PDF for the related paper: https://arxiv.org/pdf/2410.02333 https://arxiv.org/pdf/2410.02333 You'll have to read about the metric used in the paper. My summary

PDF for the related paper: https://arxiv.org/pdf/2410.02333 https://arxiv.org/pdf/2410.02333 You'll have to read about the metric used in the paper. My summary - fragility score - depends on attacked pieces that can attack/defend other pieces. "Interestingly, the maximum fragility score often highlights the most critical and notable moves in a game." Highlights: - Over 20K games analyzed (over half from Karpov, Kasparov, and Spassky, about 2% from Stockfish vs other computer chess engines) - Max fragility score in game occurred around ply 32 (move 16), std dev is 8 plies (4 moves) - Typical pieces involved (looks like normal distribution) - pawns (~60%), knights (~20%), and then bishops (~10%). queens and rooks barely show up.