I believe the USCF rating system is only valid for the case where
WhiteWin == BlackWin == 2*Draw.
Giving money prizes distorts this of course. In some cases a player will get small change for a draw but big bucks for a win. I think it is practical to rate based on the game points only and let the other chips fall where they may.
Probably not different, the question is how much distortion is caused. Regarding the top players coasting to the finish: At one time the ratings rules said a player who finished clear first in a tournament could not lose rating points. The “implementation” of this rule was a hand-correction to the ratings results. I think it was a good rule. I strongly suspect the reason they changed it was because they couldn’t modify the rating program and the hand-checking was too time consuming. Again a practical decision.
Point distributions different from 1–0.5–0 are fine for experiments, but I’m pretty conservative and would oppose any push for widespread adoption. I’m not at all concerned about pairing variations. It wasn’t 100% clear from the description whether BAP was for score or for pairings. I thought it was for score, in which case I am opposed.
What?! I don’t think you have any idea how hard it is to be a chess professional. You play all weekend and are leading the field and all you need is 1/2 point for clear first? Any self-respecting pro will be standing in front of the TD asking for their check before most of the other players have even started their clocks.
The motivation for the rule was simple. This guy/gal has played well enough to win outright, and if the opposition had been closer they might have been motivated to do even better. Let’s not penalize them simply because they gave their opponent a courtesy draw in the last round.
It wasn’t in the algorithm, it was done manually. Back in the day the USCF used to publish the ratings rules in the December issue. If I recall correctly, that’s where you could find this rule. Probably changed in the late 80s, but others would know more exactly.
My favorite Rating Legend (similar to Urban Legends) was the “rule” that you must gain at least one (or two, depending on who is telling the story) rating point for every win.
Note that in the not-so-distant past, you had to make a distinction between the “rules” and the “implementation”. There used to be a lot of “undocumented features” in the ratings program. Not to mention the frequent “undocumented manual ‘adjustments’”.
Chess Life, November 1981, p 6:
US Open Meetings
Ratings, Sustaining Memberships Among Issues
“In addition, it was decided that the clear, untied winner of a tournament of individual players should not lose rating points providing the tournament includes at least eight players. Anyone losing rating points in such a situation should request an adjustment from the ratings specialist in the New Windsor office.”
“These changes will be implemented by the office as soon as possible, probably before the end of the year.”
It’s similar in its substance but broader in its effect. Rather than affecting one or two games of a few players, it affects every game of the tournament, except maybe for those that cannot possibly count for prizes.
I’ve exchanged a lot of correspondence with Ratings Committee chair Prof. Mark Glickman over the last several years to try to ensure that the white paper describing the ratings system (see math.bu.edu/people/mg/ratings/rating.system.pdf) and the ratings programming are consistent with each other.
The non-mathematical aspects of ratings (ie, the things imposed on it by politicians) include:
Peak rating based ratings floors
Prize money based rating floors
Floating absolute floors (between 100 and 150) based on number of events, wins and draws
Restrictions on who can play in matches and limits on points that can be gained or lost from matches
Another political imposition on the ratings system is the process by which we make adjustments to the ratings of USA flag players who play in FIDE events.
All rating rules, adjustments, exceptions, etc. need to be delineated in clear English language accessable to any USCF player on request. Every USCF player has a right to know how their tournaments will be rated without ‘undocumented features’, ‘back room negotiations’, contracts with Mafia hitmen, etc. These rules should be consistent and not changed every time some rating guru changes underwear.
I think I’ve developed a pretty clear understanding of the rules, adjustments, exceptions, etc. just by reading what’s publicly available to all USCF members. I may be in the minority in that I’ve taken the trouble to actually LOOK at this stuff, but it’s there for anybody to see.
There WERE some unusual rules in the past. There WERE some pretty poor decisions made in the past (“fiddle” points?).
TODAY, I think the rules are fairly clear and reasonable. I think the ratings committee (unpaid volunteers) has been doing a fine job and that they deserve our thanks for their efforts. So do the oaid employees that have been doing the work in some recently difficult times (though they aren’t properly appreciated at the best of times, for that matter).
Maybe for the next major rule update they expand and explain clearly how ratings are calculated with any exceptions, etc.
Personally I think the USCF is doing as good a job as it can given the P/T debacle that has expended a lot of time, energy, and money out of the Federation. Hopefully with this out of the way we can concentrate more on chess.
I believe this document is a fairly complete description of how ratings are computed, though I don’t know if it deals with the adjustments made for participation in foreign FIDE rated events. (Mark is working on updating it to add that.)
I think this one is pretty much up to date, but people who find the mathematical aspects of the first document too hard to follow might have trouble with this one as well.
The ratings estimator on the USCF website does a pretty good job, especially for players with established ratings, I’ve used it on more than a few occasions to see what the possible impact of some event (real or hypothetical) might be:
A question that frequently comes up is how a rating is given to a new player. One explanation: “If someone rated 1000 had a score of 2.5-2.5 against a field whose average rating was 1500 would have an excellent tournament and their rating would go up. That same score would be a disaster for a player whose rating was 2100. A player of what rating would have that result against that field and neither win nor lose any rating points? Here the answer is 1500 and an unrated player would be given that rating based on five games. This ‘performance rating’ is averaged for all results until you have played 25 rated games. At this point your rating becomes established and your more recent results will have more effect upon your rating than games you played some time in the past.”
At one time I had a 2-3 against a 1500 field worked out to 1420, but I haven’t cranked it out recently. I’m not sure if 25 is the right number of games to become established.
There are actually two different numbers that enter into it.
Until you have more than 8 games as part of your pre-event rating (or if your games to this point have been all wins or all losses), you are rated under the ‘special’ formula, which is similar to the old ‘Win+400/Loss-400’ performance rating formula.
After that, you are rated using the ‘regular’ formula, ie, the sum of (actual - expected for each game) times the K factor in effect.
However, until you have played 26 games, your rating is still considered ‘provisional’, though the number 26 is no longer used within the context of computing your new rating.