When should you berserk in a lichess arena tournament game?

1+0 arena tournament

3+0 arena tournament

Explanation

For details on how I arrived at the above plots, including more plots and exposition, see markdown/1_analysis.md and markdown/3_analysis.md for the 1+0 and 3+0 cases, respectively. These both use the plots generated by analysis.ipynb, where you can find the actual code.

Downloading and processing the data

I downloaded all standard rated games in the lichess database from April 2017 to December 2021. All of the PGNs from April 2017 on have %clk tags, which allows one to determine whether one or both players berserked in any given tournament game by comparing each side's first-move clock time with the time control. The downloading was done in about 16 hours using $ wget -i downloads.txt, where downloads.txt was a text file containing the relevant URLs, like lichess provides here. This amounted to 715 GB of compressed files containing about 2.75 billion games. Next, I used rust-pgn-reader together with rayon to process in parallel the .pgn.bz2 files into CSVs (3 GB in total) containg the player ratings, berserk info, and result for every tournament game with a time control of 1+0 (71 million games) or 3+0 (99 million games). The rust code is in a seperate repository, pgn2csv. This was quite fast compared to other tools like python-chess; the processing took about 7.5 hours in total, which comes out to about 353 million games per hour, or 96 GB/h or 26 MB/s. Of course this will depend on hardware (this was done with an i9-9900K on a wd blue sn550).

Acknowledgements

Thanks to the lichess developers for all of their work, in particular for making the lichess database free and accessible. Thanks to niklasf for rust-pgn-reader.