Does the USCF certify tournament pairing software? Or is it that when software vendors claim that their programs comply with the USCF Swiss-pairing rules, that is based only on the vendors’ own interpretation of those rules and testing of their software?
Does the USCF maintain a list of programs which are certified (or merely claimed) to be compliant with the USCF tournament pairing rules?
When a TD uses a pairing program to pair a USCF-rated tournament, is he obliged to select the program from a list of “certified” programs? If a program mispairs a round, is the TD responsible for detecting that and correcting the problem? Can a TD be disciplined by the USCF for using a program that does not comply with the USCF pairing rules (and not detecting and correcting it)? Or is there a “safe harbor” if the TD uses certain popular programs (such as WinTD or Swiss-Sys), and that program errs? If so, what programs provide that “safe harbour”, and how was it decided that those programs are “safe”?
The TD is always responsible for the pairings. One cannot evade that by simply saying “the computer did it”.
However, some confuse the concept of the pairings they want or expect with the concept of “legal pairings.” A pairing may not be the best pairing in some players’ or even other tds’ eyes, but may still be a pairing permitted by the rules.
It is not so easy to determine from Rule 27 and Rule 29 what pairings are “permitted” by the rules.
Obviously, a pairing is permitted by the rules if it complies with all the prioritized requirements of Rule 27. That is, no player is being paired a second time with another player; all players are paired against someone with the same score; in each score group the “top half” is paired against the “bottom half” in rating-rank order; colors are all “equalized”; and colors are all “alternating”. In addition, the pairing must honor bye requests that have been permitted, and not involve any matches that have been disallowed, such as between members of the same team,
However, all that is rarely possible at the same time. An odd number of players in a score group is rather common – which means that at least one player in the group cannot play someone with the same score. Players having already met may interfere with top half playing against the bottom in rank order. And usually it does not work out that colors can all be equalized and alternated without “transpositions” in the rank ordering of a score group or “interchanges” between the top half and the bottom half. So, various compromises from the ideal are generally necessary. However, there is no mechanical or deterministic process in the rules for making these compromises, or for saying that any particular set of compromises is “legal”.
At best, it would seem that a pairing which violates the principles of Rule 27 is “legal” only if there is no other pairing possible which violates those principles less than the pairing in question. But Rule 29 pretty much leaves the director on his own to determine whether that is so. Indeed, it isn’t particularly clear how to determine which of two pairings is better with respect to Rules 27 and Rule 29.
Anyway, returning to the question of my original post: are you saying that there is no certification process for pairing programs, and that there is no pairing program available which a TD may safely suppose to be acceptable, where he will not be subject to discipline if he relies upon it?
I looked at this issue a couple of years ago and there was none then.
This like an issue that we CPA’s have to deal with in the world of finance. The accounting authorities have ruled that “the software did it” is not an excuse when financials are wrong. I assume that the logic is the same for USCF pairings, too.
Irrespective of whether the USCF has “certified” pairing programs, what programs are reputed to comply with USCF pairing rules? And, on what basis do these programs have that reputation, other than the assertions of the vendors?
The difference between the USCF’s rating rules and GAAP is that presumably it is possible to determine that financial reports are “wrong” under GAAP, independent of the software. How does one show that a particular Swiss-system pairing is “wrong” under the USCF pairing rules? How “wrong” does a pairing have to be for a TD to be subject to discipline? Has there ever been a case where a USCF TD was reprimanded or disciplined in any way for relying on a pairing program that produced a “wrong” pairing?
Yes. Caveat: “…own interpretation…and testing…” includes iterative communication and feedback from actual TDs before and after releases. Take that for what you will.
Not that I’ve seen.
No.
Yes.
It’s possible, but I think the scenario is unlikely.
Brian, I do not think you apply the comparison correctly.
The difference is not between USCF’s rating rules and GAAP but between USCF’s pairing rules and GAAP. TD’s do not rate games; hence they cannot be disciplined for that. As for your other questions, they depend on this one:
One does that by comparing the correct pairing to the TD’s pairing and evaluating the severity of the difference. - The same method CPA’s work is evaluated.
What is your real question? Why the focus on the word “discipline”? We should not be eager to “discipline” our TDs but we should always be eager to help TDs “learn and improve.” If you have an acutal case in mind, take it to the ‘Tournament Direction’ forum.
Pairng questions are often discussed the ‘Chess Tournaments’ forum. In some tight situations, there may be several correct answers with the ultimate decision sometimes being decided by the TD ‘style’.
I use SwissSys and rarely correct pairings but I do scan top to bottom every round looking for any obvious errors between score group boundaries. The last two rounds I scan the top two score groups making sure that I can explain or correct odd looking color allocations. Usually but not always SwissSys is correct.
I increase the USCF recommended point differentials for color alteration and equalization slightly as this leads to fewer late round color assignment problems.
OTOH, I know TDs that rant and rail against SysSys and it’s ‘bad’ pairings. My opinion is that everyone has to play someone every round. I have seen a rare case or two where a strong player won a campaign for a ‘patsy pairing’ and then failed to win the game. Justice!
Have you or a friend been a victim of a bad pairing? Again what is the real question? Should USCF certify pairing programs? Have you or a friend been a victim of a bad pairing? Maybe but the TD is ultimately responsible anyway.
If there is a pairing that complies with Rule 27 in all respects, then it is clearly correct.
However, the mechanical process of going through the score groups from top to bottom, dropping the odd players, and pairing the top half versus the bottom half will rarely produce a “correct” pairing. (By the way, the rules do not even specify that the director should go from high to low score groups.) Some score groups will have odd players, and the colors may not equalize and alternate.
At that point, the pairing rules do not give much guidance on which combination of various drop-downs, transpositions, interchanges etc is the “correct” one. The rules specify no deterministic order in which to make drop-downs, transpositions, and interchanges or specify in what order particular players should be paired so as to find a pairing and to be assured that it will be “correct”.
As I said above, the most that can be said is that a pairing is “correct” if there is no better pairing possible. How one is to determine that is not clear. Not is it even clear in all cases when one pairing is “better” than another one. There is certainly no algorithmic or mechanical procedure specified in the USCF pairing rules for doing the pairings. So how does the director find the “correct” pairing, and how does he know that a pairing is “correct” when he finds it – except for those rare pairings that perfectly conform to Rule 27?
No, I am trying to write a “USCF compliant” pairing program. The written USCF rules are woefully vague, hand-wavy, and underspecified. This means that the problem of writing a “USCF compliant” program is ridiculously easy because almost anything that looks like “top half versus bottom half with transpositions and/or interchanges” is acceptable. Or else it is impossible because there are in fact a whole series of unwritten rules that a software developer has no way to find out except by releasing something and “interacting” with TD’s.
I suspect that the vendors who claim to have done it are either exaggerating about being USCF compliant or else they did the USCF’s work for it and simply came up with deterministic algorithms not specified by the rules but which were more or less in the ballpark, and crossed their fingers on “compliance”. They may have adjusted the algorithms over the course of many years based on “feedback” until the complaints about “bad pairings” got down to where it was acceptable. Since the USCF apparently does not have a certification program or any means of testing these programs, they could get away with it. The programs probably just use something like the “Top Down” method, or some similar mechanical process (such as in the FIDE pairing rules) and hope that it looks close enough to what USCF TD’s expect.
The situation is very different with the FIDE pairing rules, by the way. FIDE has several different pairing algorithms for Swiss-system tournaments (Dubov, Dutch, etc). They are all designed for manual implementation with cards, but they are all algorithmically specified, deterministic, and mechanical. With a computer pairing program instead of pairing cards, there are better algorithms (such as those based on graph theory algorithms like weighted perfect matching, or on the Stable Roommates problem). But at least the FIDE rules are algorithmic. The same cannot be said for the USCF pairing rules.
The USCF pairing rules, as presented in the rule book, are just a mish-mash of rules of thumb, guidelines, consider-this, consider-that, and “tips”, and are frustrating as hell for a programmer. They are a bit of a disgrace, actually. Any software engineer who put out something like that as a “spec” would be told to go back to coding. The USCF would be better off just adopting the FIDE pairing rules. Another thing that is never going to happen.
IMO, many of these questions regarding pairings are better understood if viewed without the computer and with pairing cards. The language of pairing is more understandable when using pairing cards as a method versus a computer. When “dropping” down into the next score group, that is more visualized with a stack of pairing cards not a computer and I don’t think one really comprehends interchanges and transpositions without having done it with pairing cards.
As to which programs are “certified”, after years and years of practice, WinTD and Swiss Sys are “certified” in the sense that they are the two most used and from social mores, viewing the beaten path down to picking a program, these are the two most popular US programs. There was a director at the New York Open many years ago who had his own pairing program which paired just as well, but it was not in the least bit friendly. Swiss Sys seems a lot more user friendly but WinTD seems to consistently perform better in US Championship tests where the pairings are done manually and then done in both programs to see which comes closest to the human pairings. They use FIDE rules so the program needs the best FIDE pairing module to perform well. Both have plusses and minuses.
After the second round of a tournament I hand pair the top few scoregroups and see if Swiss Sys matches that when doing the pairings. Most of the time it does, sometimes not. When not, I look for the reason why. If the reason makes sense the pairings stay, if it does not make sense I change them to my original hand done pairings.
Brian - I good answer is also for you to experience making pairings with pairing cards and then with computers. Often after years of experience, one can just “sense” that a pairing is wrong. It will look wrong, it will violate top-half vs bottom half or some rule that really stands out. Often version updates with the software programs introduce bugs which create problems, like the odd man dropping down to the 3rd player instead of the 1st player in the next group,. something that is fixed by the programmer. At one World Open I was using Swiss Sys verson x.00 which had come out the previous day. By the time the tournament was over, the version had moved from .00 to. 01, .02, .03, .04, .05 and .06 with all the bugs found in a major event.
In the end it helps to be an active TD to know some of your answers. While directing is not rocket science, rocket science isn’t rocket science to rocket scientists who do it every day
In specifications and rules, simplicity should be a high priority to the degree that simplicity is plausible. Simplicity is a strength, but unfortunately simplicity is not valued as a strength in too many aspects of the USCF rule book.
Flawed is any pairing system that cannot be fully and correctly implemented in software by competent computer programmers as a deterministic algorithm. This fundamental principle seems to be rejected by many USCF TDs or Delegates.
There are often multiple valid ways to pair a round.
Some time back Bill Smythe advocated point-count pairings where the correct pairing would be the one with the smallest transposition total (plus other “problems” such as not alternating or equalizing colors, avoiding replaying an opponent, avoiding pairing teammates in an individual/team Swiss, etc.). I think one reason that idea never caught on is that it pretty much requires a computer or a person very fast on a calculator to determine the smallest such value, and thus a computer-less TD manually pairing would have a good chance of unintentionally violating the “best pairings” because the point count used was 0.4% higher than the point count that a program would have come up with.
Even with point-count pairings we have a situation from a US championship a couple of years ago where there were two possible 3-point transpositions to correct the raw pairing.
I agree for the most part that the rules should specify precisely the algorithm to be followed. All correct implementations of the algorithm, including manual, would produce the same result. As with FIDE, perhaps a choice of algorithms might be offered.
Alternatively, the rules might not specify any particular algorithm. Rather, the rules would specify correctness criteria. If presented with a pairing asserted to be correct, it would be possible to mechanically and algorithmically determine via the rules whether it was indeed correct.
The current situation is neither of these. The rules are simply a mish-mash of guidelines, rules of thumbs, tips, try-this, if-that-doesn’t-work-try-something else, don’t-forget-to-consider, etc.
There is neither a mechanical procedure given for producing “correct” pairings, nor any mechanical algorithmic process for determining that a pairing is “correct” and acceptable.
The USCF pairing rules don’t read like the people who wrote them have a clue about specifying algorithms or even manual card-based processes. And it isn’t as if it should be hard to see what is needed: FIDE has produced several different Swiss pairing systems that are algorithmically deterministic and specified, and suitable for implementation using cards. (For example, here are the rules for the “Dutch” system, which is probably the most similar to the USCF’s pairing rules: fide.com/fide/handbook.html? … ew=article) It is fine not to like the FIDE systems and to prefer some other pairing approach. When it comes to pairing systems, there is plenty of room for difference of opinion about what is important and/or desirable. But that shouldn’t mean replacing a specification with blather and hand-waving.
I am a Delegate, and I plan to submit some ADM’s. That is one reason why I am asking questions.
One thing I am contemplating is rewriting Rule 29 in its entirety to be more algorithmically deterministic. That has the problem that WinTD, SwissSys, etc have already had to interpret the current Rule 29, and have come up with deterministic algorithms. Nobody knows what the algorithms really are in WinTD and SwissSys (is that right?), or whether those algorithms are in all cases actually compliant with Rule 29. (The vendors say they are compliant, but how does anyone, including the vendors, know that?) Nevertheless those two programs are the defacto pairing systems for all substantial USCF tournaments, and rewriting Rule 29 to make the algorithms more deterministic would likely blow them out of the water.
Another option is leaving Rule 29 alone, and leaving WinTD and SwissSys in peace, but simply writing a rule that states that any of the FIDE Swiss pairing systems is also valid for USCF tournaments. Anybody writing a new pairing program can simply use one of the FIDE algorithms.
All that said, is it even worth the bother? is there any possibility that the Delegates would act on such an ADM?
Is that really USCF’s work? Pairing is an art not a science. Some day it may be reduced to science but not yet. At this point we rely on experienced practitioners taking many things into account and making judgments. It seems our most recognized TDs agree on most of the judgments involved.
Same with the game itself, although that’s a lot harder
Since we want our TDs to have less stress rather than more (ceteris paribus) there might be a desire to leave things undefined at the margin, so the players aren’t able to complain as much about decent pairings.
Maybe the USCF pairings are better in quality than the mandated FIDE pairings would be. What is quality? See “Zen and the Art of Motorcycle Maintenance” and/or pair some tournaments.
FIDE is going a bit far with determinism. One forfeits a game for showing up one minute late. That’s clear and simple but imho not very good.
Start formulating the ADM, find out who is on the rules committee and begin picking their brains through emails or hashing it out here, try to meet with the people that would be in the workshop before it happens at the US Open, attend the workshop and present your ideas for discussion. The workshop will take a strawman poll, discuss positives and negatives, potential ways of shoring things up, see if you’re open to friendly amendments, and then it will be a streamlined discussion during the actual delegates meeting where the strawman poll results will be provided to the full delegates. The workshops are important.
