For the sake of completeness, here is the data I sent the chair of the Ratings Committee which led to their suggestion:
The better you explain it Mike, the better I like it. I do like the idea to make it anyone that gets a rating of 500 of lower than the floor of 100. You said one member of the committee supports this idea. If I am right, the rating to start someone out would be fifty points times the age of the scholastic player. If the rating floor is set at 500, that would be the age of a ten year old. If that becomes the norm, than all the JTP tournaments would be un-rated tournaments. That would be acceptable, as I feel coaches are pushing younger and younger scholastic players to play rated tournaments.
Mike,
Has a change been made in the initial prelim rating based on age for the first 8 games been made as you indicate below? Can you explain what it is now and how it works?
Does it carry over to the first 25 games in some way?I believe more info on the prelim rating would be helpful to others in this discussion.
Bob Rieves
"Mark, while preparing the ratings distribution table, I noticed that
we appear to be getting a lot more players rated 100 than in the past,
possibly as a result of the change in the initial estimate based on age. "
A change was made in September of 2005 in the initial estimate that is used for an unrated player based on age. (This was announced in the Forums and elsewhere, and the updated and complete formula is available for download, go to the ‘ratings’ page on the old website for the URL.)
The old formula was 300 + 50 * age (in years), the new formula is 50 * age (in years), with a minimum of 100. This formula is used up to age 26, beyond that the initial estimate is 1300. (In all cases this is a rating based on zero games.)
Thus a player who is exactly 6 years old has an initial estimate of 300/0 and one who is 6 1/2 has an initial estimate of 325/0.
This change was made after data showed that most young players who played in more than one event had around a 300 point decline in their second and third events before their ratings stabilized and started moving up again, thus it appeared that the initial estimate was too high for the current pool of young players.
Douglas, I don’t understand your latest note, and I suspect you still don’t understand the proposal.
Mike, this is were I have been looking at all the time Mike. I do like the minimum rating at 500, but will accept the minimum rating of 400 better then the minimum rating of 100.
As I pointed out in the members issues, ratings under 500 in my judgment are players that should not be in a rated over-the-board tournament. If it increase from 100 as being UNR’s to anyone with a rating of 500 or less, it should make any JTP tournament always being non-rated tournaments. In my view, I like that better as coaches and directors are so obsessed with ratings in the first place with scholastic players.
If the goal is to take away the bad feelings of scholastic players with a rating of 100, why stop at the rating of 100? Why not make it the rating of 400 or of 500, ratings under these levels would be null and void. If it is a rating of 500, that would be someone under the age of ten. That should make all JTP tournaments to become non-rated tournaments, as JTP tournaments are pre-K - grade three.
There is and as far as I know has never been a proposal to have the minimum rating be 500.
A while back I expressed concern to the office about a 100 rating given to a player who happened to lose all of his games in his first tournament. It was explained to me that his rating was set at 400 points below the rating of the lowest player he faced. The player’s true strength being at least 500, I was concerned that this artificial low rating would harm him and everyone else he faced.
This arbitrary procedure ignores true playing strength and is responsible for the army of 100-rated players.
Just today, another new (to the USCF slow rating system) player ended up with a rating of 243P3 after losing all of his games in his first USCF slow tournament. This player had an established quick rating of 717. It happened that I gathered a bunch of the stronger (Chess & Math - rated) local players for an event. (The USCF netted six memberships.) He was the lowest rated of a relatively elite bunch so it was no surprise that he lost all three games. His performance rating was 460 but he’s now stuck with his 243.
The rating system is broken with respect to new players. It needs to be fixed.
My suggestion was that all provisional ratings be floored at 500. If and when players achieve established ratings, then let them sink to a floor of 100. Then they will deserve it. Now, many of the provisional 100-rated players are underrated by several hundred points and this hurts everyone.
But you understand were I was implying from Mike. If it is 400 or if it is 500, it still will make JTP tournaments to become unrated tournaments. The reason I made the rating to be 500, as a scholastic player at the age of 10 is (50 rating points X age) rated at 500 with zero games. Have always accepted that children should not play over-the-board till the age of 10.
If the minimum rating can be 400 I would accept that, prefer the rating be 500. The reason why there are so many scholastic players with the bottom rating of 100, as the movement to younger and younger players has been the obsession with a number of scholastic organizations.
In my judgment, it hurts the chance of the returning under ten year olds to return back to the USCF when the child is much older. It also hurts the over all ratings of players when a child at the age of 6 returns back to the USCF at the age of 12 with a low rating. In my judgment, very few scholastic players under the age of 10 never become an established player.
There are a few under the age of ten that are very strong players; we have one in the state of Michigan. But, he also does not play with his own age group. Most scholastic players the parents want their child to play with there own age group. If the child does show some talent, then I would not have a problem if the child does play in non-scholastic tournaments.
I’m starting to like the idea of the rating of 100 to be UNR, in fact have come to accept it after you explained it much better after your first posting. Even that it will be inflationary, the rating system is just a rating system. As you are only as good as you are with your last tournament.
I don’t know how Charles can claim that player’s “true strength” to be 500 when there is no evidence to that effect in the information we have been given.
A player who loses all his games has demonstrated no evidence whatsoever of his “true strength”. All we know is that he appears to be weaker than the weakest rated player he lost to.
The player in question has participated in 45 Chess & Math-rated events over the past three years and is now over 600 CMA. USCF ratings are usually at least 100 points higher than CMA ratings. When I say a player has a playing strength of at least 500, I know whereof I speak!
What about an attempt at insulating “ratings” from “scholastic ratings”?
How about continuing to rate scholastic players, but somehow have that rating not be considered an official rating for non-scholastic tournaments.
It is convienient in running scholastic tournaments to have a “rating” established. I know that the large number of low rated players does impact things, but having ratings does help us get people paired and adds a little fairness to tie-break situations in the larger scholastic tournmants.
I think at least one member of the Ratings Committee is looking at this proposal as more than just cosmetic. Recall that Ken Sloan responded to me as follows:
In other words, he wants sub-100 players to be able to instantly achieve their performance rating in their next tournament. What bothers me is the discontinuity – players barely above 100 would not be given the same preferential treatment.
It seems to me that a better solution would be a combination of logarithmic re-scaling AND a simple increase in the (already large) K-factor for very low-rated players. A K-factor of 800/N (where N is the number of games played in the new tournament) would achieve Prof. Sloan’s goal quite nicely. (A K-factor larger than 800/N could cause the new rating to actually “overshoot” the performance rating, which would not be desirable.)
Bill Smythe
I took a quick look at about 5000 players who began rated play during 2005 and 2006 and who had a 100 rating from their first event.
About a third of those were still rated 100 after their 2nd event.
Of those who had a rating higher than 100 after their 2nd event, the highest of those was around 600.
Charles, I’m not disputing your claim that your player may be around 500 strength, I’m just saying the USCF has no MATHEMATICAL BASIS for assuming that to be the case.
BTW, if this player had a CFC rating before his first USCF rated event, that information should have been reported to the USCF, as a CFC rating is one of the two non-USCF ratings that the ratings formula permits to be used as prior information in setting a player’s initial rating estimate. (The other is a FIDE rating, of course.)
Mike,
The player had a Chess & Math rating, not a CFC rating. The CFC doesn’t really do much in Quebec, which is the precise reason I am focusing activity in the direction of the USCF.
There are not that many good ideas how to deal with this type of a problem. If you have a scholastic only rating it is still a rating with a rating at the bottom. Sure, you could make it that the scholastic rating can get down to 1000 or just a little lower, or even have a scholastic master rating on the other end.
The problem I have with scholastic tournaments, it is very hard to transfer scholastic players to the regular tournaments. If there was a scholastic rating, it could make it much harder to transfer the best players or the players that are ready to play in the regular tournaments.
Since players can have an obsession with their rating, it can make them active because they feel they are underrated or less active because they understand they are at the top of there game. If there was a scholastic rating, they can become obsessed with that rating. Would these players want a regular rating were they can be hundreds of points lower then their scholastic rating?
The only reason this became an issue, as it looks bad to have the worst rating of 100, why not make it UNR were the kid does not have to feel so bad. Does a player low rating make it necessary to have a therapist?
The office has the discretionary authority to use other national rating systems, but no predefined formulas for them.
TD or players who wish to have a non-USCF rating other than FIDE or CFC used to initialize their rating need to contact Walter Brown in the USCF office, and should be prepared to show how ratings in that system should be converted to appropriate USCF ratings.
I strongly disagree with a blanket statement that children under ten should not play over-the-board tournaemnts. You can suggest that parents and coaches be more diligent in determining whether or not a child is ready for a tournament, but that blanket statement is too severe.
In our area there are a number of kids under the age of ten that are quite a bit higher rated than 500. One organizer has for 15 or more years been running a tournament with three broad grade-based divisions (K-3, 4-5, 6-7) and 8 or 9 rating based sections per division. The most recent one (2/5/06 - event 200502057271 for affiliate H6009015) had 137 K-3 kids with 57 in the sections for those rated over 500. 7 started or finished the tournament rated over 1000, and one of those seven plays regularly at the local community club (the 1300s have to take him seriously if they want a win, and those lower rated are in for a real struggle.
I know that Pennsylvania also has an organizer that runs a similar tournament, and I would expect that there would be others around the country.
Moving the discussion away from scholastic-only tournaments, my community club has a number of young players. We may give more respect to child players than an average club, and that may be the reason we have a noticeable presence of children at our meetings (generally well-behaved - sometimes quieter than the adults). Kids that first show up around 6-8 years old will lose most games, but if they stay to go over the games with their opponents then you start seeing them win around 10-13 years of age, and later competing in our local league when they get over 16 years of age.
Most of the adults will treat the kid-players much more like players than like kids. Those who don’t end up getting very embarrassed WHEN the kids eventually start beating them.
One concession we make to the kids is to adjust (for games involving kids) from our standard Game/90 time control to a more kid-bedtime-friendly Game/60 time control. This is done for our serious but non-USCF-rated ladder play. There is no such adjustment for our rated tournaments, but we still get a number of kids playing in those.
If you want to see an example of a high rated young player, check out Nicholas Nip, who recently represented the USCF in the under 8 section of the World Youth Championships. (He turned 8 earlier this year.)
Nicholas has a very good chance of being an Expert before he turns 9, which we think would shatter the old record for achieving an Expert rating by at least a year!
When thinking about this, first I did not like the idea, and then started to accept the idea, now after thinking about it have rejected the theory. The reason was thinking what Arpad Elo (1903 - 1992) would think about the idea. The rating system was approved in St Louis in 1960, in an era were there was very few scholastic players in the first place. The system was designed for adult players in mind, not designed to deal with players at kindergarten or pre-kindergarten, as this would be the age group were most 100 players should be at their developmental stage.
The rating system is just a method for calculating the relative skill levels of the players. The rating is just really a probability theory, with any probability theory the system can be broken down to fail. Whatever factors are used with a logistic distribution, it is just randomness or uncertainty.
Will say the model has broken down when you start to get to the bottom of the rating system of a rating of 100. It is not the era of the rating system; it is just that the players have transferred from being more of an adult based model to a mixed model of adults and the very young.
I doubt that the members of the Ratings Committee, most of whom have PhDs in areas such as mathmatics or computer science, would agree with your assessment.
The chair, Dr. Mark Glickman, has published several papers on the statistical theory behind ratings systems and their application to areas other than chess, work that takes him well beyond where Dr. Elo was 45 years ago.
If there is a statistical problem with trying to track the ratings of younger players (which are supposed to be predictive), it is that their progress often tends to be in quantum leaps rather than incremental.
While sometimes this happens with adults, it is far more common in players under a certain age, probably 18 or lower. (Now that’s depressing, since I didn’t even start playing rated chess until I was 18, and that was a LONG time ago.)
When that happens, their past results may no longer be all that helpful in predicting their future results, which means that something other than an incremental change in their ratings may be necessary as well. (I think the GLICKO II system, which has been adopted by the Australian Chess Federation, has some provisions for that, and it is one of the justifications for the USCF"s bonus formula.)
Perhaps a more comprehensive resolution to this phenomenon, such as a change to GLICKO II, will be needed in the long run, but I think the current proposal may be a reasonable first step.