lichess.org
Donate

Decisive Advantage but Lost on Time?

Thank you for yours data analysis series.

Ideas of questions:

  1. I understand that there might be lots of data in the bullet class of data. And also a diverse set of position explored compared to other less impulsive testing time controls. But, then I wonder how much, in each case of your series, hypotheses or question and their tendencies are also reflected in other time controls. I guess one could apply similar methods as you did, although all time related parameters in the questions might have to be adapted.

  2. I have a more basic question: that of game depths distributions in any time control, and as a function of 2 variables that of the rating pair of variables.. I guess bullet might be already a place with great spread.

Has it been done elsewhere and published, or is it common knowledge already? I guess one could look at LC0 training games while the self play batches progress, but there might be a tendency for the 2 variables to not vary too much apart as part of the batching criteria (something like win ratio threshold)...

But with humans would be better defined rating ranges. I would be glad for pointers to data analysis reports with enough display of underlying data sources and manipulations used.

Or you might already have done such things any way on your tribulations behind the blog presentations.

maybe other formulations of similar questions might be more informative. or done. I am not done reading your blog presentations here on lichess. I only looked fast at a few of them. to see the themes.

Thank you for yours data analysis series. Ideas of questions: 1) I understand that there might be lots of data in the bullet class of data. And also a diverse set of position explored compared to other less impulsive testing time controls. But, then I wonder how much, in each case of your series, hypotheses or question and their tendencies are also reflected in other time controls. I guess one could apply similar methods as you did, although all time related parameters in the questions might have to be adapted. 2) I have a more basic question: that of game depths distributions in any time control, and as a function of 2 variables that of the rating pair of variables.. I guess bullet might be already a place with great spread. Has it been done elsewhere and published, or is it common knowledge already? I guess one could look at LC0 training games while the self play batches progress, but there might be a tendency for the 2 variables to not vary too much apart as part of the batching criteria (something like win ratio threshold)... But with humans would be better defined rating ranges. I would be glad for pointers to data analysis reports with enough display of underlying data sources and manipulations used. Or you might already have done such things any way on your tribulations behind the blog presentations. maybe other formulations of similar questions might be more informative. or done. I am not done reading your blog presentations here on lichess. I only looked fast at a few of them. to see the themes.