Time-Eval Panel: Clock Advantage and Engine Evaluation in 97,943 Classical Chess Games (Lichess Broadcast Relays, Jan–Jul 2026)

Player-game-checkpoint panel underlying "Down to the Wire: What ChessEngines Miss by Ignoring the Clock" (MIT Sloan Sports Analytics Conferenceresearch paper abstract). One row per player, per game, per 5-movecheckpoint (moves 10–60), for 97,943 classical games from Lichessbroadcast relays (January–July 2026, ≥60 starting minutes per player).Positions were re-evaluated with Stockfish 19 at depth 18. 1,353,364 rows. Columns: game_id unique game identifier move_n move number of this checkpoint (10, 15, ..., 60) color "white" or "black" -- whose row this is elo this player's rating opp_elo opponent's rating time_deficit_pct (own remaining time - opponent remaining time) / starting time, as of this player's own move eval Stockfish evaluation, this player's perspective (pawns) is_mate whether eval is a forced-mate score wdl_win_pct Stockfish win-probability estimate at this position wdl_draw_pct Stockfish draw-probability estimate wdl_loss_pct Stockfish loss-probability estimate score_loss_this_move expected-score loss on this specific move cum_score_loss cumulative expected-score loss up to this checkpoint win 1 if this player won the game, else 0 decisive 1 if the game ended decisively (not a draw), else 0 Source games: Lichess broadcast relays (database.lichess.org), CC0.Code to reproduce this panel from raw PGNs, and every figure/validationscript that reads it, is at:https://github.com/Thunderwood50/Chess-Time-Difference-SSAC License: CC0 1.0 (matches the underlying Lichess data license).