Live NFL and college stats, factor-based projections that show their work, an engine built to disagree with the crowd, and a simulator that models how football outcomes actually move together. Every number on this site can explain where it came from.
Weekly fantasy scoring has a large irreducible random component. Any product that hides that from you is selling confidence. These tools are built to find a small, persistent edge and to let you verify it.
Fifteen weak signals, ridge-fitted weekly
Standardized within position, weights learned from this season's own results, refit every week, predicting one week ahead. It reports its holdout error instead of hiding it.
Prices being different from the crowd
Models projected ownership, then divides ceiling equity by it. Leverage, fades, traps, and the expected-duplicate math that tells you whether a lineup can win a large field at all.
25,000 correlated slates
Real correlation structure — QB-WR stacks at +0.46, split backfields at −0.38 — factored with Cholesky and drawn jointly. Turn correlation off and watch your ceiling collapse.
Walk-forward, against a hard baseline
Weights refit on strictly earlier weeks, scored on weeks the model never saw, compared against a three-game average that is much harder to beat than most sites admit.
19 feeds from the sites that actually move fantasy markets — analysis desks separated from transaction wires, headlines tagged with the players they mention, dead sources flagged instead of silently dropped.
Full player board, the whole news wire, the top of the edge board every week, and a real simulation. Paid tiers exist for the complete model surface and the tools to audit it — not for a better model.