it would imply that #5 and #6, the two first round byes, had already faced each other, and therefore couldn’t be paired later, but that was just not true. So then I reverted to full Swiss and crossed my fingers.

9

and

Thank you- this seems to answer my first question, and the answer is no, I was not reading the rules correctly! I thought that the book meant that I had to pair 2-6, even if it was not my own personal preference. But 29E5 says “Colors vs. ratings. Correct Swiss pairings should consider both colors and ratings, so a tournament director should exercise care not to distort either unduly.” I guess that I didn’t want leeway, and wanted to make a decision that was absolutely decided by the rulebook.

True. I didn’t think that the players knew the pairing rules better than I did, so I was not using their opinions to decide which was technically correct. I only mentioned that in my original post to emphasize that the 2-4 pairing seemed to have some legitimate justification.

and

AND

I know that this will seem like just a matter of wording, but the way that I was thinking, it was not like me asking #4 “Who would you rather play” and just giving him that choice to make him happy. As I said, I personally wanted to pair 2-4, based on rating, NOT just a personal whim (regardless of whether it was a good or bad inclination, and different from Smyth Dakota’s choice above). Then I thought (incorrectly) that rule 29E5b could force me to pair 2-6 in order to avoid giving #4 BWBB. When #4 said that he was OK with BWBB, it meant that he was not going to officially protest what I wanted to do anyways, so I was safe to go ahead with 2-4.

This TD stuff is all extremely subtle and complicated, isn’t it?

10

Given that #4 was the only player who would “suffer” from 2-vs-4, and that he wanted to suffer, your pairing may have been more or less justified. It bothers me just a bit, though, that you wanted 2-vs-4 when 2-vs-6 would have solved a color problem.

If there had been a 5th round coming up, I would feel even more strongly that you should go with 2-vs-6. Pairing distortions in round N can cause further problems in round N+1.

There may be a subtle subconscious psychological effect at work here. Some TDs don’t like to make transpositions when they involve different score groups – they prefer to simply pair the lowest player in the high score group against the highest in the low group, period, not realizing that you can (for example) pair down the second-lowest instead of the lowest, or pair up the second-highest instead of the highest, in order to improve colors, following the same transposition limits (80 or 200 points) that apply to same-group transpositions.

As for the 6-player trap, you really don’t need to use the round-robin tables. It is a theorem that, with 6 players, you can pair the first two rounds any (legal) way you want, without falling into the trap. So all you need to do is, after you have made tentative 3rd-round pairings, just make sure there exist round 4 pairings. (If not, just change the 3rd-round pairings and check again.)

(Another 6-player theorem: If there is a way to make round 4 pairings, then there are two ways to make round 4 pairings. This comes in handy if there is a 5th round coming up. 6-player corollary: If there is a way to make round 4 pairings, then there is also a way to make round 5 pairings. So the above precaution (checking before finalizing the round 3 pairings) also suffices for a 5-round event.)

Bill Smythe

11

Sorry to jump on this thread but I have yet another “simple looking pairing question.” In a tournament this past weekend we had the following crosstable at the top of the standings after 5 rounds of a 6 round event.

1. Player A	1698	B63+	W49+	B21+	W8+	 B7+	 5.0
2. Player B	1734	W62+	B28+	W18+	B9+	 W5=	 4.5
3. Player C	1715	B32+	W27+	B19+	W10+	B6=	 4.5
4. Player D	1710	W73+	B50+	W20+	B14=	W15+	4.5
5. Player E	1675	B64+	W51+	B10+	W16+	B2=	 4.5
6. Player F	1673	W65+	B11+	W22+	B17+	W3=	 4.5

The natural pairings for round 6 are AvB, CvE and FvD but this leaves two bad color match-ups, CvE and FvD.

To correct the colors WinTD paired AvF, CvB and EvD, which looked weird giving the 5 point player the lowest player in the 4.5 group. I personally would have gone with AvB, CvD and EvF which corrects the colors and pairs up the top 4.5 player.

Is either way specifically more correct than the other?

12

CvD and EvF involves an interchange. Since there is a transposition available that is <= 80 points, the transposition is superior to the interchange.

I agree that it looks weird, but it follows the pairing rules more closely.

13

But doesn’t the way WinTD did it also involve an interchange for the remaining players in the group when pairing CvB and EvD? Both ways involve an interchange except my version doesn’t have the transposition.

14

Yes - which is why the players usually have no idea. They understand some very basic things - maybe - but not the complications. It is sort of fun to take someone who asks why they didn’t get the pairing they expected in a later round of the tournament and sit down with them and go through all of the color issues and previous opponents - and then throw in teammates in the same score group just for fun and watch the eyes glaze over in about 25 seconds. after that I rarely get asked the question again.

15

The natural pairings were:
1698 - 1734
1715 - 1675
1673 - 1710
A two-point swap for colors gives
1698-1734
1715-1673
1675-1710
Now you have a previously played conflict on board 2. Swapping black on boards one and two is only a 17-point swap while swapping blacks on boards two and three is a 37-point swap. 29E7, example 5, gives an example where the lower point count trumps playing the bottom of the previous group (the other side of the coin from playing the top of the next group). So WinTD saw a change that greatly reduced the total points swapped.

16

Well, let’s see. We have:

colors 1 Player A 1698 W63 W49 W21 W8 W7 bwbwb 5.0 2 Player B 1734 W62 W28 W18 W9 D5 wbwbw 4.5 3 Player C 1715 W32 W27 W19 W10 D6 bwbwb 4.5 4 Player D 1710 W73 W50 W20 D14 W15 wbwbw 4.5 5 Player E 1675 W64 W51 W10 W16 D2 bwbwb 4.5 6 Player F 1673 W65 W11 W22 W17 D3 wbwbw 4.5
TDs should train themselves not to find it “weird” to pair player A (the only 5.0) against player F (the lowest 4.5). This is, after all, only a 61-point swap (1734 minus 1673) from the “raw” pairing, A vs B. (I prefer “raw” to “natural” because such pairings, or their consequences, can actually turn out to be highly unnatural, e.g. pairing two players who have already met.)

In this example, however, there are other considerations.

Swaps within a score group can be either transpositions (respecting top-half vs bottom-half) or interchanges (violating top-half vs bottom-half). By the same token, swaps across two score groups could be regarded as either “score transpositions” (pairing a player in the bottom half of the higher score group against a player in the top half of the lower score group) or “score interchanges” (violating the above).

With A vs F we have a “score interchange”.

Two alternatives have been presented:

(1) A-B, C-D, E-F. One interchange, point swap 35 (1710 minus 1675).

(2) A-F, C-B, E-D. One score interchange, point swap 61 (1734 minus 1673), plus one “regular” interchange, point swap 5 (1715 minus 1710).

So alternative (1) seems better, but not exactly for the reasons originally given.

I do not agree with the argument that (2) can be justified by viewing it as a compound swap of 2 points (1675 minus 1673) and 17 points (1715 minus 1698). The latter two are in different score groups, so their ratings should not be compared.

Bill Smythe

17

If memory serves, WinTD does work step by step, so I think that is the sequence it used to get the pairing.
Also, the example I cited explicitly compares ratings across two scoregroups (swapping the top of the lower scoregoup with the middle of the three in the top scoregoup and treating it as a 30 point swap instead of looking at only the other two in the top scoregroup and considering it a multi-hundred point swap).

The A-B, C-D, E-F pairing would be a normal top-down pairing.

18

I’m all in favor of looking ahead rather than using top-down pairings. And, in the case of transpositions within a score group, I’m also in favor of using the smaller of the two rating differences to evaluate the transposition (29E5c).

There are four cases:

(A) a transposition within a score group
(B) a transposition across score groups
(C) an interchange within a score group
(D) an interchange across score groups

In case (B), only the rating difference of the two players actually being swapped should be used to evaluate the transposition:

“29D. The odd player. … Pairing players out of score group. … [29D1]b. … you should look only at the rating difference of the players being switched.”

If, for example, the second-lowest, instead the lowest, player in the upper group is paired against the highest player in the lower group, then only the rating difference of the two players in the upper group should be considered, not the rating difference of their opponents (in different score groups).

Symmetrically, if the lowest player in the upper group is paired against the second-highest, instead the highest, player in the lower group, then only the rating difference of the two players in the lower group should be considered.

In case (C), again, only the rating difference of the two players actually being swapped should be used to evaluate the transposition (29E5d).

Case (D) is sort of a combination of (B) and (C), so it should be doubly true that only the rating difference of the two players actually being swapped should be considered.

The attempt to justify the WinTD pairing was based on the following sequence:

  1. Raw pairings: A-B, C-E, F-D.

  2. Transpose E with F for colors: A-B, C-F, E-D. (2-point swap)

  3. Interchange B with F to avoid repeat pairing: A-F, C-B, E-D. (61-point swap)

To refer to step 3 as a 17-point swap seems in violation of 29D1b and/or 29E5d, since it is based on a comparison, not of the players being swapped (B and F), but rather of their opponents (A and C), in a situation where (according to the aforementioned 29D1b and 29E5d) only the players themselves should be compared.

So, I’ll still go along with the more obvious idea:

  1. Raw pairings: A-B, C-E, F-D.

  2. Interchange D with E for colors: A-B, C-D, E-F. (35-point swap)

Bill Smythe

19

But 29E5c gives an example of swapping a 1500 and an 1800 and treating it as a 20-point swap (based on their opponents rated 2000 and 1980). That example says that the 1980 and 2000 are the ones really being swapped, but it is the players in the lower half of the score group that are moved. Thus the B-F swap is really an A-C swap (of 17 points).

I think the A-B, C-D, E-F pairings are easier to understand and conform more to what many players would expect (the top player playing against the top of the next scoregroup), but I can see why WinTD did the pairings it did.

20

I don’t doubt that WinTD did what it did for exactly the reason you cite, but I still say it’s wrong.

There are contradictions in the rules. 29E5c says that all transpositions should be evaluated based on the smaller of the two rating differences. Yet 29D1b says that, in the case of transpositions across score groups, only the difference between the two players actually being swapped (i.e. the players in the same score group) should be considered. One wonders what the rulemakers had in mind.

The example in 29E5c does not necessarily support the WinTD pairing, because in that example, all four players are in the same score group. (At least it would appear so, as it talks about implementing the actual swap “by switching the cards in the lower half of the score group” [emphasis mine].)

In the case of cross-score-group transpositions, if you consider both differences, you can get some really absurd results.

For example:

Upper group:
player 1: 1801

Lower group:
player 2: 1900
player 3: 1850
player 4: 1800
player 5: 1750
player 6: 1700
player 7: 1650
player 8: 1600

Raw pairings:
1-2 (bottom player in upper group vs top player in lower group)
3-6 (top half vs bottom half)
4-7 (top half vs bottom half)
5-8 (top half vs bottom half)

But, suppose that (for some reason) it is highly undesirable to pair 1 vs 2. Maybe they’ve already played each other, or both have just had two blacks in a row, etc.

So we look for another opponent for player 1, i.e. we figure out whom to transpose player 2 with. Following the WinTD theory, we conclude that the best pairing is 1-7, because then the opponents (1 and 4) of the players being swapped (2 and 7) differ by only 1 rating point.

Does anybody really believe that pairing 1 vs 7 is better than pairing 1 vs either 3, 4, 5, or 6? I certainly don’t.

Bill Smythe