Original studies on betting market inefficiencies

The MLB paper proved CLV is ~2.5× more predictive of cohort ROI than win rate. Basketball has fundamentally different market structure: tighter spreads, faster line movement, more public-money-driven steam. We test whether the empirical relationship holds in a different sport — and quantify how much of the "CLV is the only metric that matters" claim is universal vs MLB-specific.

NRFI/YRFI is the fastest-growing MLB bet type. The market mostly prices it off season ERA, but first-inning ERA is often dramatically different — pitchers with shaky early innings vs steady starters, lefty/righty splits, etc. The gap is the edge.

Sharps have long suspected books over-juice unders on low-run games (recreational bettors love the under on a pitchers’ duel). We measure the actual hold, the actual ROI, and identify which dates and pitcher matchups have been the most exploitable.

In the playoffs, coaches play their starters 4-6 minutes more per game on average. Prop lines lag this by 24-48 hours. There is a structurally short window every series where points/rebounds/assists overs are mispriced — we map exactly when and where.

Most bettors use FanGraphs/BR park factors. Both compress real effects — Coors at +112 should be closer to +124 once you blend altitude, wind, humidity, and handedness. We give the corrected factors and the historical totals ROI of using them.

The market knows about B2Bs but mostly prices them with a flat 1-2 point adjustment. Our backtest carries actual travel miles and rest days per team per game — we measure the real ATS hit by fatigue cohort, including the under-priced "tired road favorite" trap and the over-priced "rested home dog" pop.

Books move totals down ~3 points and spreads tighter once playoffs start, but the underlying scoring distribution shifts more than the line. The result: a measurable historical edge on specific series-stage / road-team combinations that we identify with 17 years of receipts.

Q1 prop and total markets are growing fast on PrizePicks / Underdog / DraftKings. Books mostly price them as ¼ of the full-game total. The actual Q1 distribution is wider, has a different skew on rested-vs-tired teams, and produces a structural edge we quantify game-by-game.

Conventional wisdom says big spreads "get backed down" because the public takes the dog. The data says the opposite for a specific cohort: rested home favorites of 10+ vs road dogs on the second night of a B2B blow the spread out by ~3.4 points on average. We show every cohort and the historical ROI of fading the dog.

Most "model picks" services give one number. Ours decomposes into three independent signals — and the data shows ROI is roughly flat when 1 component agrees, climbs sharply at 2, and inflects at 3. This is the empirical case for picky bet selection over volume — measured, not asserted.