lichess.org
Donate

Lichess Marathon Statistics

Well now I know which marathon to play if I don't want to again spend 15 hours to win a trophy.

Well now I know which marathon to play if I don't want to again spend 15 hours to win a trophy.

@Whitedancingrockstar said in #2:

Well now I know which marathon to play if I don't want to again spend 15 hours to win a trophy.

The answer being "none of them"? :) In general the increment arenas seem a bit easier than the non-increment marathons. (Maybe because people like me try catching up by berserking, get flagged by lower-rated players and ragequit.)

@Whitedancingrockstar said in #2: > Well now I know which marathon to play if I don't want to again spend 15 hours to win a trophy. The answer being "none of them"? :) In general the increment arenas seem a bit easier than the non-increment marathons. (Maybe because people like me try catching up by berserking, get flagged by lower-rated players and ragequit.)

No no, see, playing an increment marathon would decrease the 15-hour work-day to the fantastic 13 or 14 hours instead.
Or maybe I should just git gud.

No no, see, playing an increment marathon would decrease the 15-hour work-day to the fantastic 13 or 14 hours instead. Or maybe I should just git gud.

Hey, I'm the guy that lowered the rating "requirement" to get top 50 in bullet marathon! Kinda proud to "falsify" the statistics haha. I wish everyone (myself included) to stay healthy for the next one cuz we need it lul

Hey, I'm the guy that lowered the rating "requirement" to get top 50 in bullet marathon! Kinda proud to "falsify" the statistics haha. I wish everyone (myself included) to stay healthy for the next one cuz we need it lul

I suppose I should inquire what source code produced these charts and graphs, in case people want to add machine learning-based predictions but don't want to rewrite Lichess API client code.

I suppose I should inquire what source code produced these charts and graphs, in case people want to add machine learning-based predictions but don't want to rewrite Lichess API client code.

Good work, blessings!

Good work, blessings!

Great job! Was really helpful!

Great job! Was really helpful!

@Toadofsky said in #6:

I suppose I should inquire what source code produced these charts and graphs, in case people want to add machine learning-based predictions but don't want to rewrite Lichess API client code.

I've uploaded the code and the data to https://github.com/tmmlaarhoven/lichess/tree/main/marathons now. Some comments:

  • The folder /sheets contains the score sheets of 10 players each, e.g., the string "022444444440023000210224400...". I fetched the top 500 score sheets to get the number of games played by each player, as this is not returned via /api/tournament/{id} or /api/tournament/{id}/results.
  • The folder /api was used as the starting point for the .json and .ndjson files. With the score sheets downloaded, these .json and .ndjson files were updated/overwritten to the files in the folders /1+0, /2+1, etc., so the folder /api is the "backup" folder of the original data from the API.
  • The folders /1+0, /2+1 etc. contain the tournament .json and (modified) .ndjson files with the score sheets and some additional data.
  • The folder /plots contains various figures, including the ones in the blog post.
  • The file logo.png is used to add the Lichess logo in the background of the plots.
  • The file script.py was used to update the .ndjson with these score sheets, to add more metadata, to calculate trophy rankings, and to generate plots.
  • The file README.md is a copy of the blog post.

Let me know if you have issues reproducing the same plots and ranking.

@Toadofsky said in #6: > I suppose I should inquire what source code produced these charts and graphs, in case people want to add machine learning-based predictions but don't want to rewrite Lichess API client code. I've uploaded the code and the data to https://github.com/tmmlaarhoven/lichess/tree/main/marathons now. Some comments: - The folder `/sheets` contains the score sheets of 10 players each, e.g., the string "022444444440023000210224400...". I fetched the top 500 score sheets to get the number of games played by each player, as this is not returned via `/api/tournament/{id}` or `/api/tournament/{id}/results`. - The folder `/api` was used as the starting point for the .json and .ndjson files. With the score sheets downloaded, these .json and .ndjson files were updated/overwritten to the files in the folders `/1+0`, `/2+1`, etc., so the folder `/api` is the "backup" folder of the original data from the API. - The folders `/1+0`, `/2+1` etc. contain the tournament .json and (modified) .ndjson files with the score sheets and some additional data. - The folder `/plots` contains various figures, including the ones in the blog post. - The file `logo.png` is used to add the Lichess logo in the background of the plots. - The file `script.py` was used to update the .ndjson with these score sheets, to add more metadata, to calculate trophy rankings, and to generate plots. - The file `README.md` is a copy of the blog post. Let me know if you have issues reproducing the same plots and ranking.

This is really interesting data! Thank you so much!

This is really interesting data! Thank you so much!