|
|||||||||
|
|
|||||||||
|
Chessville
Advertise to Single insert:
|
The Black Hole in Chess Ratings By Russ Mollot (founder, Chess Express Ratings) Miss reading the original
articles? Reader's Responses
Gene Milener writes: I read your article about ratings deflation on the Chessville website. I must be missing something because I do not see any problem. My probably flawed understanding is: When a player enters his first rated USCF tournament the system treats him as a presumed 1500, right? It has nothing to do with his age or scholastic chess activities. I have never understood why USCF feels it must add or subtract rating points from a fully rated player based on his winning or losing against an new unrated player. The system should use the win or lose to affect the provisional rating of the unrated player, and do no more than a token +1 or -1 for the fully rated player. That way skews and guesswork are eliminated. As long as a player has a merely provisional rating, his fully rated opponents should have minimal changes in their ratings from playing him. REQUEST: Could you please reply pointing out to me my misunderstandings or things I am overlooking. I find the topic interesting. I like your article, and I will re-read it. But for me at least it seems to assume too much prior knowledge in the audience about the mechanics of the rating system. Thank you. Russ Mollot replies: This idea has much merit. It also happens that CXR uses a similar approach: when a provisionally-rated (or unrated) player plays an opponent who has an established rating, the result is heavily weighted for the new player, but is "light-weighted" for the established player. The veteran's rating gets tweaked, as Gene suggests, with a low "K" factor (K=6), but not just by 1 point as you suggested. But we're on the same wavelength. You also suggest that unrated players should start with a 1500 rating. I presume you used this figure because Arpad Elo intended 1500 to represent an "average" rating. Actually, the USCF does not do this, but has a mathematically sound alternative that takes a long, long time and many tournaments and a lot of entry fees, game fees, and membership fees. At CXR we generally recommend an initial provisional rating of 1200 for new players (somewhat lower for scholastic players, as will be discussed in Part III of the article).
Brad Rosen writes: I found your piece in Chessville to be interesting. However, I didn't quite understand what you meant by unregulated scholastic ratings. Are you referring exclusively to a players USCF rating, or are you making reference to the various local rating systems out there like those for Washington State scholastics and various highs school systems? And who are you suggesting step in and regulate scholastic ratings? --- or are you just suggesting that the USCF (or other chess raters out there) revisit the way they calculate ratings? I look forward to
your response. By the way, are responses to your piece posted anywhere?
Thank you for attention and consideration. Yes, by unregulated scholastic ratings, I refer to the myriad local rating systems (Washington/Oregon systems are among the best of these, I believe). The problem is not so much with the formulas they use; it is the widespread practice of assigning rock-bottom initial ratings to the kids, when it is nearly impossible for them to climb out of this hole. Even the strongest of those players graduating from their school chess programs end up severely under-rated. I do think that scholastic ratings need to be regulated, but by whom? I don't know. My main purpose at this point is to stimulate some discussion, and seeing your letter has given me some encouragement. Your comments are much appreciated!!
James Sawaski writes: I saw Russ Mollot's article on chess ratings at Chessville and I could not agree more with him. The ratings issue is serious for the serious tournament player. A lot of rated chess players would agree that ratings have become very badly distorted. Many of the players I used to play with in the 1980s either (a) don't play OTB chess anymore because of the system or (b) are rated about 100-300 points less than they were before. That's not right. The whole rating deflation promotes negativity towards going to tournaments and this costs tournaments and tournament directors and tournament organizations (such as USCF) money and attendance. I am an 1858 rated player and have been victim of a deflationary rating curve for a long time now. About 15 years ago I was higher rated out of high school (1895) than I am today ... and I am MUCH stronger today than I was 15 years ago. As a 20 year tournament veteran, I remember when 600 was a pathetically low rating. NOW I see kids at tournaments that are rated 100!! There can't possibly be enough bonus points built into USCF's system to cover this type of player’s improvement – even if he improves a little bit, the points have to be made up from somewhere. The fact is – the rating curve is STILL deflationary. Often I think one very simple solution is for Performance Ratings to be used in certain circumstances. We all hate that 800 rated chess player that has a 2050 performance of 4 out of 5 in a weekend Swiss. Nothing is more devastating than to lose 32 rating points to that person. If the performance rating was used in that special circumstance, then the sting wouldn't be so bad for the loss and the bonus points would start to kick in for the 800 fast as deservedly so. I don’t mind losing the actual game, but give me credit for losing to his 2050 performance rating, NOT his 800 deflated rating that he earned months before downloading a chess training program and honing his tactics to ‘killer’ level. Another issue I feel that is important is the master title. Yes, I believe it is statistically impossible for the average ‘Joe’ to work hard and become a master at chess in USCF’s deflated rating curve. Online I can hold high ratings of 2100 and 2200. I know online ratings can't really ever be official because some people cheat and use computers, but there still has to be some contradiction in achieving a master or expert rating online and NOT being able to do it over-the-board if the system is really so perfect. Of course, one does not play the number of OTB games that they do online, so maybe the online rating is actually more accurate? Yeah, I’ve read all the complaints about the so-called dumbing down of the USCF ratings, “We have to protect our master title!” Well, if you think about it, a 1600 rated player today when compared to the general public is light years beyond anyone at their skill. Therefore, I think 1800-2000 should be considered Expert; 2000-2100 a one star Master, 2100-2200 a two star Master, 2200-2300 a three star Master, etc. up to IM and GM. This would enable players to achieve higher "ranks" more quickly, provide more incentives than what exists today and the ‘master’ rank would be available and meaningful to hard working group chess players. Plus, there would still be a way to distinguish a strong master, from a weak one and many variations of course to keep the best players at the top and the less talented below. This is not unreasonable when you think of players being rated 100!! Look at other things that offer a master title. A master plumber certainly wouldn’t have to work 1/10 as hard as I have at chess to become a master at what he does. Yeah, I understand chess is not plumbing – but statistically, my meager ‘Class A’ rating is good enough to defeat 99%+ of all the people in the world. As it is, on USCF’s 2002 rating distribution chart, an 1800-1899 player was in the top 92 percent!!! From 1900-1999 it’s 94.8%!! It is absurd and crazy to think that there is no master recognition or some other honorary title for achieving such a tremendous statistical goal! It’s not like I’m asking to be a Grandmaster – yes, keep that for elite players, but please acknowledge that quality chess is played at the 1800+ level in this day and age of computers, DVD instruction, databases and actually quite well written books. The bottom line is, better ratings means more chess players and more chess players means more revenue for USCF. I hope the CXR system is successful in fixing the rating problems. I like the idea of multiple gauges (separate metrics for different areas). I have a master’s degree in mathematics and I’m an avid mathematician (not just a writer) and I can read and understand the math papers of Mark Glicksman, but I can't do much past that. I do want to note that I am pro-USCF and I am a life member. They have my loyalty. However, they need to pay attention to this. I am not weaker than I was 15 years ago. I have dedicated myself for 25 years to studying and playing the game of chess. I deserve some sort of recognition for the hard work I have achieved thus far as I have played well in tournaments and have proven myself to an acceptable level of quality. If USCF continues to make this an elitist’s game, they will continue to struggle financially and people will play online where they CAN earn their master ratings. There is room in USCF membership for more people to have a master title and I think the star ranking of masters makes for a good compromise to keep the strong players at top, yet allow the hard workers the just reward they have deserved all these years. Remember how fair bell curves were 30 years ago? That’s what master ratings are today – just one big bell curve. A tiny portion ‘IN’ and the rest is nothing. Chess is still in the dark ages regarding ratings and USCF really owes it to its playing members to update. If they refuse, well, then maybe they someone else should do the ratings for events? This is the 21st century and as chess players we should have well formulated ratings that are properly calculated, are fair, accurate and actually mean something. Thank you for listening, I like what you guys are doing I hope to see more articles on ratings in the future. [Editor's note: Mr. Sawaski is the author of the book The Chess Team, and numerous short stories.] Russ Mollot replies: Dear James, As you point out, there is something wrong with a rating system where veteran players can lose 100-300 points even though their game has improved (or, at least we can presume that so many players didn't suddenly get "dumber"). I am convinced that the "Black Hole Effect" from low scholastic ratings is responsible for this monstrous deflation. As you point out, the "online ratings" are not very accurate, but generally do result is higher ratings than USCF's system for OTB chess. Still, I have my doubts about the online rating systems and wonder whether they bias the results so that most people's ratings have an upward tendency. It is in their interest to have people play more, and higher ratings do encourage folks to play. USCF's monopoly on OTB chess ratings is being challenged by Chess Express Ratings ("CXR") which can be found at: www.chess-express.com. Part II of the article will probably appear a week from Sunday; Part III will probably appear one week later. Thank you again for your most interesting comments!
|
The
|
|||||||
|
|||||||||
