Your network blocks the Lichess assets!

lichess.org
Donate

Why chess ratings don't mean what they used to

This article gonna change the history of chess

This article gonna change the history of chess

Well done, great research work and interesting read for any tournament and club player!

Well done, great research work and interesting read for any tournament and club player!

Everyone is saying god reasearch so good reasearch sooooo... good research ... seriously tho good reasearch :D

Everyone is saying god reasearch so good reasearch sooooo... good research ... seriously tho good reasearch :D

The spanish vs indian example is on point haha

The other day, I was speaking with GM Julen Arizmendi. He shared a similar example where a 2300-rated Spanish player faces a 2000-rated Indian player who has a "true" level of 2300. When the Indian player wins, the Spaniard's rating drops significantly below his actual ability.

Now, GMs like Julen are forced to face these FMs who are rated well below their true strength. In these matches, a draw is not a miracle yet it hurts the GM's rating even more.

Julen explained that this creates a cycle where he would need to win an unrealistic number of games just to "break even" on points.

The spanish vs indian example is on point haha The other day, I was speaking with GM Julen Arizmendi. He shared a similar example where a 2300-rated Spanish player faces a 2000-rated Indian player who has a "true" level of 2300. When the Indian player wins, the Spaniard's rating drops significantly below his actual ability. Now, GMs like Julen are forced to face these FMs who are rated well below their true strength. In these matches, a draw is not a miracle yet it hurts the GM's rating even more. Julen explained that this creates a cycle where he would need to win an unrealistic number of games just to "break even" on points.

IMHO there is a second factor that makes the problem worse: FIDE rated events (and games) are way more common in some countries than in others. In our country, essentially every classical open is FIDE rated (and most rapid ones, except small regional or club ones); out of seven levels of team leagues, five are FIDE rated. But that's not the case everywhere and often it's those countries known for their "Elo hungry" heavily underrated (mostly) young players where FIDE rated events are not so common.

Why does it matter? I believe your map showing the "rating offsets" is not telling the whole story. I'm pretty sure 2400-2500 rated Indian players are not that much underrated because most of them already have plenty of international experience and their rating has mostly adjusted to the internationally compatible level. On the other hand, for a young 1500 FIDE rated Indian, the offset may be easily 200 rather than 100. In other words, the offsets for 1500 rated, 2000 rated and 2500 rated are likely to be different. If there were enough FIDE rated games within the country, there would be at least indirect connection for those lower rated players to other countries through higher rated players with enough international experience. But without sufficient mixing within the country, the isolation effect is amplified.

IMHO there is a second factor that makes the problem worse: FIDE rated events (and games) are way more common in some countries than in others. In our country, essentially every classical open is FIDE rated (and most rapid ones, except small regional or club ones); out of seven levels of team leagues, five are FIDE rated. But that's not the case everywhere and often it's those countries known for their "Elo hungry" heavily underrated (mostly) young players where FIDE rated events are not so common. Why does it matter? I believe your map showing the "rating offsets" is not telling the whole story. I'm pretty sure 2400-2500 rated Indian players are not that much underrated because most of them already have plenty of international experience and their rating has mostly adjusted to the internationally compatible level. On the other hand, for a young 1500 FIDE rated Indian, the offset may be easily 200 rather than 100. In other words, the offsets for 1500 rated, 2000 rated and 2500 rated are likely to be different. If there were enough FIDE rated games within the country, there would be at least indirect connection for those lower rated players to other countries through higher rated players with enough international experience. But without sufficient mixing within the country, the isolation effect is amplified.

I'm genuinely excited to see the effect this book has on FIDE.

I'm still personally unsure if the solution of modifying ratings for several key countries is politically viable, even though the data shows it would certainly help solve the underlying deflation tax across the entire system.

Perhaps there would be more buy-in to my suggestion: asymmetric K-factors whenever two players from different countries play each other until you reach master (or close to it). That is, when two players from any different country face each other, the winner gains more points than the loser loses, thus avoiding a "tax" on the more stable player's rating. Even a change of K_winner1.25 and K_loser0.8 would be enough to equalize the geographically disparate regions over time without highlighting specific regions as the "culprit" in an unequal system.

Great lichess blog and great book.

I'm genuinely excited to see the effect this book has on FIDE. I'm still personally unsure if the solution of modifying ratings for several key countries is politically viable, even though the data shows it would certainly help solve the underlying deflation tax across the entire system. Perhaps there would be more buy-in to my suggestion: asymmetric K-factors whenever two players from different countries play each other until you reach master (or close to it). That is, when two players from any different country face each other, the winner gains more points than the loser loses, thus avoiding a "tax" on the more stable player's rating. Even a change of K_winner*1.25 and K_loser*0.8 would be enough to equalize the geographically disparate regions over time without highlighting specific regions as the "culprit" in an unequal system. Great lichess blog and great book.

@mkubecek said ^

IMHO there is a second factor that makes the problem worse: FIDE rated events (and games) are way more common in some countries than in others. In our country, essentially every classical open is FIDE rated (and most rapid ones, except small regional or club ones); out of seven levels of team leagues, five are FIDE rated. But that's not the case everywhere and often it's those countries known for their "Elo hungry" heavily underrated (mostly) young players where FIDE rated events are not so common.

Why does it matter? I believe your map showing the "rating offsets" is not telling the whole story. I'm pretty sure 2400-2500 rated Indian players are not that much underrated because most of them already have plenty of international experience and their rating has mostly adjusted to the internationally compatible level. On the other hand, for a young 1500 FIDE rated Indian, the offset may be easily 200 rather than 100. In other words, the offsets for 1500 rated, 2000 rated and 2500 rated are likely to be different. If there were enough FIDE rated games within the country, there would be at least indirect connection for those lower rated players to other countries through higher rated players with enough international experience. But without sufficient mixing within the country, the isolation effect is amplified.

Great observation! I am addressing the breakdown by rating band directly in the book. Your intuition is spot-on, most of the Indian underrating signal is concentrated in 1400-2000 players, with more balance as you go higher.

@mkubecek said [^](/forum/redirect/post/DozeRhvJ) > IMHO there is a second factor that makes the problem worse: FIDE rated events (and games) are way more common in some countries than in others. In our country, essentially every classical open is FIDE rated (and most rapid ones, except small regional or club ones); out of seven levels of team leagues, five are FIDE rated. But that's not the case everywhere and often it's those countries known for their "Elo hungry" heavily underrated (mostly) young players where FIDE rated events are not so common. > > Why does it matter? I believe your map showing the "rating offsets" is not telling the whole story. I'm pretty sure 2400-2500 rated Indian players are not that much underrated because most of them already have plenty of international experience and their rating has mostly adjusted to the internationally compatible level. On the other hand, for a young 1500 FIDE rated Indian, the offset may be easily 200 rather than 100. In other words, the offsets for 1500 rated, 2000 rated and 2500 rated are likely to be different. If there were enough FIDE rated games within the country, there would be at least indirect connection for those lower rated players to other countries through higher rated players with enough international experience. But without sufficient mixing within the country, the isolation effect is amplified. Great observation! I am addressing the breakdown by rating band directly in the book. Your intuition is spot-on, most of the Indian underrating signal is concentrated in 1400-2000 players, with more balance as you go higher.