This came up tonight at our club. The only 3.5 had had WBWB, while the top 2.5 (no 3.0) had had WhhB, that is two half-point byes. The 3.5 was approximately 2150 and the 2.5 2050, not that I think that matters. This was also the last round.
SwissSys paired them with the 2.5 as White. We couldn’t figure out why, and the TD eventually (after much prodding from the 3.5) decided to overrule. Why would the program do that?
Alex Relyea
Because they were both equal white/black. So you start going back using full color history. In round 4 they were both black so that is a tie. In round 3 the 3.5 was white while the 2.5 had no color. So that is the round to alternate from, giving the 3.5 black and leaving white for the 2.5.
Because it was correct?
Looks like the computer pairing was correct based on 29E4.
Both players had equal colors and the latest round where the colors differed was round 3 - W and h, so the 3.5 reverses color to B. (The rule specifically includes an example comparing a color with an unplayed game.)
OK. That’s what I said, but neither the TD nor the player bought it, and I didn’t have my rule book with me. Thanks.
Alex Relyea
Maybe the TD would have let you borrow his Rulebook.
Or maybe the TD could have been prodded to read it himself.
Bill Smythe
Maybe this particular TD should be very careful before overruling a computer pairing in the future.
As should most TDs, for that matter. 90% of the time the computer makes better pairings.
Bill Smythe
And most of the time it doesn’t is due to setting the pairing options incorrectly. It’s reached the point where a strange looking pairing usually causes me to check the options before doing anything else. I’ve had cases like a few years back with my back room person showing me a strange looking pairing, my redoing the pairing with an interchange, and then the back room person changing a setting to finally allow interchanges and letting WinTD redo it to end up with exactly my pairing.
At my last weekend tournament WinTD came up with a totally bizarre looking pairing. After spending some time delving into it (the options had been verified earlier), it turned out to be the best pairing based on the rulebook.
People using SwissSys probably have similar stories.
bingo! WinTD has this nasty habit of doing what you tell it to do.
Yup. Some of these difficulties can be traced to a wrong turn taken in the rulebook, about 2-3 rulebooks ago. Thinking that they were making things easier for computers, the rulemakers specified a procedure for generating pairings instead of criteria for good pairings. WinTD finds pairings using a procedure that humans would find…challenging to implement. As a result, it often finds excellent pairings that a mediocre hand-pairer will never even consider.
And then there are the players, who think they can compute their pairing without considering the rest of the field.
I don’t think I’ve seen an incorrect pairing from WinTD in more than 5 years.
One thing to double-check in a Swiss is when there are two players in the exact middle of a score-group that have the same rating (i.e. players 5 and 6 out of 10). Some versions of WinTD would pair those two players with each other and then pair the rest of the score-group normally (i.e. 1-7, 8-2, 3-9, 10-4, 5-6 for the first round of a 10 player tournament with players 5 and 6 having the same rating).
I don’t remember seeing any other pairing errors as long as the options are set correctly.
Then don’t try a hybrid RR-swiss. I have almost a whole paragraph describing the bug fixes for that if the “wrong” player happens to have the highest score.
The problem Jeff describes was due to a flaw in the “scoring” system for pairings. With the cut point falling among players with the same rating, those players can be either top half or bottom half. It was incorrectly counting pairings as interchanges except if 5 played 6.
I remember a similar problem a couple of decades ago. Jeff and I were playing, somebody else was directing. (I think Swis-Sys was being used.)
There were 38 players in (the Friday night section of) the Master Challenge. The computer was set to do accelerated pairings. The program, at the time, accomplished this by assigning an extra “ghost” point to everybody in the top half, in order to cause the top quarter to play the second quarter, etc.
Jeff was number 19 (just above the midpoint), I was number 20 (just below the midpoint). The “ghost” point resulted in 19 “winners” and 19 “losers”, an odd number of each. So the program paired the bottom “winner” (Jeff) against the top “loser” (me), even though we had adjacent ratings on the wallchart. The rest of the pairings came out by quarters, as expected.
Bill Smythe
did you find anything wrong with this pairing? Seems like perfectly natural accelerated pairings, to me.
Please make my day and tell me that your game ended in a draw.
I don’t remember what the result was. I suppose you’re trying to say that, if the game was a draw, then the pairing vindicated itself. If your idea is to create as many draws as possible, to reduce the perfect-score count, then why aren’t you advocating doing all pairings in this manner, i.e. 1-vs-2 pairings?
A much more logical approach, when accelerated pairings are used with an odd number of pairings (4n+2 players), would be to make the top and bottom “halves” slightly unequal, e.g. 20 players in the top half and 18 in the bottom, or vice versa.
Bill Smythe
Because I recognize that such decisions must BALANCE competing forces. 1-vs-2 pairings do too much violence to the players’ expectations about a Swiss event. Almost by definition, in the perfect SWISS event, 1 and 2 play each other on Board 1 in the LAST round - not the first round. Accelerated pairings take the point of view that the players in the TOP half are credible candidates to finish high on the crosstable, while the players in the BOTTOM half are essentially playing in a qualifying event. At the start of round 3, the qualifiers from the bottom half are allowed to play with the big boys, and those who have demonstrated that they have no chance are never relevant to the question of “how many perfect scores will there be when the event is over”.
There’s a subtle trap here as well. Like most players, I think you don’t see it because this particular trap didn’t happen to you, personally.
True. Of course one would use 1-vs-2 pairings only if mentioned in pre-tournament publicity.
By any chance, are you referring to the fact that, with the top and bottom “halves” unequal, you can’t pair top-half losers vs bottom-half winners because those two groups are different sizes? If so, then that problem often comes up anyway, because there may be more (or fewer) draws in the top half than in the bottom. Solutions (contrived) are suggested in the rulebook.
Whatever accelerated-pairings trap you are referring to, I’ve probably fallen into it more than once. I dislike accelerated pairings to begin with. The whole concept is one big trap, if you ask me.
Bill Smythe