Methodology

How the model works, why we publish losers, and where the edge actually is.

What a "pick" is

For each game the model computes a recent-form edge — a player's or pitcher's last-N hit frequency at a line, versus the book's vig-included implied probability. We bucket results into tiers and publish them. A hit rate is a descriptive past frequency, not a calibrated probability.

Why we publish NO-BETs and losers

Most nights, the disciplined call is to pass — the top "edges" are often heavy-chalk hot streaks that revert, or stale data. We publish those NO-BET nights and count them, and every graded loss stays on the record. The transparency is the product.

Backtest-gated markets

A market is only promoted to "production" after it beats the vig in a real backtest with a meaningful sample. Today the one validated lane is MLB pitcher Outs Over (backtest n=688: +21% ROI blind, +17.5% on the filtered tier). Other markets — MLB hit/run unders, WNBA, tennis, soccer — are shown as research/exploratory and are not claimed as edges until they clear.

Freshness & honesty gates

• Stale player logs or no confirmed lineup → the pick is held, not shown as live.
• Games already started are excluded.
• Heavy-chalk and known −ROI line buckets are flagged.