I’m using SwissSys 8.4, and after two rounds the 2.0 group is as follows:
2589 BW
2445 WB
2319 BW
1554 byeB
The computer paired 2445-2589 and 1554-2319. I thought that this was clearly incorrect, and I overruled the computer and made it 2319-2589 and 1554-2445. Why did the computer do it that way in the first place?
Alex Relyea
Colors.
Obviously, but doesn’t this exceed the transposition limit?
Which is why he changed it.
Perhaps we are talking at cross-purposes. Since the color business was obvious, I assumed that Alex’s question was whether there was something in the rules which would justify that large a transposition. I suspect it’s just a bug. One of the bug fixes in 8.5 was “Was a little too quick to switch for color when an equalizer met an alternator.”
Perhaps the software limits for switches should be changed. That is something that is in the user’s control (Rules For Pairing option).
I ran a test with 8.5, and it paired 2319-2589 and 1554-2445. Unless there’s some data I don’t have, this is just a bug, fixed in the last update.
I use WinTD rather than SwissSys. In WinTD the pairing logic can be set to different specific limits for transpositions and interchanges. If the interchange limit is set too high then the pairing program will do what this one did. Does SwissSys have the same options? If so then how were they set?
what’s wrong with the computer pairing? top half vs bottom half. There’s no transposition involved unless you change the natural pairings
oops, nevermind 
Yes. The defaults are 80 alternation/200 equalization, and I’ve never seen any reason to change them. I seem to recall that Jerry Weikel in Reno didn’t believe in making any adjustments for colors, so I suppose he’d set both to zero.
The pairing chosen had a 136 point interchange. Flipping the 2319 and 1544 would have been a 144 point transposition (based on their opponents’ ratings - not the 775 points based on their ratings). Since (as John noticed) the 1544 had a bye I wonder if the equalization routine was triggered to use a 200 point limit (and whether it compared a “valid” 144 to 136 or a “non-valid” 775 to 136).
Since each player is more than 80 points from the next, any transposition (or interchange) violates the 80-point rule. Is this justified because one player (1554) has had unequal colors? No, because in all proposed pairings, this player gets his due color. So the whole transposition is being made merely for alternation, rather than equalization, so should obey the 80-point rule.
In small tournaments (or small sections), under about 20 players, it may be advisable to change the 80 default to 0, but leave the 200 as is. If colors work too well in the early rounds of a small event, serious pairing problems can occur in later rounds. It’s good to have a few bad colors in round 3, to avoid setting up two camps, those who started with white and those who started with black. With two camps in a small event, you’ll eventually run out of desirable inter-camp pairings (where the colors work) and be forced to make some intra-camp pairings (where they don’t). It’s better to have the bad colors in round 3 (alternation) rather than in round 4 (equalization).
As an extreme example, try pairing the first 3 rounds of a 6-player Swiss, making all the colors alternate, and then see what you can do in round 4. It isn’t pretty.
Bill Smythe
Controlled round-robin (29S).
I observed a similar event and the pairing software (WinTD) paired the same two players again in the last round. I was surprised. Has this happened to anyone else?
Many times. I call it the Dennis Keen trap, after the first TD I saw fall into it. I’m sure, however, that he was neither the first nor the last.
Bill Smythe
Bill, I have no idea what you are trying to say. I expected the software to make round robin like pairings since I thought not replaying the same player was a higher priority.
The TD has to make the decision to use RR-like pairings at round 1. If he waits until it’s too late, it’s not the computer program’s fault!
The point is that, in a small tournament with too many rounds, if you make normal Swiss pairings and normal adjustments for colors, it is possible to reach a situation in which there are no legal pairings in the last round. Preventing this requires that you make a conscious decision in round 1 or 2 to violate the standard pairing rules. Building this into a program would require either a simulation three or four rounds ahead (unrealistic even with very fast processors) or including a special “exception” subroutine (possible, I suppose, but hardly worth it for something that hardly ever comes up).
When you tell the computer to pair as a Swiss, it will just blithely pair each round as a Swiss, just as you told it to do. It won’t look ahead to the next round to determine whether any future pairings exist. Then, in a small tournament, when the next round comes, there may be no pairings at all, except those which pair some players against an opponent they have already played.
Here is a six-player example:
colors score
Player 1 W4 L2 W6 wbw 2.0
Player 2 W5 W1 W3 bwb 3.0
Player 3 W6 W4 L2 wbw 2.0
Player 4 L1 L3 W5 bwb 1.0
Player 5 L2 L6 L4 wbw 0.0
Player 6 L3 W5 L1 bwb 1.0
I dare you to pair round 4, without pairing the same players twice. I’ll even allow you to ignore scores and colors if necessary.
Bill Smythe