Alex McFarlane wrote: ↑Sun Sep 05, 2021 9:46 am
It does, but only fairly recently.
However, in theory, you can ask for your pairing system to be accepted. It was suggested that if the CAA method was computerised there was a fair chance that it would be. Unfortunately, the attempt to automate it failed to achieve the required level and was abandoned when the programmer lost interest.
FIDE from time to time does suggest that it might introduce the median of the score group as the player to get the bye.
There are a number of 'problems' with the FIDE system which the CAA one tried to avoid. The FIDE system (almost) always works down and ignores the actual rating. This means that if the pairing should be 5 v 23 but these players had already met, the FIDE system will look at 5 v 25 (assuming 24 is due white) and pair them rather than looking at 5 v 21 as well. The CAA system considers both options and chooses the one which minimises the rating difference. So even if 21 and 23 had the same rating the FIDE system would not swap the players. However, it is this that makes it so difficult to program.
If there are any programmers out there willing to give it a go ...
Okay! Do you have a preference as to programming language?
I have a feeling that the way to go with this is not to create a "smart" pairing algorithm that tries one pairing, calculates that it's undesirable, and tries another one. Instead, to start with a matrix of costs for every possible pairing, and have the algorithm solve for the lowest total cost. So 5 v 23 might have a cost of -100 for being the correct pairing according to the score groups, but +1,000,000 for having already been played. 5 v 25 might get -1 for being the next pairing "down" from the correct pairing, but +10 for a rating difference of 10, whereas 5 v 21 would get 0 on both position and rating difference, so 5 v 21 ends up as the cheapest pairing for now. If 5 v 21 causes the appearance of a +1,000,000 elsewhere in the pairings, the algorithm would eventually reject it and replace it with 5 v 25.
Then you can implement any change to the pairing rules by changing the cost matrix instead of changing the algorithm, which is miles easier. You can also ask the program "why is 5 not playing 21?" and it will tell you the cheapest set of pairings containing 5 v 21, sorted by cost. If there's a +1,000,000 in there, you have your answer as to why 5 is not playing 21.
Donate to Sabrina's fundraiser at
https://gofund.me/aeae42c7 to support victims of sexual abuse in the chess world.
Northumberland webmaster, Jesmond CC something-or-other. Views mine. Definitely below the Goodall Line.