I can’t see any justification for the pairing, but I’m not willing to say flatly that it’s wrong without having the file for analysis. Given only the information above, I’m fairly sure that I would have overruled the program and paired 1737 vs 1776 and
1909 vs 1683.
Of the three players due white, 1737 is “least” due white. If, for example, he were paired against either of the other two, he would get black under rule 29E4, “Pairing players due the same color”, point 4.
This rule, as worded, is customarily used to determine which of two players should get due color if they are paired against each other. It’s not that much of a stretch, though, to also use this rule to decide which player, among several, should be given the “wrong” color whenever colors don’t work perfectly in a score group.
I think that’s what SwisSys did – it gave black to the player least strongly due white. If so, the program is thinking just slightly outside the box, or perhaps barely inside the box, in applying 29E4.
I actually like this application of 29E4. When I was on the 5th edition rulebook revision committee, I was the one who proposed points 1 through 5 of this rule (the 4th edition had left several points unclear). Now, however, I regret the wording “Pairing players due the same color”, as it would have been even better to apply these criteria also to players not paired against each other.
But in doing so, SwissSys also dropped the top player from the 4.0 score group to the 3.5 score group without improving color allocation. Should giving the wrong color to the player “least due” the due color extend across score groups?
First, “extending across score groups” isn’t really the issue here. With three players at 4.0 and one at 3.5, there will automatically be one cross-score-group pairing no matter what.
If there were four at 4.0 and two at 3.5, then of course I would agree that each score group should be kept intact if at all possible, even if pairing two of the 4.0s against the two 3.5s would improve colors.
Second, there is nothing particularly wrong with dropping the top player into the next score group, as long as that player is within 80 points of the player who would have been dropped according to the “raw” pairings.
Third, the Swis-Sys pairing did improve color allocation – it gave black to BBWBW rather than to BWBBW.
Having said all that, however, I have now come to the conclusion that Swis-Sys should not have made that pairing, but ONLY because it drops the 1776 rather than the 1683 – a transposition of 93 points (compared to the “raw” pairings), slightly in excess of the 80-point limit for such color improvements.
Possibly, Swis-Sys first tried to give black to BBWBW by dropping the 1737 rather than the 1683, a transposition of only 54 points. But that caused two players to be paired who had already played each other, so Swis-Sys then made an additional transposition of 39 points, dropping the 1776 instead of the 1737. These two “permissible” transpositions were each less than 80 points, but they added up to the not-so-permissible 93-point transposition.
Bill’s point exactly, isn’t there a provision in the pairing rules (I don’t have a rulebook handy, so I can’t cite the exact section), that says there are no rating difference limits to transpositions if they are made to avoid rematches (i.e. 1737 vs 1909)?
Note that the 1909 has had two blacks in a row and can’t get a 3rd. So if the 1683 plays the 1909, you have bad colors and have to give the extra black to the higher ranked player. It’s possible that in the pairing logic, that qualifies as really bad colors and is avoided in favor of simply bad colors. If that’s why the pairings fell the way they did (and that’s just speculation), IMO it’s an error. The ban on three in a row means that you wouldn’t pair two players both of whom have the same problem with consecutive colors.
That difference shouldn’t matter when determining who plays whom. It only comes into play if the two of them are paired and you are determining which of the two gets due color.
It’s a question of color equalization (200-point switch allowed) vs color alternation (80-point switch allowed). In this case, we have an equalization problem, but NO transposition solves the equalization problem, so the best we can do is optimize the color sequence. This type of optimization is actually even lower on the totem pole than color alternation. So an 80-point (or smaller) max seems more appropriate here.
True. But in this case (if Swis-Sys did as I speculated), the first transposition was made to improve colors, then the second transposition was made to avoid the rematch caused by the first transposition. However you look at it, the net total effect was a transposition of more than 80 points to minimize a color problem even less serious than alternation, let alone equalization.
I agree, that would be erroneous logic – especially since, no matter what transposition (or no transposition) is made, you end up necessarily giving the wrong color to a 4.0 and the correct color to the 3.5.
That’s true with the present wording of the rule (“Pairing players due the same color”). I have been arguing, though, that it would make sense to also apply the five points of 29E4 when deciding who plays whom to begin with – and I stand by that suggestion as a slight improvement over the present rule.
I had paired it originally with version 7.51. Then I downloaded version 7.56. I got the same pairings. JonH paired it with 7.56 and got the pairing of:
1737 vs 1776
1909 vs 1683
These were the hand pairings I had come up with at the club. Some of the other pairings in other score groups Jon had were different then mine. The only difference in his settings and mine were that he had 3 colors in a row allowed. This tournament had a number of color issues because of make up rounds where the unplayed games were treated as draws. Also there were a few upsets in those make up games that threw off what score groups people ended out in when the last round was paired.
I decided to use the hand pair because I wanted the top two 4s playing each other with equal colors.