—
Match source
On Flashscore, open the match's Point by point tab and copy every set. Include the summary and stats sections too, and names, tournament, surface and net-point rates are picked up automatically, and the result is cross-checked against Flashscore's stats.
By default the win probability is market-anchored, the same way the Shanghai generator's run_model() works: de-vig the two-way odds, then shift by CONF_MULT × adj on the logit scale. The players' serve/return stats only set the serve environment (how many holds, tiebreaks and total games), and the point probabilities are solved so the simulation hits that exact win probability. Switch to Stats only to let the serve/return numbers decide the winner too.
With Custom inputs the stat fields start as placeholders; the presets load real 52-week stats. Set adj from your Form-vs-Model review to move the anchor off the market.
Model pricing
Total games
Momentum Cumulative point difference · click to jump
Game flow
Stats so far
In a real-match replay, the score, serve order, point winners and break/set/match points come from the data. Rally length, ball paths, net approaches and how each point ends (winner or error) are not in Flashscore's data, so they are illustrative; net-approach frequency follows Flashscore's net points when the stats are pasted. Simulated matches are generated point by point from the model's serve probabilities. Shortcuts: Space play, ← → point, Shift+arrow game.
