Hello, nice graphs!
In the analysis section you write about "180+0" games, while referring to them before as "3+0". For the average player, it's more usual to speak in minutes.
As for the model the most successful, given how arbitrary the weights seem to be and the low number of players used, isn't that a classic case of overfitting?
In the analysis section you write about "180+0" games, while referring to them before as "3+0". For the average player, it's more usual to speak in minutes.
As for the model the most successful, given how arbitrary the weights seem to be and the low number of players used, isn't that a classic case of overfitting?
Nice analysis! I tried to do something similiar myself- predicting a result of a game based on position and elo, time of the players and etc.
I think the parameters the model got was a little to sparsed (move on time x, eval on turn x and so on).
I think finding a way to measure and make few features out of this can help the model for example you can graph the time disturbution and find a simple function that close to the actual times and make this into a 3 features for example if its a function from order 3 or simpler- features about in which move number the theory ended or first blunder occured, The precentage of time that got spent until move 20 and so on.
This is a great project and I would love to see the code in Git, Great job :)
I think the parameters the model got was a little to sparsed (move on time x, eval on turn x and so on).
I think finding a way to measure and make few features out of this can help the model for example you can graph the time disturbution and find a simple function that close to the actual times and make this into a 3 features for example if its a function from order 3 or simpler- features about in which move number the theory ended or first blunder occured, The precentage of time that got spent until move 20 and so on.
This is a great project and I would love to see the code in Git, Great job :)
Can you provide Github link? I would like to see your methods.
<Comment deleted by user>
Way too complicated for my little brain. Great job though!
ыаа
To make an obvious observation.... Lichess uses Glicko, not elo.
good
nice