Pairing Problem

[Brian_Mottershead]

In a tournament this weekend the situation preceding the last round was as follows, in rank order:

#5 Alpha (969) 3.0 B10 W6 B3
#1 Baker (1517) 2.5 B6 W4 B2
#2 Charlie (1460) 2.5 W7 B3 W1
#7 Delta (749) 2.0 B2 W10 B9
#3 Echo (1142) 1.0 B8 W2 W5
#4 Foxtrot (1042) 1.0 W1 B9 W6

In short, there was one player (Alpha) with 3 points, and 2 players with 2.5 points (Baker and Charlie). Swisssys paired Alpha against Baker. This resulted in Charlie becoming an odd player in the 2.5 score group, and since he had already played Delta and Echo, he was paired against Foxtrot, a double drop.

The father of Charlie, and various coaches and other Swiss System pairing experts joining the discussion opined that this pairing was “obviously” wrong, that no “human TD” would ever pair it this way. Their reasoning was that the pairing Alpha-Charlie, Baker-Delta would have resulted in only single drops, and no multiple drops. Also, Alpha-Charlie would have been a better pairing from the point of view of colors than Alpha-Baker. (The fact that Alpha was a lower-rated player who was having a very good day and was a bit of juicy target for Charlie was of course irrelevant, and did not enter into the discussion.)

What I said was that the ratings should stand because the round had already started, and that the SwissSys pairings were acceptable under the pairing rules. In fact, the SwissSys pairings were the “natural pairings”. There were no transpositions or interchanges from the natural pairings which were legal under the 80-point and 200-point rules and which avoided rematches.

I also said that I would have paired it the same way as SwissSys, in that there is no provision in Rule 29 which would deprive Baker (as the highest ranked player whom Alpha had not already played) of his opportunity to meet Alpha in the final round. The rules state that Alpha, as an odd player, should be paired against the highest-rated player in the next score group, which was Baker. Avoiding multiple drop-downs further down the table is not mentioned as a reason to make exceptions to this rule.

This type of question comes up fairly often in the last round, because we frequently run four-round Swiss tournaments with eight to ten players, or less. The local pairing experts all seem to believe that SwissSys should be “looking ahead” and picking pairings which minimize the number of multiple drop-downs and that this is a higher priority than pairing an odd-player against the highest ranked player in the next score group. I can’t find this rule in the book.

So, what do the real pairing experts think?

[ericmark]

@Brian Mottershead

In a tournament this weekend the situation preceding the last round was as follows, in rank order:

#5 Alpha (969) 3.0 B10 W6 B3
#1 Baker (1517) 2.5 B6 W4 B2
#2 Charlie (1460) 2.5 W7 B3 W1
#7 Delta (749) 2.0 B2 W10 B9
#3 Echo (1142) 1.0 B8 W2 W5
#4 Foxtrot (1042) 1.0 W1 B9 W6

In short, there was one player (Alpha) with 3 points, and 2 players with 2.5 points (Baker and Charlie). Swisssys paired Alpha against Baker. This resulted in Charlie becoming an odd player in the 2.5 score group, and since he had already played Delta and Echo, he was paired against Foxtrot, a double drop.

The father of Charlie, and various coaches and other Swiss System pairing experts joining the discussion opined that this pairing was “obviously” wrong, that no “human TD” would ever pair it this way. Their reasoning was that the pairing Alpha-Charlie, Baker-Delta would have resulted in only single drops, and no multiple drops. Also, Alpha-Charlie would have been a better pairing from the point of view of colors than Alpha-Baker. (The fact that Alpha was a lower-rated player who was having a very good day and was a bit of juicy target for Charlie was of course irrelevant, and did not enter into the discussion.)

What I said was that the ratings should stand because the round had already started, and that the SwissSys pairings were acceptable under the pairing rules. In fact, the SwissSys pairings were the “natural pairings”. There were no transpositions or interchanges from the natural pairings which were legal under the 80-point and 200-point rules and which avoided rematches.

I also said that I would have paired it the same way as SwissSys, in that there is no provision in Rule 29 which would deprive Baker (as the highest ranked player whom Alpha had not already played) of his opportunity to meet Alpha in the final round. The rules state that Alpha, as an odd player, should be paired against the highest-rated player in the next score group, which was Baker. Avoiding multiple drop-downs further down the table is not mentioned as a reason to make exceptions to this rule.

This type of question comes up fairly often in the last round, because we frequently run four-round Swiss tournaments with eight to ten players, or less. The local pairing experts all seem to believe that SwissSys should be “looking ahead” and picking pairings which minimize the number of multiple drop-downs and that this is a higher priority than pairing an odd-player against the highest ranked player in the next score group. I can’t find this rule in the book.

So, what do the real pairing experts think?

Disable at least the 80-point limit and possibly the 200-point limit as well, for a smallish local tournament, one section, where players of a wide range of ratings compete.

That’s what I do ‘in my head’…our events are so small we do not bother with pairing software. My gut reaction would have been the alternate pairings suggested—not the ones the software chose—but I have not stared at it long enough, maybe.

I am not an expert in the sense you mean, but as someone who has paired by hand at small events for 25+ years, I see the merits both of using software and being ready to over-rule the pairings it spits out.

[wilecoyote]

Keeping the players in the same score group takes priority over color allocation. When it is not possible to keep score groups together, the priority becomes minimizing the drops in score group. I believe the pairings Alpha-Charlie and Baker-Delta are correct in this situation. However, I am not surprised by SwissSys’s pairings, although if I were the TD, I would have overridden them.

(As an aside, this is one of the things I most dislike about having no time to do pairings and to review the results, and having someone hovering to snatch the pairings and put them up on the wall as fast as humanly possible, shaving seconds off the time. But I digress …)

[fantasychess]

@Brian Mottershead

I also said that I would have paired it the same way as SwissSys, in that there is no provision in Rule 29 which would deprive Baker (as the highest ranked player whom Alpha had not already played) of his opportunity to meet Alpha in the final round. The rules state that Alpha, as an odd player, should be paired against the highest-rated player in the next score group, which was Baker. Avoiding multiple drop-downs further down the table is not mentioned as a reason to make exceptions to this rule.

Please go back and carefully reread section 29D (especially the last sentence), paying particular attention to subsection 29D1b.

[Brian_Mottershead]

@fantasychess

@Brian Mottershead

I also said that I would have paired it the same way as SwissSys, in that there is no provision in Rule 29 which would deprive Baker (as the highest ranked player whom Alpha had not already played) of his opportunity to meet Alpha in the final round. The rules state that Alpha, as an odd player, should be paired against the highest-rated player in the next score group, which was Baker. Avoiding multiple drop-downs further down the table is not mentioned as a reason to make exceptions to this rule.

Please go back and carefully reread section 29D (especially the last sentence), paying particular attention to subsection 29D1b.

Chris, what would your reasoning be from section 29D be in this situation?

The last sentence of 29D states that it is “acceptable” to pair the odd player against a lower-ranked player if that would improve colors. That is apparently to be treated as a “transposition” and must fulfill the 80-point and 200-point rules regarding transpositions. What does “acceptable” mean in this context? (As an aside, as a TD I don’t want the rules to tell me what is “acceptable”. When it comes to pairing, I want the rules either to tell me what I must do, or what I must not do. I only want the rules to tell me what is “acceptable”, when I would otherwise think that something is not acceptable. Otherwise, I don’t need to be told what is “acceptable”.)

In any case, what agitated the assembled pairing experts more than the color allocation problem (which was minor), was that SwissSys did not “look ahead” to avoid the double drop down of player C, and that is what I am more interested in here. In the situation, one of Baker or Charlie was going to be paired “up” (against Alpha), and therefore the other was going to be paired “down”. If it was Baker paired up, Charlie would be paired down two groups. If it was Charlie paired up, Baker would be paired down, but only one group. Is there a rule which requires the pairing in that situation, ignoring colors, to be Alpha-Charlie? In other words, is there a rule on the basis of which one can say that the SwissSys pairing was “wrong”, and “should” have been corrected by the TD? Please point to the rule.

Ken mentions some important practical issues, such as the fact that with ASAP round starts, there is no time for the TD to do more than cursorily spot check the pairings done by the pairing program.

Leaving those issues aside, it seems to me that the TD should only “correct” the output of the pairing program when the automated pairings are “objectively wrong”. Otherwise, the TD is open to charges of playing favorites – of changing the computer pairings when they negatively impact certain players (for example the players with the most opinionated parents or coaches) and not others. It seems to me that the conclusion to be drawn is that the TD should not second-guess the computer pairings unless he can point to a rule which is violated by those pairings. A rule which states that such and such a procedure is “acceptable” is not a rule which makes a different procedure wrong, even if the “acceptable” procedure is the one which most “human” TD’s would follow.

[eastside]

First, a specific comment. The only pairings I think of as being “obviously” incorrect are ones which violate USCF pairing rules. The pairings SwissSys gave are within the rules, and as such don’t strike me as being obviously incorrect. Also, if the round had already started, I absolutely agree with not changing the pairings. That said, I do believe that Alpha-Charlie and Baker-Delta would have been better, for the following reasons.

(1) The top player (Alpha) would have received an opponent very similar in rating in either case.
(2) Baker wouldn’t be stuck with three blacks in four rounds (which I tend to avoid, unless absolutely forced).
(3) Baker would not have to drop multiple score groups (which I also tend to avoid, unless absolutely forced).

Now, for a few general comments.

In any tournament with n rounds, where 2^n is greater than the total number of competitors, I find it very useful to review the pairings manually each round (after round 1, of course) prior to posting. It won’t totally avoid pairing controversy (especially where chess parents are involved), but it does allow the TD to be fairly sure he’s getting it right.

For a small event, such as the ones you have regularly, Eric Mark’s suggestion to disable the switch limits is a good one.

A useful exercise might be to take the last few tournaments you’ve run, and try the “look-ahead” method yourself, using ye olde tyme pairing cards. This way, you can see how each round’s pairings might have changed if you do it sans SwissSys. You might find some surprising results.

[wilecoyote]

I think Chris might have been a little oblique in his reference to the last sentence of rule 29D, but that is the governing rule in this case:

29D. The odd player.

Pairing players out of score group. There will often be situations where some players cannot be paired within their score group. This will always be true if there is an odd number of players with a given score, and can also happen when players within a score group have already played each other, or are otherwise restricted (family members, same scholastic team, requested non-pairs, etc.) from playing. At least one player, and possibly more, will have to be dropped to play in a lower score group. In such situations, the first priority (other than avoiding restricted pairings) is to have players play as close to their score group as possible. (emphasis added)’

The SwissSys pairings failed to have players play as close to their score group as possible by dropping Charlie two score groups down. Clearly, there must be two players dropped down to lower score groups; there was a legal pairing available by which both players dropped one score group, rendering the SwissSys pairings incorrect.

Computer generated pairings should be overridden when they are provably wrong, of course. But I also am willing to override computer generated pairings when there is a legal pairing available which has fewer color problems than the computer generated pairings (all other factors being equal).

[wilecoyote]

Um, “crosstalk” … Boyd Reed and I apparently were writing our replies simultaneously. Interestingly, Boyd and I disagree about the correctness of SwissSys’s pairings. On the other hand, I’ve heard it said that you can put three NTDs in a room with a pairing question and end up with four answers/opinions.

[wilecoyote]

Regarding Eric’s suggestion to disable rating limits: I suspect that is a good suggestion because I am not sure that pairing programs (SwissSys in particular) handle 29D1b correctly. Note that 29D1b explicitly states that there is no rating limit on permitted switches of players in order to keep score groups intact. (I extrapolate that to mean pairing players as close to their score group as possible, or, in other words, minimizing drops).

The other concern I have is that in the specific problem posed at the start of the thread, with Charlie having already been paired against Delta and Echo, the pairing Charlie-Foxtrot gives Charlie (2.5) a player with a score of 1.0 (there being no 1.5 score group). That has a definite chance of skewing the results for first and second place, since Charlie may be getting a much easier game than Alpha and Baker. Foxtrot is not in the running for the place prizes (1st/2nd/3rd); even if he wins and has a final score of 2.0, there are three players going into the last round with at least 2.5 points.

[Brian_Mottershead]

Allow me to define “total drops” as the sum of “drops” over all match-ups in a round. If players are paired with the same score, that contributes zero to the “total drops”. If the higher ranked player has been dropped one score group, that contributes 1 to “total drops”, and so forth.

Ken and Chris seem to be saying that it follows from 29D1b that if pairing A has a lower “total drops” than pairing B, then pairing A should be used, not pairing B, provided the two pairings are equal as regards the number of rematches. “Total drops” trumps color allocation. Moreover, it seems Ken and Chris think that the existence of pairing A would make pairing B “wrong”, implying that if the TD notices the possibility of “A” in time, he should change the pairings, if the computer has chosen B.

How far does this extend? For example, suppose the “improvement” in “total drops” is near the bottom. Leaving aside the question of how you might notice it, would you pair the odd player against the second player in the next score group in order to obtain an improvement in “total drops” is in a much lower pairing group? If not, why not?

[eastside]

@wilecoyote

Um, “crosstalk” … Boyd Reed and I apparently were writing our replies simultaneously. Interestingly, Boyd and I disagree about the correctness of SwissSys’s pairings. On the other hand, I’ve heard it said that you can put three NTDs in a room with a pairing question and end up with four answers/opinions.

Directors are much like economists - place all of them end-to-end, and you’ll never reach a conclusion.

I tend not to blame pairing programs for situations like this - I believe it generated pairings that were within the scope of the limitations placed upon it by its settings. And in a six-player, four-round Swiss, dropping a player more than one score group is a very likely occurrence, especially in the last round.

I don’t think the pairings were illegal. But they certainly weren’t optimal.

[wilecoyote]

@Brian Mottershead

Allow me to define “total drops” as the sum of “drops” over all the pairings in a round. If player are paired with the same score, that contributes zero to the “total drops”. If the higher ranked player has been dropped one score group, that contributes 1 to “total drops”, and so forth.

Ken and Chris seem to be saying that it follows from 29D1b that if pairing A has a lower “total drops” than pairing B, then pairing B should be used, not pairing A, provided the two pairings are equal as regards the number of rematches. “Total drops” trumps color allocation. Moreover, it seems Ken and Chris think that the existence of pairing A would make pairing B “wrong”, implying that if the TD notices the possibility of “A” in time, he should change the pairings, if the computer has chosen B.

How far does this extend? For example, suppose the “improvement” in “total drops” is near the bottom. Leaving aside the question of how you might notice it, would you pair the odd player against the second player in the next score group in order to obtain an improvement in “total drops” is in a much lower pairing group? If not, why not?

If you have rematches, you’re in very deep doo-doo (or you have six players in a four round event and walked into a pairing trap in the third round . Since not pairing players twice has the highest priority of all the pairing rules, you can (and must) break up score groups in any way necessary to avoid rematches.

Otherwise, there is no limit as to how far down the wall chart rule 29D applies. You would try equally hard to keep the very bottom score groups together as you would the very top. It’s not just about who is in contention for prizes.

[Brian_Mottershead]

@wilecoyote

@Brian Mottershead

Allow me to define “total drops” as the sum of “drops” over all the pairings in a round. If player are paired with the same score, that contributes zero to the “total drops”. If the higher ranked player has been dropped one score group, that contributes 1 to “total drops”, and so forth.

Ken and Chris seem to be saying that it follows from 29D1b that if pairing A has a lower “total drops” than pairing B, then pairing B should be used, not pairing A, provided the two pairings are equal as regards the number of rematches. “Total drops” trumps color allocation. Moreover, it seems Ken and Chris think that the existence of pairing A would make pairing B “wrong”, implying that if the TD notices the possibility of “A” in time, he should change the pairings, if the computer has chosen B.

How far does this extend? For example, suppose the “improvement” in “total drops” is near the bottom. Leaving aside the question of how you might notice it, would you pair the odd player against the second player in the next score group in order to obtain an improvement in “total drops” is in a much lower pairing group? If not, why not?

If you have rematches, you’re in very deep doo-doo (or you have six players in a four round event and walked into a pairing trap in the third round . Since not pairing players twice has the highest priority of all the pairing rules, you can (and must) break up score groups in any way necessary to avoid rematches.

Otherwise, there is no limit as to how far down the wall chart rule 29D applies. You would try equally hard to keep the very bottom score groups together as you would the very top. It’s not just about who is in contention for prizes.

This rather shows the problem with the USCF pairing rules. No algorithm is specified. Simply a set of principles, which (more or less) allow you to state that one pairing is better than another, the implication being that if there is a pairing A which is better than pairing B according to the principles, then B is incorrect. Of course, that does not mean that pairing A is correct, since there may be a yet better pairing C which makes pairing A incorrect. Never mind that there may be no practical way ever to find the better pairing A (or C).

Unless you enumerate all possible pairings, and pick the best, or it so happens that the “natural pairings” are perfect as regards “drops” and “color allocation”, there is no way to know whether you have found the “correct” pairings. It may not be possible in a reasonable amount of time to know that the chosen paring is the best of all possible pairings and therefore “correct”.

Contrast this situation with the FIDE Dutch System pairing rules. (The Dutch System is one of the FIDE rule sets for ratings-based Swiss pairings.) The Dutch System rules are algorithmic in form. In fact, as has been pointed out in this forum, there is a little bit of hand-waving at a certain key point in the Dutch System, which renders it less algorithmic than it may appear on first sight. However, the situations where that fuzziness becomes important don’t occur frequently and in most situations the Dutch System is an algorithm. You follow the algorithm, your pairings are correct. If you don’t follow the algorithm, your ratings are not correct. Score one for the Dutch System.

[hmb]

@Brian Mottershead

In a tournament this weekend the situation preceding the last round was as follows, in rank order:

#5 Alpha (969) 3.0 B10 W6 B3
#1 Baker (1517) 2.5 B6 W4 B2
#2 Charlie (1460) 2.5 W7 B3 W1
#7 Delta (749) 2.0 B2 W10 B9
#3 Echo (1142) 1.0 B8 W2 W5
#4 Foxtrot (1042) 1.0 W1 B9 W6

Here’s a tongue in cheek suggestion for you, Brian:

Of the four players in striking distance of the lead, Baker and Charlie cannot play each other, and there is no one within 400 rating points of them left for them to play.

So perhaps you might consider giving them “meaningful” games by pairing them with Echo and Foxtrot, while giving Alpha a chance to try to stay ahead of them and win the event by pairing him with Delta.

After all, this is a “ratings controlled” pairing system, is it not?

[fantasychess]

@Brian Mottershead

Ken and Chris seem to be saying that it follows from 29D1b that if pairing A has a lower “total drops” than pairing B, then pairing A should be used, not pairing B, provided the two pairings are equal as regards the number of rematches. “Total drops” trumps color allocation. Moreover, it seems Ken and Chris think that the existence of pairing A would make pairing B “wrong”, implying that if the TD notices the possibility of “A” in time, he should change the pairings, if the computer has chosen B.

Actually all I did was show that your statement below was incorrect

@Brian Mottershead

there is no provision in Rule 29 which would deprive Baker (as the highest ranked player whom Alpha had not already played) of his opportunity to meet Alpha in the final round. The rules state that Alpha, as an odd player, should be paired against the highest-rated player in the next score group, which was Baker. Avoiding multiple drop-downs further down the table is not mentioned as a reason to make exceptions to this rule.

However, should I have come up with the pairings then I would have changed them to the alternate version, which minimized the number of people being paired out of their scoregroup but maintained the same number of bad color matchups. The switch of pairing Alpha with Charlie and not Baker meets the criteria set out in 29D1b and also has the added plus of avoiding the further out of scoregroup pairings.

I must, however, also note that should the question have been brought to my attention after play had started that I would not have changed the pairings. Even though SwissSys came up with what I think are not the “best” pairings, they could still be argued as a “legal” set of pairings and the issues given do not seem to be so drastic as to constitute stopping all the games and having them restart again.

[jwiewel]

@Brian Mottershead

…
This rather shows the problem with the USCF pairing rules. No algorithm is specified. Simply a set of principles, which (more or less) allow you to state that one pairing is better than another, the implication being that if there is a pairing A which is better than pairing B according to the principles, then B is incorrect. Of course, that does not mean that pairing A is correct, since there may be a yet better pairing C which makes pairing A incorrect. Never mind that there may be no practical way ever to find the better pairing A (or C).

Unless you enumerate all possible pairings, and pick the best, or it so happens that the “natural pairings” are perfect as regards “drops” and “color allocation”, there is no way to know whether you have found the “correct” pairings. It may not be possible in a reasonable amount of time to know that the chosen paring is the best of all possible pairings and therefore “correct”.
…

That may be true when doing top-down pairings, but look-ahead pairings would catch that.
You don’t have to try all possible pairings, but it doesn’t take that long to look at all the applicable pairings. WinTD actually does that and minimizes the problems, which results in it sometimes making strange-looking pairings that turn out to be very good (as long as the settings are correct).

I would have guessed that if SwissSys had been set with the 80/200 limits then it would have done the transposition that avoided having to pair a 2.5 vs a 1.0. Setting those limits to zero resulted in an inferior pairing that was still legal.
(And the option of setting the limits to zero can be virtually necessary when doing something like a double-round blitz tournament).

@Brian Mottershead

…
You follow the algorithm, your pairings are correct. If you don’t follow the algorithm, your ratings are not correct. Score one for the Dutch System.

I am guessing you meant pairings and not ratings.
As far as scoring one for the Dutch System goes, that may depend on whether or not the algorithm results in pairings that are very specific but inferior to a good pairing under the Swiss System.
For that matter, setting the Siwss pairing limits to 0/0 can result in a very specific and deterministic pairing that is quite inferior to ones obtained by using 80/200 limits.

[Brian_Mottershead]

@jwiewel

That may be true when doing top-down pairings, but look-ahead pairings would catch that. You don’t have to try all possible pairings, but it doesn’t take that long to look at all the applicable pairings. WinTD actually does that and minimizes the problems, which results in it sometimes making strange-looking pairings that turn out to be very good (as long as the settings are correct).

When I read 29E6a, which describes the “Look Ahead” method, it seems only to deal with color equalization and alternation and not number of drops. There doesn’t seem to be any discussion of “look ahead” in the context of drops. Regarding WinTD, it does not really do “look ahead” either. As explained by the developer of WinTD in this forum, it uses a simulated annealing algorithm. This involves generating a baseline pairing and then doing a random change to it. (analogous to “interchanges” and “transpositions”). The new pairing is scored and its score is compared to the previous pairing. If the new score is within a specified range, the revised pairing becomes the new baseline. The specified range is slowly changed. At first it is wide, including negative changes, and it slowly gets tighter and positive. This is the “annealing” analogy. The process ends when there have been no changes to the baseline for some number of iterations.

I would have guessed that if SwissSys had been set with the 80/200 limits then it would have done the transposition that avoided having to pair a 2.5 vs a 1.0. Setting those limits to zero resulted in an inferior pairing that was still legal.
(And the option of setting the limits to zero can be virtually necessary when doing something like a double-round blitz tournament).

I guess the way SwissSys works is that it only does “transpositions” and “interchanges” from the natural pairings. It considers the Alpha-Baker, Charlie-Foxtrot pairing to be the natural pairings. But there is no sequence of transpositions to get from the “natural” pairing to Alpha-Charlie, Baker-Delta consisting only of transpositions which are legal and improve the situation at every point along the way.

The difficulty with pairing is that there are situations where in order to find the better pairing you initially have to head in a direction which looks worse. The computer does not see when pairing Alpha that if it chooses Baker, it is going to get into problems pairing Charlie. It is not looking ahead. The “natural pairing” which results may not be the best,and there may be no way to improve it by a sequence of operations such that every one is an improvement.

[wilecoyote]

@jwiewel

I would have guessed that if SwissSys had been set with the 80/200 limits then it would have done the transposition that avoided having to pair a 2.5 vs a 1.0. Setting those limits to zero resulted in an inferior pairing that was still legal.

While I was not present at the tournament, I am fairly confident in stating that the 80/200 limits in SwissSys were not changed for these pairings. The suggestion to set these limits to zero was raised by Eric Mark in this thread.

[ericmark]

@wilecoyote

@jwiewel

I would have guessed that if SwissSys had been set with the 80/200 limits then it would have done the transposition that avoided having to pair a 2.5 vs a 1.0. Setting those limits to zero resulted in an inferior pairing that was still legal.

While I was not present at the tournament, I am fairly confident in stating that the 80/200 limits in SwissSys were not changed for these pairings. The suggestion to set these limits to zero was raised by Eric Mark in this thread.

I suggested that in response to Brian’s statement “There were no transpositions or interchanges from the natural pairings which were legal under the 80-point and 200-point rules and which avoided rematches.”

Since then I looked closely at the original post and the MSA crosstable for the section. It seems to me that 1. The quoted statement is not accurate—and the widely suggested alternate pairing would fall within the 80/200-point rule; and 2. Thus, it would not have mattered whether the 80/200 switches were set or not.

The more I look the less I understand the pairings the software created. (Note that I likely have less experience using pairing software than anyone else who has posted to this thread.)

As a human TD, I see there are four players in the top three score groups who have a chance to win the event. My reasoning, without consulting the rulebook, runs:

A. These four players should play on the top two boards, barring re-matches among the top four or very strange things in lower score-groups.

B. Someone (of the top four) must get an extra Black; it happens. As long as it’s not three Blacks in a row, that’s just how it goes.

C. The 3-pointer must play one of the 2.5-pointers. Dropping him to the lower-rated of the two possible opponents improves colors on top board—equalization, not just alternation—and is only a 57-point switch. If works, whether or not you use the 80/200-point rules.

D. That leaves someone on board two with an extra Black—but that had to happen to one of the top four-ranked players, thus cannot be helped.

Note that my human TD brain sees the ‘alternate’ pairings as superior, even without the two-group (and 1.5-point) drop factored in.

All I can guess is that the software found it better to give the ‘extra’ Black to one of the board one players rather than on board two. I do not know why. Is there an obscure pairing rules variant that can be programmed into pairing software that applies here?

As for the merits of 80/200 in general, that has been discussed before on the forums. I find the numbers too arbitrary and restrictive, especially for the small one-section Swisses I direct. For larger events it might make more sense.

Of interest here is that FIDE sets no such limits, as I understand it, as color issues are given higher priority in FIDE Swiss pairing rules than in USCF. I think I read that setting the 80/200 limits to 0 is called the FIDE option, or something close, in pairing programs.

[ericmark]

@wilecoyote

@jwiewel

I would have guessed that if SwissSys had been set with the 80/200 limits then it would have done the transposition that avoided having to pair a 2.5 vs a 1.0. Setting those limits to zero resulted in an inferior pairing that was still legal.

While I was not present at the tournament, I am fairly confident in stating that the 80/200 limits in SwissSys were not changed for these pairings. The suggestion to set these limits to zero was raised by Eric Mark in this thread.

To clarify: It’s probably obvious, but what I meant to say was “set the 80/200 transposition limits to infinity—or 3000 points, (not “zero points”)—so you are not limited by ratings when you equalize colors.”

The FIDE way, IOW.

However, it looks like that was not the issue here…