Controversial pairings from SwissSys

Running Chess Tournaments
21

Well, the part of 29D (“The odd player”) which says “… switches to correct colors should stay within appropriate limits (29E5)” means, to me, that 80- and 200-point switches are just as appropriate with inter-group pairings as with intra-group pairings, i.e. “the rule book says to” make such transpositions to improve colors.

This is certainly true, but perhaps because the complainers haven’t read, or don’t understand, the rules.

Rotsa ruck. First of all, a whole bunch of consultants (like the majority of the 30-person group that advised the current edition) would have to agree on the details. That would be a biggie in itself. Second, point-count pairings wouldn’t do human TDs much good, except to decide between two (or more) proposed sets of pairings. Somebody (whether human or silicon) would have to come up with the different proposals in the first place.

The point-count concept might, however, serve well as a guide to writers of pairing programs (they probably already do something like it). Human TDs often have difficulty coming up with pairings as good as those produced by the programs, and calculating the points by hand (in a manually paired tournament) could delay the next round considerably.

It’s mainly a way of settling arguments – but even then it won’t work, because those who disagree with the results will simply claim the point-count formula was flawed to begin with. (But let them come up with their own formulas, then.)

Bill Smythe

22

Were there an odd number in the score group, or did the program pair down TWO players just to improve colors?

Bill Smythe

23

I worry about this too – and for the present as well as the future. Worse yet, these same TDs (with minimal manual pairing experience) are overruling pairing programs whenever they don’t immediately understand the logic behind the computer’s pairings. Then the players will usually end up with worse pairings, not better ones, and the TD won’t even understand why.

Unless a TD is VERY experienced, it’s almost always better to let the program pairings stand.

Bill Smythe

24

Those same TDs also don’t understand the settings in their pairing programs.

That’s why the first question I always ask about a so-called ‘bad pairing’ from a computer is what the settings were.

25

There was an odd number. The best colors resulted from taking the highest player down into the next score group, so that’s what the software did. The players were both late for the round, so no one noticed until it was too late, but neither was very happy.

Alex Relyea

26

Sounds as though this tournament might be the basis for yet another interesting and fruitful new conversation in this forum. Can you post the complete crosstable (in a new thread), including colors?

Bill Smythe

27

Let’s check out the point-count pairings for the crosstable that started this thread. Here it is again, but re-arranged in standings order. Color history is on the right:

1. Alpha 2638 W11 W 5 W 8 --- 3 wbw 2. Echo 2407 W13 D 3 W10 --- 2.5 wbw 3. Bravo 2535 W19 D 2 H - --- 2 bw- 4. Delta 2410 D12 W14 D 6 --- 2 bwb 5. Foxtrot 2302 W15 L 1 W16 --- 2 bww 6. Golf 2287 H - W18 D 4 --- 2 -bw 7. Lima 2119 L 8 W15 W17 --- 2 bwb 8. Charlie 2508 W 7 D10 L 1 --- 1.5 wbb 9. Hotel 2244 D18 H - D12 U - 1.5 w-b 10. India 2236 W16 D 8 L 2 --- 1.5 bwb 11. Juliet 2232 L 1 D12 W18 --- 1.5 bwb 12. Mike 2086 D 4 D11 D 9 --- 1.5 wbw 13. November 2069 L 2 W19 H - --- 1.5 bw- 14. Romeo 1967 B - L 4 H - --- 1.5 -b- 15. Oscar 2048 L 5 L 7 W19 --- 1 wbb 16. Quebec 2015 L10 B - L 5 --- 1 w-b 17. Sierra 1832 H - H - L 7 --- 1 --w 18. Papa 2037 D 9 L 6 L11 --- 0.5 bww 19. Kilo 2200 L 3 L13 L15 --- 0 wbw
First, here are the “raw” pairings and the corresponding undesirability score. In each pairing I have listed the higher-ranked player (not necessarily the player due white) on the left:

[code]Alpha - Echo 3200 (3000 score, 200 color equalization)
Bravo - Foxtrot 80 (color alternation)
Delta - Golf 50000 (already played)
Lima - Charlie 53200 (50000 already played, 3000 score, 200 color equalization)
India - Mike 0
Juliet - November 0
Romeo - Oscar 3200 (3000 score, 200 color equalization)
Quebec - Sierra 0
Papa - Kilo 3200 (3000 score, 200 color equalization)

TOTAL: 112880 undesirability points
[/code]
Now here are wilecoyote’s “natural” pairings:

[code]Alpha - Echo 3200 (3000 score, 200 color equalization)
Bravo - Golf 95 (80 color alternation, 15 transposition)
Delta - Foxtrot 15 (transposition)
Lima - India 3233 (3000 score, 200 color equalization, 33 transposition)
Charlie - Mike 33 (transposition)
Juliet - November 0
Romeo - Oscar 3200 (3000 score, 200 color equalization)
Quebec - Sierra 0
Papa - Kilo 3200 (3000 score, 200 color equalization)

TOTAL: 12976 undesirability points
[/code]
And the Swis-Sys pairings:

[code]Alpha - Echo 3200 (3000 score, 200 color equalization)
Bravo - Delta 108 (transposition)
Foxtrot - Lima 206 (transposition)
Golf - Charlie 3098 (3000 score, 98 transposition)
India - Mike 0
Juliet - November 0
Romeo - Quebec 3113 (3000 score, 80 color alternation, 33 transposition)
Oscar - Sierra 33 (transposition)
Papa - Kilo 3200 (3000 score, 200 color equalization)

TOTAL: 12958 undesirability points
[/code]
Finally, the Swis-Sys pairings except with transposition X instead of transposition Y:

[code]Alpha - Echo 3200 (3000 score, 200 color equalization)
Bravo - Lima 27 (transposition)
Delta - Foxtrot 15 (transposition)
Golf - Charlie 3098 (3000 score, 98 transposition)
India - Mike 0
Juliet - November 0
Romeo - Quebec 3113 (3000 score, 80 color alternation, 33 transposition)
Oscar - Sierra 33 (transposition)
Papa - Kilo 3200 (3000 score, 200 color equalization)

TOTAL: 12686 undesirability points
[/code]
Looks like the computer wins again.

Bill Smythe