Thanks a lot for that explanation. It's a really interesting approach.
Intuitively, I don't believe an increase in the amount of FIDE rated tournament in countries like Sri Lanka and India helps to rectify the problem. These countries simply lack the total amount of points overall they need to be accurate. This is caused by fewer master level players to begin with (as a percentile relative to Europe) to gain points from and a tremendous amount of underrated juniors. More tournaments will likely lead to more deflation as Elo struggles to keep up.
I've scanned this quickly and looks very interesting and well written, but the USCF does not use the GLICKO system. There are similarities, but the RD value in the GLICKO systems does not exist in the USCF system or at least it didn't the last I looked about a year ago.
The USCF rating system can be found here https://new.uschess.org/sites/default/files/media/documents/us_chess_rating_system_specs-2024-03-01.pdf
@spockmscs said in #33:
> I've scanned this quickly and looks very interesting and well written, but the USCF does not use the GLICKO system. There are similarities, but the RD value in the GLICKO systems does not exist in the USCF system or at least it didn't the last I looked about a year ago.
>
> The USCF rating system can be found here new.uschess.org/sites/default/files/media/documents/us_chess_rating_system_specs-2024-03-01.pdf
You are correct, USCF does not use a pure Glicko or Glicko-2 implementation, but they still use dynamic K-factors, which is one of the key improvements suggested. Thanks for reading and pointing this out!
Just a few methods that can be considered that have been used in national systems.
1. Junior addition - For this the Elo calculations of the opponent of a junior uses a higher grade than their published rating. For example if playing a 14 year old that is rated 1500 the rating used for your rating calculation adjustments could be 1640.
2. 200 up rule - This basically makes it so if a player increases their rating by over 200 points in a year their rating is calculated from scratch and all opponent's rating calculations are based on their new rating.
3. K Factor - If you play at least 30 rated games a season your new rating should be very close to your performance rating. This prevents legacy data having an outsized impact.
FIDE Rating system is faulty - Chess Bureau of Investigation
https://lichess.org/@/chess-technik/blog/fide-rating-system-is-faulty-chess-bureau-of-investigation/rcGABwgk
.....
.....
This one is all right. They are higher rated players.
The 1443 was low rated. So one loses, the rating decreases.
Here is the problematic occurence that requires remedial action
Losing to 1443 rated player is worse than winning 2 players with rating 1430 and 1434 ??
How can 2 wins against two 1400+ rated players be less than one loss against a 1400+ rated player with reference to increase of rating number
Winning 2 players : 1430 rated and 1434 gives +13 rating
Losing to 1 player : 1443 rated gives a decrease of -13.20 ???
What type of system is this ?
winning and losing should have same increase or decrease of number
The player has lost. FIDE Rating system creates a deeper wound ?? and pours salt into it
Whatever the explanation is, it will be faulty
A player works hard for say, 3 hours + 3 hours for 2 rounds and wins them, 1400+ rated players
The same player loses after 3 hours to 1 player, his/her rating does not increase but decreases ??
So if a players wins 20 rounds against 1400+ rated players and loses to 10 rounds to 1400+ players, his/her rating will not increase, it will decrease ??
A player plays 200 rounds against 1400+ players, wins them and at the same time, he/she played 100 rounds against 1400+ players and lost them, the rating will not increase but decrease ?
200 round wins against 1400+ rated and 100 round losses against 1400+ rated - will decrease one's rating ??
Nope it works as it should. Remember rating are not money they are not rewards or payments. Point is only estimate your strength
In the example there over 100 giving expected score around 0.7 point in a game. Losing two and winning one = 0.5 point per game and to break even you should have scored 0.7 point per game. so you rating should obviously drop. Amount of points lost is high but given strength of players
@chess-technik said in #37:
It's completely fine. The rating change depends on your opponent's rating relative to yours: If you win against a stronger player, you gain more rating than if you win against a weaker player. If you lose against a stronger player, you lose less rating than if you lose against a weaker player.
This will create an effect that if you win against a weak player, you gain little. But if a weak player wins against you, they gain a lot. If the rating difference between the two of you is large enough, then it can be that one win gives you less than two losses cost you.
Think of it this way: If you played a match over 10 games against a someone rated 2500 Elo, and the match ended 5-5, wouldn't you expect to gain rating through that match? After all, you played 10 games at a performance of 2500, which is much higher than your 1900 (taken from your Lichess profile), so you played better than expected, and should gain rating. At the same time, the 2500-rated player made 5/10 against a 1900, so much less than expected, and they should lose rating.
Following your logic, if you make a 5-5 against a 2500-rated player, both your ratings should remain the same. And equivalently, if you make a 5-5 against a 1200-rated opponent, both of you should also remain the same. So, your rating change would be the same regardless of if you get this result against a beginner or a grandmaster. Would that make any sense?