37A - Underrated Juniors

[Bob Gillanders]

After all the controversy concerning the rating adjustment for Jason Cao, can we now have a civilized debate as to how to handle the issue of underrated juniors?

Is it appropriate to make a rating adjustment for extreme cases?
If so, what is the criteria? Who makes the decision?

Your comments please.

[Christopher Mallon]

I don't have a problem with ratings adjustments where needed, but I do have a problem with using a FIDE rating as the basis. If someone is badly underrated, then give them a new rating based on their most recent 24 games, as if they had started over provisionally.

[Garland Best]

I firmly believe that rating corrections should naturally occur within the system, and not by bypassing the system. As such I believe that the role of the rating auditor is to confirm that the system is actually working as described and not make ad-hoc changes to one's rating. This whole thing was bollixed.

Best solution someone suggested to me was to leave the ratings of juniors as-is, but if their ratiings were below 1000, then use 1000 as the "rating floor" when calculating the effect on an opponent's rating. The result would be that those with ratings below 1000 points would add points to the system as they improved, not take them away.

Regardless, any solution should be STUDIED and not randomly implemented. It appears there are skilled people out there for evaluating these things. Roger Patterson has shown a talent for it, and I'm mathematicaslly inclined. My weakness is a lack of knowledge on ELO`s theories and system. We would also need someone with the programming skills to generate the routines and then run them on the existing database of results gathered in the CFC rating database, and analyse the results.

[Michael Barron]

Thank you, Chris!

You suggestion makes sense, but it doesn't address the root cause.

The problem is - the new players got ridiculously low provisional rating, even if they won all their games.

Just look at this tournament:
http://www.chess.ca/xtableSQL.asp?TNum=201011054

The winner - Qty Zhou - won all 5 games and lost ~500 rating point!
How it could be?

Even Carlsen couldn't perform better!

Qiyu already is A class player.
How could we assign her 1464 rating because she beat unrated opponents?

Unfortunately, even if we use you suggestion, we get the same result.

To correct the situation, we need to change provisional rating calculation - for example, ignore games won against unrated and underrated (more than 400 points difference) opponents.

[Bob Gillanders]

The winner - Qty Zhou - won all 5 games and lost ~500 rating point!
How it could be?

Michael - please, I beg of you, stop quoting this example. The rating drop is completely understandable when you realize the provisional rating of 1946 is based on only one game. The system calculated the new rating correctly.

Citing this example repeatedly unjustly criticizes our rating system.

[Christopher Mallon]

It IS a problem though Bob, with the way the system handles unrated and provisional players. There are other examples. There was an event I played in to get my first Active rating, we had a round robin with ONE rated player. He managed to lose rating points, despite playing all unrated people. That however is unrelated to juniors being underrated.

[Fred McKim]

There is no rating protection in the first 25 games. The formula has always worked this way. That's why it's called provisional.

[Aris Marghetis]

I firmly believe that rating corrections should naturally occur within the system, and not by bypassing the system. As such I believe that the role of the rating auditor is to confirm that the system is actually working as described and not make ad-hoc changes to one's rating. This whole thing was bollixed.

Best solution someone suggested to me was to leave the ratings of juniors as-is, but if their ratiings were below 1000, then use 1000 as the "rating floor" when calculating the effect on an opponent's rating. The result would be that those with ratings below 1000 points would add points to the system as they improved, not take them away.

Regardless, any solution should be STUDIED and not randomly implemented. It appears there are skilled people out there for evaluating these things. Roger Patterson has shown a talent for it, and I'm mathematicaslly inclined. My weakness is a lack of knowledge on ELO`s theories and system. We would also need someone with the programming skills to generate the routines and then run them on the existing database of results gathered in the CFC rating database, and analyse the results.

I agree with Garland. Fix/improve the system, and if appropriate, include mechanisms within the system, to handle extreme cases. Never bypass the system in an ad-hoc manner. Instead, fix/improve the system again. The Jason Cao thing was bollocksed!

[Bob Gillanders]

It IS a problem though Bob, with the way the system handles unrated and provisional players. There are other examples. There was an event I played in to get my first Active rating, we had a round robin with ONE rated player. He managed to lose rating points, despite playing all unrated people. That however is unrelated to juniors being underrated.

Chris - that is a different kettle of fish entirely. When there is a large number of unrated players, the program stops and asks the operator to enter an estimate of the ratings of unrated players. It is a necessary requirement of the programming. The ED enters a value based on the results of that tournament using his best judgement. In your example, the rated player could easily lose rating points if he lost some games. Right!

[Egidijus Zeromskis]

There is no rating protection in the first 25 games. The formula has always worked this way. That's why it's called provisional.

It might be that this drags a player's rating down for a long period. I would be interesting to study the rating settling with only 10 provisional games but larger coefficients (e.g., 50 for 400-1000; 40 - 1001-1401; 32 1401-2199; 16 - 2200-...)