(I suppose if there ever would be a thread about being against everything he would be against that too.) @krasnaya
HaHa! Good one. Now it's getting a bit more complicated.
HaHa! Good one. Now it's getting a bit more complicated.
@krasnaya , yes true. Different time categories require different skills, and produce a different type of battle. For instance Andrew Tang (penguingim1) has once scaled #1 bullet player in chess.com, when they had introduced 10s games or so, beating the ratings of Magnus and Naka, who have FIDE ratings over 350 points above him.
As opposed to him, my ultrabullet rating is 400 points below my rapid rating, which is 300 points below my correspondence rating. Maybe correspondence ratings can also compare well to standard FIDE ratings. Maybe even tactical ratings can help to increase precision.
As opposed to him, my ultrabullet rating is 400 points below my rapid rating, which is 300 points below my correspondence rating. Maybe correspondence ratings can also compare well to standard FIDE ratings. Maybe even tactical ratings can help to increase precision.
@EvilChess
I have made the same argument with no success. May as well bang the head against a wall. Off course a correspondence rating would be a far better data set to use than a blitz rating when a comparison is sought to OTB 90 games.
Have you plugged in the ratings to the formula and observed the resulting trends and predictions?
For example try 1078 blitz rating and 1169 classical (when the OP started the thread, classical was any game over 8 minutes, practically blitz) and view the predicted Fide rating. (should predict a Fide 1000) However, the result is much higher than the blitz rating! Try 2278 and 2369 and the prediction (should predict 2200) is much lower than the blitz rating! Makes sense right? LOL
+78 and +169 are from the OP's original premise on 1st post.
"Since there are many (fake) outliers, summarizing these data using the average would be misleading. We use the median instead:"
* A typical (median) user's FIDE rating tends to be 78 points lower than her Lichess Blitz rating
* A typical (median) user's FIDE rating tends to be 169 points lower than her Lichess Classical rating
OP
The formula makes the expected prediction for players rated 1900/2000. Ratings higher/lower and the predicted Fide rating becomes increasingly skewed towards each end of a plotted line.
If the formula is valid for only 1 class of players that fall within the median of +78 and +169, how can it possibly be valid for players that have a different skill set? ie: A blitz rating that is higher than their classical rating, which certainly applies to a great many players.
I have made the same argument with no success. May as well bang the head against a wall. Off course a correspondence rating would be a far better data set to use than a blitz rating when a comparison is sought to OTB 90 games.
Have you plugged in the ratings to the formula and observed the resulting trends and predictions?
For example try 1078 blitz rating and 1169 classical (when the OP started the thread, classical was any game over 8 minutes, practically blitz) and view the predicted Fide rating. (should predict a Fide 1000) However, the result is much higher than the blitz rating! Try 2278 and 2369 and the prediction (should predict 2200) is much lower than the blitz rating! Makes sense right? LOL
+78 and +169 are from the OP's original premise on 1st post.
"Since there are many (fake) outliers, summarizing these data using the average would be misleading. We use the median instead:"
* A typical (median) user's FIDE rating tends to be 78 points lower than her Lichess Blitz rating
* A typical (median) user's FIDE rating tends to be 169 points lower than her Lichess Classical rating
OP
The formula makes the expected prediction for players rated 1900/2000. Ratings higher/lower and the predicted Fide rating becomes increasingly skewed towards each end of a plotted line.
If the formula is valid for only 1 class of players that fall within the median of +78 and +169, how can it possibly be valid for players that have a different skill set? ie: A blitz rating that is higher than their classical rating, which certainly applies to a great many players.
FIDE Rating = 187 + Lichess Classical Rating X 0.38 + Lichess Blitz Rating X 0.48
Here is the problem with the conversion formula:
A number (187) is being added for all examples.
For ratings approx. 1000 adding 187 is almost 20% of the resulting prediction. For ratings approx. 2000 adding 187 is only 10% of the resulting prediction. This explains simply why for ratings of 1000 the predicted Fide rating is higher than the starting blitz rating and for ratings of 2000 the predicted Fide rating is lower than the starting blitz rating.
This number 187 MUST BE written as a fraction of the starting blitz/classical ratings and can not be applied equally to all ratings as a constant number (it will have a greater effect the lower the rating), but must be a variable to make for consistent results.
Next, common sense indicates .38 x classical and .48 x blitz is backwards. It places a greater emphasis on blitz than a classical rating, which to anyone's logic is more closely related to OTB play.
Here is the problem with the conversion formula:
A number (187) is being added for all examples.
For ratings approx. 1000 adding 187 is almost 20% of the resulting prediction. For ratings approx. 2000 adding 187 is only 10% of the resulting prediction. This explains simply why for ratings of 1000 the predicted Fide rating is higher than the starting blitz rating and for ratings of 2000 the predicted Fide rating is lower than the starting blitz rating.
This number 187 MUST BE written as a fraction of the starting blitz/classical ratings and can not be applied equally to all ratings as a constant number (it will have a greater effect the lower the rating), but must be a variable to make for consistent results.
Next, common sense indicates .38 x classical and .48 x blitz is backwards. It places a greater emphasis on blitz than a classical rating, which to anyone's logic is more closely related to OTB play.
@mdinnerspace ,
If you have a better idea, kindly ask for the pyton source code he offered, collect the data and dump it into excel then try your own formulas. That's easy.
Note that the sum 0.38 + 0.48 is just 0.86, so this formula decreases Lichess ratings by 14%. This is why it can add 187 and still produce a lower Fide rating.
For whatever formula you try, calculate the standard deviation (equation below) between calculated FIDE rating of each player and the reported FIDE rating of that player. That will measure your formula's precision.
wikimedia.org/api/rest_v1/media/math/render/svg/00eb0cde84f0a838a2de6db9f382866427aeb3bf
If you have a better idea, kindly ask for the pyton source code he offered, collect the data and dump it into excel then try your own formulas. That's easy.
Note that the sum 0.38 + 0.48 is just 0.86, so this formula decreases Lichess ratings by 14%. This is why it can add 187 and still produce a lower Fide rating.
For whatever formula you try, calculate the standard deviation (equation below) between calculated FIDE rating of each player and the reported FIDE rating of that player. That will measure your formula's precision.
wikimedia.org/api/rest_v1/media/math/render/svg/00eb0cde84f0a838a2de6db9f382866427aeb3bf
@EvilChess
Wrong. It produces a higher Fide rating (than the inserted blitz rating for all players <2000. I have given the numbers numerous times. Do it yourself. Observe the results.
Wrong. It produces a higher Fide rating (than the inserted blitz rating for all players <2000. I have given the numbers numerous times. Do it yourself. Observe the results.
@EvilChess is right, of course. And it turns out that what he calls "the standard deviation between calculated FIDE rating of each player and the reported FIDE rating of that player" is minimized by the least squares estimator (which is what I used).
But please never change @mdinnerspace. We love you.
But please never change @mdinnerspace. We love you.
The least squares estimator is primarily concerned with +/- 100. If a profile says their Fide rating is 1850 but their classical lichess rating is 1150... the "least squares estimator" assigns a 1750 estimate. (The estimator is of course missing a few squares in the brain.) It truly is astounding the amount of 3000+ Fide rated players (according to profiles) that play on lichess. 100's of Magnus Carlson's, all in disguise.
"I have made the same argument with no success. May as well bang the head against a wall. Off course a correspondence rating would be a far better data set to use than a blitz rating when a comparison is sought to OTB 90 games." @mdinnerspace
No.
Correspondence Chess is 95%+ determined by:
- How good your opening sources are
- How much time you spend on the games
This isn't comparable to OTB at all, where your opening skill is related to your memory (rather than just what books you own) & the vast majority of people take the games relatively serious, rather than just Blitzing out random moves (plus of course, your time is limited).
Many people who play Correspondence just play it as a way to get some moves in during work break or similar. Naturally, if you actually spend a day+ thinking about your move, you will just beat those people, even if they're actually rated 400pts higher OTB than you are.
I climbed to 2200 Corr "trivially" (and only stopped there because I stopped playing it); with the vast majority of those games being simple opening crushes. This might have happened OTB too, but you could've exchanged me for a random 1300 with access to the same books, and it wouldn't have changed the result (while of course the 1300 OTB won't be remembering any of the lines).
Corr Rating has pretty much nothing to do with any "real" rating at all.
No.
Correspondence Chess is 95%+ determined by:
- How good your opening sources are
- How much time you spend on the games
This isn't comparable to OTB at all, where your opening skill is related to your memory (rather than just what books you own) & the vast majority of people take the games relatively serious, rather than just Blitzing out random moves (plus of course, your time is limited).
Many people who play Correspondence just play it as a way to get some moves in during work break or similar. Naturally, if you actually spend a day+ thinking about your move, you will just beat those people, even if they're actually rated 400pts higher OTB than you are.
I climbed to 2200 Corr "trivially" (and only stopped there because I stopped playing it); with the vast majority of those games being simple opening crushes. This might have happened OTB too, but you could've exchanged me for a random 1300 with access to the same books, and it wouldn't have changed the result (while of course the 1300 OTB won't be remembering any of the lines).
Corr Rating has pretty much nothing to do with any "real" rating at all.
It is certainly not by accident that all players are arranged in a cloud or a stripe say about 100 points above their FIDE elo. It‘s the sheer mass that counts and not the outliers. So the fit is good, the prediction should work well for most people and poor those who think that outliers wag the dog.
This topic has been archived and can no longer be replied to.