An excellent point, and one that deserves its own thread.
What sort of pairing criteria should be specified? Presumably, the list would include:
Not pairing the same players twice if possible.
Pairing players within the same score groups whenever possible.
Minimizing score differences whenever 2. is not possible.
Equalizing colors whenever possible.
Alternating colors whenever possible.
Within each score group, pairing top half vs bottom half, sequentially, to the extent possible.
Once the list of criteria is decided, the next step would be to decide how much of one criterion can be traded for how much of another. The present 200-point transposition limit for color equalization, and 80-point limit for color alternation, would be an example, and a good place to start.
First, should anything be added to the above list?
There are only a couple of extras I can add. First, the order of pairing the score groups, starting from the highest and working down. Then who is dropped to a lower group in case of an odd-numbered score group, and the idea that there are no rating difference constraints if transpositions are made to avoid rematches in 1.
The idea of “making things easier for computers” should be replaced by an idea of making things easier for flesh-and-blood directors. Those of us directors who have been around before computers can remember difficult situations to pair in which we were left with trying to choose the objectionable pairing. A simple order of how to deal with these situations will make it easier for beginning TDs. The old idea in Swiss pairings of “rank has it priviledges” was easy to follow in cases of conflicting color allocations, along with equalization having priority over alternation. Today’s rules of taking into account color history may make for better pairings but puts extra pressure on inexperienced directors. All of us TDs above the club level at one time had to pass an exam with at least one pairing problem in it, often as full of twists and turns as any real-life tournament. Although the levels of TD certification make these problems less of a factor, at least for me, it was a mental freeze-up when, as a new Club director, I had for the first time to pair a score group with multiple color conflicts and could not make the idea pairings to resolve them because of rematch issues.
But let’s stick with Prof. Sloan’s “criteria” idea for the moment.
Here is a simple example. Suppose the “raw” (untransposed) pairings in a 6-player score group are as follows:
2000 WBW vs 1880 WBW
1950 BWB vs 1850 BWB
1900 WBW vs 1800 xBW
(In each case the player listed on the left is the higher-rated player, not the player to be assigned white. x stands for no color – bye, forfeit, unplayed game, etc.)
As it stands, colors are bad on all three boards. They can be made better on two of the three, by switching the 1850 with either the 1880 or the 1800. Which of these two switches is better?
I’d switch with the 1880. Colors aren’t that bad on the bottom board anyway because of the unplayed game. And looking ahead to the next round, if this the top score group, the only person why could have an awkward color pattern going into the last round is the fellow with the bye. Since everyone else’s colors are OK, we are assured of being able to give him his due color in the last round.
But then, I’m not a TD. I bet you have a point to be made with this example.
You must have been a TD in a previous life. I am a TD and this is the same pairing I’d make. Using look-ahead, three players are due for black, two for white, and one for neither. Since this is not a situation where more than half are due the same color (in this case, exactly half), the procedure is to correct as many color conflicts as possible using the permitted transpositions and/or interchanges. Swithing the 1880 with the 1850 is a 30-point switch and corrects two conflicts. The bottom board has one player already even in color history because of the unplayed game, so giving 1900 his due color of black does not distort color allocations too much. Switching the 1850 with the 1800 is a 50-point switch for either side, and is less desirable than the previous 30-point switch.
Based on your answer the differential is important, i.e., 50 for the 1850 switching with the 1800 and 30 for the 1880 switching with the 1850. Would you propose a different pairing if the differential was reversed, say the 1800 was 1849 and the other two players stayed the same?
Absolutely. With the former 1800 player rated at 1849, I’d do a three-way transposition. I would pair 1849-2000 and move the other two players down one slot: 1950-1880 and 1850-1900. This makes for better pairings than the previous example since now all color conflicts are resolved.
The correct pairings are 1850-2000, 1950-1880, and 1800-1900. Note that it is an error to claim that the 1800 player is not due a color. That player is due black (see rule 27A5). In fact, the only way a player can be due no color is to have no played games as of the round being paired.
With four players due black and two due white, at least one player will not receive due color. Since equalization takes priority over alternation, transposing the 1880 and 1850 players is correct, since the one remaining color problem is alternation.
This continues to be the correct answer even if the rating of the 1800 player is changed to 1849. The change in rating does not alter the fact that four players are due black and two are due white. Nor does it alter the priority of equalization over alternation.
Starting from the top and working down describes a process, not an objective. If you enshrine that as a “rule”, you can very easily end up getting boxed in when you get to lower score groups.
What you’re describing is how the rule books tended to read through the 4th edition. And they were completely inadequate as a description of how pairings could be done and were being done at the time by both humans and computers. For instance, consider the following top vs bottom half with colors:
1900 W 1780 W
1870 W 1760 B
1830 B 1710 W
1810 B 1640 B
If you paired according to the procedures described in the rule book, you would probably interchange the 1780 and 1810. The one direct switch to correct the color problems is to switch the 1640 with the 1780, which grades out as a 90 point switch on the left side-more than the 80 point limit. However, most experienced TD’s would end up with
1900 W 1760 B
1870 W 1640 B
1830 B 1780 W
1810 B 1710 W
which requires a whole slew of swaps, many of which break up perfectly good natural pairings. Since any permutation can be created by a series of adjacent swaps, you can’t break this down into an evaluation of the permissibility of pairwise swaps. The 5th edition provides a method for “scoring” the changes which comes up with this being a 70 point rearrangement (board 3 has 1830 playing 1780 rather than 1710 and 1780 playing 1830 rather than 1900). Would a human faced with this go through the process of evaluating this to see if it’s actually permitted as an alternative to the interchange? Probably not. And they certainly wouldn’t do this on a 20 or 40 or 100 player score group that might come up in a larger tournament. The human TD would simply have a feel for not having moved anybody all that far. Computer programs, which are the ones actually doing the pairings on the 20 and 40 and 100 player score groups, can’t operate on “feel”.
Regarding the complete color history, you need to keep in mind that that is not used in determining who plays whom. It only comes up in the simpler task of determining who gets which color given a pairing.
Hi Tom:
You espouse keeping the rating difference to the minimum when swapping for color even when requiring multiple swaps. I agree with this philosophy. I have a couple of questions on how you handle this in different context.
In a small score group, the interchange is a 1 point difference and the transposition is 80. Do you believe that the interchange would be a better pairing regardless of our current rules? Under our current rules, would you still use the transposition? If not, where do you draw the line rating difference wise.
In an odd group, you have two satisfactory methods. The first is to drop the bottom player in the higher score group to meet the second player in the lower. The second is to drop the next to last in the upper to meet the top in the lower. Does the rating differential make a difference here?
Regards, Ernie
WinTD has three levels of support for interchanges.
“Avoid Interchanges” means that they will never be used for color corrections (except three in a row), and will be avoided unless needed to fix a more serious problem, which is basically FIDE’s rule regarding this.
“Limit Interchanges” means that they will never be used to fix an alternation problem, even if no set of transpositions under 80 is available.
The standard USCF treatment, where they will not be used if a transposition of less than 80 points is available. Your case is right on the cusp. With the standard USCF rules, the interchange would be taken. Quite a few TD’s probably wouldn’t do that, which is why I offered the two higher levels of avoidance.
Other things being equal, yes. It would take the smaller of the two adjustments.