I’m toying with different pairing systems, and came up with things that might result in the same players being paired against each other more than once during a tournament. I know, from previous threads, that at some point, the rating software automatically changes certain games from tournament games into match games.
Is there someplace I can read about exactly when the games submitted for an event get treated “normally”, and when the match rules kick in? I’m also interested in the effects on category norm calculations if multiple games are against the same opponent, or if the games played at an event are declared (by the rating software) to be matches.
You’re asking about two different but somewhat related situations.
One is a section in which, for a variety of reasons, players A and B face each other more than once. Ladders, for example.
If a player has the same opponent more than twice in a section, then that player is not eligible for bonus points from that event. That’s true even if the player has numerous other opponents in that section.
The other question is ‘what is a match?’ That’s a bit harder to define, but like the Supreme Court, most people know it when they see it.
The simple case is easy: A and B play 2 or more games and those are the only games in that section. Anything else is a hybrid and not everyone might interpret a given situation the same way.
It is ultimately the TD’s responsibility to report matches as such, though the validation software tries to help the TD out. (I"m being kind in using the word ‘help’ here.)
Thanks. So, no bonus points if they meet more than twice, but otherwise, all games are rated normally.(Except as follows.) If they are the only players in a section, then it’s a match, not a section. (I don’t know if the software is sufficiently sophisticated to catch it, but a four player section in which 1 always plays 2, and 3 always plays 4, ought to be two matches. Rest assured I would submit it as matches, not tournaments, if I set that up.)
Other than that, they are just games like any other games? I submit it using the “swiss” format, but in reality, they aren’t played swiss style. I just do manual pairings every round, and probably get lots of warnings for having repeat players?
Could you achieve a norm in a four round tournament when you played only two players, each of them two times?
This is equivalent to a team vs. team match with, say, 1 & 3 on one team and 2 & 4 on the other team. If either of the pairings violate match rules, you will have to clear it with the office. See Mr. Smith’s thread above.
If, in a 4 player section, players 1 and 2 only play each other multiple times and players 3 and 4 only play each other multiple times, then the validation program should recognize that as as match (and so should the TD.)
A norm is possible if you play another player up to twice, but not if you play someone more than twice. Events coded as matches do not earn norms.
The ugly cases are when players 1 and 2 play each other several times but may have games against a couple of other players, none of whom play the same person more than once.
The validation program may not catch such an event, but does it pass the smell test? (Is something fishy going on here?)
Well, I haven’t fleshed out details, and it may never come up in a real event, but here are the thoughts I’m working on right now, where that situation could come up fairly innocently.
I’m working on a modification to my pairing program that runs an optimal matching algorithm. Each player has a cost to be paired with another player, and the goal is to minimize total cost. If the cost assigned is the square of the difference in ratings, you get 1-2, 3-4, pairings.
What I’m thinking about doing is running that same idea, but add an additional cost penalty to any players who had already played each other. In typical pairing systems, they would just not allow that pairing unless there was no choice. In the system I am describing, pairings of similarly rated players who haven’t played each other would be preferred, but repeats would not be prohibited. It would not be in any way “Swiss”, because that day’s record wouldn’t be used to determine pairings at all.
If you used that system and two high rated or two low rated players showed up, with a big gap between those two and the middle of the pack, they could easily be paired more than once, or even twice depending on the size of the rating gap, the number of rounds, and the size of the penalty. The idea behind the pairing system would be to simply eliminate the huge rating swings that sometime occur at my tournaments, which have small numbers of players and a wide range of ratings. It would be a lousy way to award prizes, and would reduce some of the “drama” of the tournament, but it would eliminate 1000 point rating differences, which sometimes happen. Since I don’t award prizes, that’s no big deal.
In such a case, you would have exactly what you described. There might be two players who play each other three times in a four round tournament, while everyone else had no duplicates. Would the rules and/or the rating software decide that those games shouldn’t be treated normally? (Other than no bonus points, and no norms possible for the two who played each other three times.)
Either the section is a match or it is not. If players are participating in some kind of attempt to get around the validation checks for a match (there are some truly egregious events out there in MSA), they should be treated as if their games are a match too. Why reward them for their complicity?
Unfortunately, any attempt to drop the hammer on the bad guys with more stringent validation checks is likely to catch some innocents too. That’s why the office has the discretion to waive the match rules and also the option to cause an event to be rated as if it was a match even if the TD says it was not.
Also, keep in mind the issues raised several match rule threads ago. There are two reasons behind the match rule limitations. One has to do with deliberate attempts to manipulate someone’s rating, the other has to do with the statistical validity of the rating system, both for large numbers of games between two players and for players with provisional ratings.
Too many games against just one opponent is also why players don’t get bonus points if they face the same opponent more than twice in a section.
I think that if it is a tournament, ie you’re pairing whoever shows up not arranging a match, that the office will waive the match rules and rate it. You might just have some explaining to do.
An example might be if you have, say, a U1200 section and only 2 players enter.
Suppose only two players show up for an Under/XXX section and play 4 games against each other. If they have established ratings and are within 400 points of each other, I would see no reason to request a waiver to code such an event as a non-match.
Remember, the match rules are there not just to prevent deliberate ratings system manipulation.
It says in the match rules that the definition of a match “would exclude any games from events that arise as a result of specific pairing situations during a tournament, such as extra games involving a house player, poor turnout in a section, etc. Since the players didn’t enter the event with the intention of playing one person multiple times, the rules regarding participation in a match should not apply.”
True, but that information is not obvious from the crosstable, so any automated review needs to classify a 2 player 4 round section as a match. The TD can request the office waive the match rules limitations, explaining the situation. There are actually solid mathematical reasons why the 50, 100 and 200 point limitations should probably still apply, but that would likely require redefinition of the rating system formula and/or the match rules plus some amount of reprogramming.
