Oh, I had a model some years ago that would download a user's fine grained time data for all games. I had time series in python for my and my opponent's time and used it to animate some bullet games!
I stopped the animation just short of when I realized I had to implement promotion animation and I was like lazzyy. And put it off forever. You guys do things similar sometimes our best projects are just waiting unfinished?
Ratings only make sense within the rating pool they were measured in, they're not universal measures of playing strength.
And besides... Lichess does not use elo. So don't call it elo.
And besides... Lichess does not use elo. So don't call it elo.
hello
Wwhhyy??
<Comment deleted by user>
Maybe try an lstm for the time series analysis?
I would imagine that the evaluation score and move time deltas would relate to a players rating.
You also need to train it on a large dataset. I think significantly larger than what you used.
I would imagine that the evaluation score and move time deltas would relate to a players rating.
You also need to train it on a large dataset. I think significantly larger than what you used.
good point noticing lichess uses "ELO" word in own database. Perhaps it might be PGN header tradition?
@Molurus said in #13:
>
yes about the pools, it does not change the possibility that within the pool context of the database there might be predictor signals among the data, other than the win ratio. It is a legitimate questino to ask or explore, and the purpose is not necessarily to find a positive answer, it is equally informative to assess hypotheses in any direction.
It does not mean that the statiscial associations that could have been found could have been generalized to other pools with their ratiings though.. Which is why people invoked the pool dependency.. If alll the data is from the pool of games which is at the basis of the ratings themselves, then I think one should ask about any sort of associations.. I just wonder what does eval by time mean?
Also, anywhere in the analysis is SF position evalution involved? (such a blunder, inac.. mistake SF score converted binning might be). How much of a game terminal end point is related to each position moves quality along the whole game... "reversing" the question, how much of engine ELO ratings (based on game terminations) is making it a reliable move quality measure for all moves in a game? Are there not some critical position where a bad move might seal the deal, even if the agony might be stretched with prefect moves from there, by the losing side?
>
yes about the pools, it does not change the possibility that within the pool context of the database there might be predictor signals among the data, other than the win ratio. It is a legitimate questino to ask or explore, and the purpose is not necessarily to find a positive answer, it is equally informative to assess hypotheses in any direction.
It does not mean that the statiscial associations that could have been found could have been generalized to other pools with their ratiings though.. Which is why people invoked the pool dependency.. If alll the data is from the pool of games which is at the basis of the ratings themselves, then I think one should ask about any sort of associations.. I just wonder what does eval by time mean?
Also, anywhere in the analysis is SF position evalution involved? (such a blunder, inac.. mistake SF score converted binning might be). How much of a game terminal end point is related to each position moves quality along the whole game... "reversing" the question, how much of engine ELO ratings (based on game terminations) is making it a reliable move quality measure for all moves in a game? Are there not some critical position where a bad move might seal the deal, even if the agony might be stretched with prefect moves from there, by the losing side?
anyone care?