Question Regarding Evaluation Scores
Hi,
Thank you for uploading and sharing this dataset — it’s potentially very useful for my research.
I had a quick question regarding the evaluation scores included in the data (in column "movetext"):
Could you clarify how these scores were computed? For instance, were they generated using a specific chess engine (e.g., Stockfish 16), and if so, under what settings (depth, skill level, etc.)?
Or were they created via another method?
Thanks in advance!
Best regards,
Yifei
Hi Yifei, glad you find the data useful!
The games are evaluated using fishnet (distributed Stockfish) and the exact settings and version will depend on when the broadcast was analyzed. Current settings are 5M nodes per position and a Stockfish version of at least 16 based on the current min version of fishnet.
Thanks a lot!