Putting our money where our mouth is.
Model projections vs. the sharp books and Kalshi, every game day — every play graded and published.
Spread Scalp runs a proprietary projection model on every FBS and NFL game and compares its numbers to the sharpest line in the market and to the prices on Kalshi. When the model and the market disagree by enough, the play goes on the board — before kickoff, with the price it was flagged at. Once a play is on the board it stays there at that number: if the market moves against it later in the week it is marked, not removed, so the record shows the cost.
Every play is then graded against the final score and against the closing line, and published here whether it won or lost. The scorecard, the picks ledger, and Bankroll v1 (a flat-stake simulation) are the whole point of the site: a track record you can check rather than a highlight reel.
In general terms, the model builds a projected score for each game from team performance data and turns that into a line and a cover probability. A second, independent check looks for Kalshi prices that are out of step with the sharp market. What we don't publish is the recipe — the inputs, weights, and thresholds are ours; the results are public.
The model is not finished and never will be. It is reviewed every week against its own results and the closing lines, and it changes when the numbers say it should: one adjustment at a time, each one dated and kept. Early in a season it leans on preseason ratings; as games are played, real results take over. Expect the plays to get more selective as it learns.
Plays marked paper are in calibration — logged and graded but not bet. Tier A plays are the ones we'd put money on; Tier B leans are tracked for measurement only. Nothing here is financial advice.