FIDE Pairing Rules

Running Chess Tournaments
1

Does anyone know whether there was an intent to make a dramatic change in the Dutch pairings when revised in 2016? The previous description was, in effect, a description of the programming in one specific program. From what I can tell (generally) the 2016 revision replaced that with a something that looks more like “rules” rather than “algorithms”. However, the description of what used to be called “backtracking” seems seriously lacking.

(Italics indicate the part in question).

For example, suppose that there are an odd number of players, there are three below 2.0 but all those have already had a bye, so you hit the end and discover that you’ve failed to complete the round pairing. The PPB is 2.0 and the 1.5’s and below are the collapsed scoregroup. OK. Then what? “The pairing process resumes with the re-pairing of the PPB”. Except (based upon the counts of the players here), the 2.0’s could be paired—the problem lies south of 2.0. So how are you supposed to “re-pair” it. Any TD would certainly look for a 2.0 who could still take a bye, and would have to downfloat her and one other out of the 2.0’s. Except the whole point of the FIDE Dutch system is that the pairings are deterministic—find something that let’s you pair the rest isn’t part of that. The 2012 rules had a very specific (maybe overly wordy) method for dealing with this type of backtracking, but the 2016 rules are basically silent on the specifics. Does anyone know if that was intentional? (I could see taking the attitude that these are the low scores in late rounds so the need for specificity isn’t important).

2

I think this is yet another reminder that FIDE misses Sevan Muradian more than it realizes.

Alex Relyea

3

Yes, plus I think there are multiple FIDE pairing methods, some of which contradict others, and many of which keep changing day by day.

Not too long ago, it seemed FIDE pairings tried much harder than U.S.Chess pairings to get the colors right. FIDE would go to great lengths to make colors work, and if the colors couldn’t be made to work, it would transpose wildly in order to push the bad colors down to the bottom of the score group. Thus, by pushing the bad colors down, it effectively pushed the bad pairings (i.e. the transpositions) up.

At least it recognized that bad equalizations are worse than bad alternations. If a player due white had to be paired as if due black in order to balance the score group, FIDE recommended that the player getting the color shaft should be one of those due white to alternate rather than one due white to equalize.

But it never took this idea as far as it should. After, say, 4 rounds (with no byes or unplayed games), if all players were in balance (2 whites and 2 blacks), it seems to me that, to be consistent, FIDE should have said that a player with BWWB is “less due” white than a player with WBWB, and that if somebody has to get the shaft, it should be one of those “least” due white.

And then there’s a whole new variation, I think it’s called Dubai or Dubov or something, wherein the pairing program is supposed to keep track of transpositions by rating-point differences. Those who have been victimized by large unfavorable color transpositions in previous rounds should be compensated by being given favorable color transpositions in future rounds – or something like that.

If so, then FIDE seems to have completely reversed itself, now placing more importance on rating differences and less on colors than before. Either that, or it is giving arbiters a lot more choices than it did previously.

Bill Smythe

4

I have found a major problem is that the commissions conduct business in English, but for most commission members English is not their first language. Then some of the people they rely on to put stuff into understandable English aren’t very good at it (i.e. Mr Rueben of England)

That is one reason we fought hard to get some pretty key US people onto commissions like rules, qualifications, ethics, and constitutional.

5

Partially right, partially wrong. Yes, colors are king. Ratings are ignored except to assign pairing numbers. Any transposition or any interchange is permitted if it improves colors. However, in the Dutch system the bad colors (if a bad color is unavoidable) generally end up on a high board. (Transpositions which don’t involve top boards take precedence).

That’s not true (at least any longer). BWWB vs WBWB gives the W to the second player. As with US rules, you back track through the colors until they differ and alternate from that. The only difference is that in FIDE rules, you don’t count x’s, so xBxW is effectively the same as BWBW. Under US rules, the round 3 x is compared with the round 3 B, so the second player gets W.

Dubov. It tries to equalize (to the extent possible) the competition averages of the players in the score group by having the players who have (so far) faced weaker opposition play higher rated players. The attempt at equalization is done for players due White, so you order those by CA from low to high and the players due Black by rating high to low and try to pair those line by line. From what I can tell, it probably works OK for a big tournament where the score groups will tend to be fairly large. In a smaller tournament, if it’s not easy to find opponents, the whole idea behind it breaks down.

6

IA Stuart Reuben has more experience organizing and running tournaments for the last several decades than almost anyone else in the world. He was the ONLY member of the Rules Commission who voted for the proposal to loosen the restrictions on half-point byes, which, despite having been approved by those in attendance at the Rules Commission, was nevertheless voted down in a rather whimsical manner by the Rules Commission itself.

I saw for myself some of the monkey business that went on in the Commission Meetings, and how problematic it is to make any constructive change among the all the competing national interests, let alone the language issues. Stewart’s experience in navigating these political complexities, which long preceded Sevan’s, was a significant asset.

It was quite obvious to me that he knew his stuff inside and out.

7

He may well, but his drafting skills are an issue. In the rule of which you were very concerned we fixed the language, then he took it and got it totally messed up. We had to go back again to the chair and get the language our working group had figured out put back in. What he came up with would have not answered your concern.

8

I had it reversed, then, and I stand corrected. It’s still true, though, that pushing transpositions down (or up) is approximately equivalent to pushing bad colors up (or down, respectively).

That’s true in both FIDE and U.S.Chess rules if two players are paired against each other. What I’m saying is that the FIDE rule (and the U.S.Chess rule too, for that matter) ought to go a bit further, by declaring that a player with BWWB is less strongly due white than a player with WBWB even if they are not paired against each other.

In other words, when it is necessary to give black to a player due white, priority for this “shafting” should go to the player least due white. This is already true when the question is equalization vs alternation, so why not also when the question is color history?

Let’s suppose we have ten players in our score group:

[size=150][code]

  1. WBWB 06. WBWB
  2. BWBW 07. BWBW
  3. WBWB 08. BWWB
  4. BWBW 09. BWBW
  5. WBWB 10. WBWB
    [/code][/size]
    where we have 6 players due white and 4 due black. One player due white must be given black.

The tendency would be to pair:
1 vs 7
6 vs 2
3 vs 9
8 vs 4
5 vs 10 (bad colors)

Or, if you prefer to push transpositions down and bad colors up, as mentioned above:
1 vs 6 (bad colors)
8 vs 2
3 vs 7
10 vs 4
5 vs 9

But I say, better than either of these would be to assign the black pieces to player number 8, the one player (of the six due white) who is least strongly due white:
1 vs 7
6 vs 2
3 vs 8 (bad colors)
10 vs 4
5 vs 9

Bill Smythe

9

No. Just no.

10

Players should be concerned less about pairings and more about good moves.

11

The English language issues pertain to many of the commissions. It is understandable why it happens. Hopefully, US Chess can help there.

12

Yes. Just yes.

I’m talking about the Dutch system, not the Dubov system. My idea is a simple extension of the Dutch system philosophy, which is to transpose as much as you need to, within a score group, in order to improve colors as much as possible. FIDE should add this little modification to the next version of its Dutch system.

Players, yes. Organizers, arbiters, and authors of pairing programs, no.

Bill Smythe

13

The same would also apply to players due a color on equalization. And in a later round:

BWBWB <> WBBWB <> BWBBW <> xxBWB etc.

Instead of having players divided into six basic classes (due white/black absolute/strong/weak), you could easily have, in a single score group, 15 or more classes of players, many of whom are trivially different. Your idea looks like a piece of cake in a 10 player score group—try it with 30, not so much.

In practice, with the Dutch, you don’t have to go through all 15! (roughly one billion) transpositions to pair a 30 player score group because you can do a look ahead and figure out how many bad colors of what type are expected and then work from top to bottom. In looking for a pair for 1 in the top half, look for the first player in the bottom that can play 1 and where the colors are OK or have the expected bad color problem. Move on to 2 in the top half and do the same. Allow pairings to pass with the expected bad color problem until you’ve hit the number you’ve determined to expect. (As I said, bad colors go towards the top). Thereafter, all pairings must have correct colors. If you split color claims into a much larger set of possibilities, then it becomes much less likely that something short of a full-on enumeration of possibilities will work.

In short, no. Just no.

14

Arbiters should be concerned less about good moves and more about correct pairings.

15

Note that the USCF pairing rules also have numerical rating limits for transpositions, though these are often ignored on the local level in small events. Trying to keep colors good without creating mismatches or breaking up score groups is a challenge. In using pairing cards over the years, I found it necessary to pair score groups at the top and also pair the bottom score groups at the same time to have reasonable and fair pairings. This is what some of the older TDs did when I started out as a TD. Their concern was to limit mismatches after the first round and were less concerned with the niceties about color allocation. Of course, there were a few who did not care about what happened to the lower score groups at all which led to a lot of problems in finalizing a legal pairing. I remember that the practice in last rounds when players in a pairing had a similar color history to allow them to flip for color rather than to assign the color. That resulted in less complaints about pairings and colors and bias by the TD. Some computer allocations of pairing and color remain mysterious. Pairing programs never explain why certain choices were made. I have watched TDs scratch their heads and argue at times over what programs have spit out. When said TDs reverted to a manual pairing, the players had reason to ask why one pairing was favored over another.

16

One can be an effective commission member but not be the best person to draft things in common modern English.

17

Aha, now I see. You are afraid that this extra complication would be difficult to program.

But not if it’s handled properly, I think. One could come up with a formula for “due-ness score” and then rank all the players (in the score group) in order by their due-ness scores to decide which ones are to be deprived of their due color.

For starters, you could give 1000 due-ness points to a player each time he has played black, and -1000 each time he has played white. This would take care of equalizations.

Then you could add 10JJ due-ness points (J is the round number, 1 thru N) for each round J where the player has had black, and subtract 10JJ in each round where he has had white. This would place more weight on the later rounds, and (I hope) be completely consistent with FIDE and U.S.Chess procedures. For example, WBWB would be more strongly due white (-10+40-90+160, or +100 due-ness points) than would BWWB (10-40-90+160, or +40 due-ness points).

With 10 players, for example, you might have a score group as follows, after arranging them in order by due-ness scores:

BWBB (+2220 due-ness points)
WWBB (+200 due-ness points)
WWBB (+200 due-ness points)
WBWB (+100 due-ness points)
WBWB (+100 due-ness points)
WBWB (+100 due-ness points)
BWWB (+60 due-ness points)
BWBW (-100 due-ness points)
BBWW (-200 due-ness points)
WBWW (-2220 due-ness points)

With 7 players due white and 3 due black, two players due white must be given black. These should be those due white with the lowest (positive) due-ness scores, i.e. the +60 and one of the +100’s. (Which of the +100’s? Whichever the usual Dutch rules prescribe, which I think would be the highest rated, since apparently Dutch likes to push bad colors up in order to push transpositions down.)

It seems to me the complications are only slightly greater here than with the conventional Dutch pairings.

Bill Smythe

18

No. It’s an utterly pointless complication. And it would be a lot more complicated than you think. Give it up.

19

You pair the PPB (2.0 scores).
You pair the CLB (lower than 2.0 scores) and find no legal pairings.
No amount of repairing the PPB (2.0 scores) will get you a legal pairing in the CLB.
The 2.5 (or higher) score group now becomes the PPB.
The 2.0 and below scores now become the CLB and you pair them together.

20

Sometimes a task is easy enough (or at least doable) if you want to do it, but extremely difficult if you didn’t want to do it to begin with.

Bill Smythe