Your network blocks the Lichess assets!

lichess.org
Donate

From 0.7% to Champion: Pragg Wins the UzChess Cup

@GiveEverythingToWin said in #8:

Lets say someone has a score of 5-1 vs somebody, but has a lower rating. Then the ratings wouldn't be the entire truth of the probability of the outcome of the game. I'm no expert in this area but I was thinking like giving the player with the higher score (+3 could be the minimum) could ge like a 20 elo rating buff.

The problem I see is that there is no "proper math" for this. Just adding a few Elo points because the player did good in the past is just not very accurate. Also what if the games are old? What if he had a worse Elo back then? So many things to keep into account to only make the end results less accurate IMHO.

@GiveEverythingToWin said in #8: > Lets say someone has a score of 5-1 vs somebody, but has a lower rating. Then the ratings wouldn't be the entire truth of the probability of the outcome of the game. I'm no expert in this area but I was thinking like giving the player with the higher score (+3 could be the minimum) could ge like a 20 elo rating buff. The problem I see is that there is no "proper math" for this. Just adding a few Elo points because the player did good in the past is just not very accurate. Also what if the games are old? What if he had a worse Elo back then? So many things to keep into account to only make the end results less accurate IMHO.

This is very coolllł

This is very coolllł

Carlsen, Caruana, Nakamura, Gukesh, MVL, Nepo, Giri, I feel like this tournament is missing a few people.

Carlsen, Caruana, Nakamura, Gukesh, MVL, Nepo, Giri, I feel like this tournament is missing a few people.

@ChessMonitor_Stats said in #11:

The problem I see is that there is no "proper math" for this. Just adding a few Elo points because the player did good in the past is just not very accurate. Also what if the games are old? What if he had a worse Elo back then? So many things to keep into account to only make the end results less accurate IMHO.

I agree with what you're saying but what I mean is that the elo doesn't say everything about the encounter. I don't have a genius idea how to implement past results etc., that's where I would hope you had an idea :)

Another thing is that at their level it is quite an advantage to be white, so maybe one should add that into the formula you have. For example in the masters database white has around a 33% win rate, a 44% draw rate and a 23% loss rate.

@ChessMonitor_Stats said in #11: > The problem I see is that there is no "proper math" for this. Just adding a few Elo points because the player did good in the past is just not very accurate. Also what if the games are old? What if he had a worse Elo back then? So many things to keep into account to only make the end results less accurate IMHO. I agree with what you're saying but what I mean is that the elo doesn't say everything about the encounter. I don't have a genius idea how to implement past results etc., that's where I would hope you had an idea :) Another thing is that at their level it is quite an advantage to be white, so maybe one should add that into the formula you have. For example in the masters database white has around a 33% win rate, a 44% draw rate and a 23% loss rate.

@GiveEverythingToWin

It's a good idea but I also see no real way to implement it. Maybe one day someone has a good idea on how to approach it ;)

Regarding white's advantage: I actually add 35 Elo points to the white player in the simulation resulting in slightly higher win chances for white.

@GiveEverythingToWin It's a good idea but I also see no real way to implement it. Maybe one day someone has a good idea on how to approach it ;) Regarding white's advantage: I actually add 35 Elo points to the white player in the simulation resulting in slightly higher win chances for white.