Player-prop value finder (model hit-rate vs market) with backtest

Player-prop value finder: flags props where a player's empirical over/under hit rate (model probability) beats the de-vigged over/under juice (market probability). Edge = model minus market. Two windows: upcoming (default) flags props on games in the next 24h using a trailing lookback-game model; all runs a point-in-time historical backtest and populates a backtest block (record, ROI, average edge). The crude empirical model flags generously, so treat the backtest as a calibration check, not an edge guarantee. Response object keys: model, window, min_edge_pct, lookback, min_games, evaluated, value_found, backtest (null for upcoming), props[]. 5-minute Redis cache. Solo plan or higher: prop value signals start at Solo; Free receives a 403 naming the tier and linking the docs and pricing pages. (The Edge in "Edge = model minus market" above is the statistical quantity, not the plan.)