Does this dataset contain all/most of the evaluations embedded in the movetext of the standard-chess-games dataset?
Hey there! Just a question out of curiosity: does this dataset contain all (or most) of the evaluations embedded in the movetext for the games in this dataset? https://huggingface.co/datasets/Lichess/standard-chess-games
And what, if any, relation does it have with the fishnet-evals dataset? https://huggingface.co/datasets/Lichess/fishnet-evals
Trying to figure out if the evaluations in this dataset could be supplemented by evaluations from the movetext in the games dataset or if I'd just be wasting my time because they're coming from the same source (Requested cloud-based analysis of games). Also trying to use evaluations from either this database or the fishnet-evals dataset to prevent the need for evaluating the positions manually in a backwards induction based opening repertoire building algorithm I've been working on. So far I've been using this dataset, but it seems like fishnet-evals has way more positions evaluations in it (granted, without listed multi-PVs) so I'm wondering if I should switch to that instead.
Nevermind! Upon deeper investigation I think I've answered my own question: it seems like fishnet-evals is what matches the movetext in standard-chess-games, NOT chess-position-evaluations like I originally thought. That also makes sense given that the eval graphs from the Analyze Game requests at the end of games are essentially just position-level evaluations rather than multi-PV evaluations and tracks with the "move" being what the human played next for the schema in the fishnet-evals schema. Hopefully this discussion helps someone else who is curious!