Each strategy has a methodology page (the math) and a live picks feed (today's signals). Tracked performance for each one is in Track Record.
The foundation of every strategy on the site. After de-vigging the market consensus across every book, +EV picks are bets where one book is offering odds longer than the consensus fair price implies.
A 0–100 composite score over four signals: odds discrepancy across books, underdog value, market divergence, and EV strength. Picks scoring above tier thresholds (55 / 42 / 30) are flagged Elite / Strong / Moderate.
The detector looks at every market with at least 4 books posting, calculates the median American price, and flags any book offering a noticeably better number for the bettor than the consensus.
When public bet percentage and sharp money diverge, books move lines against the public to limit exposure to professional bettors. The Public Fade Index combines bet count, dollar handle, and line movement direction into one score.
When a team loses by a sport-specific blowout margin (NBA 20+, NFL 21+, NCAAF 24+, NCAAB 25+, MLB 7+, NHL 4+) and is priced as an underdog in their next game, the fade triggers. Behavioral-finance edge — the public over-adjusts to dramatic losses.
When parlay legs are positively correlated (QB + WR yardage, weather games, division-rivalry under stacks), the true co-occurrence rate is higher than the independent product the book uses to set the price.
Anytime HR props ranked by modeled edge. Combines season Statcast (barrel rate, hard-hit, xISO), the day's pitcher matchup with HR/FB regressed to a league prior, humid-air carry physics + per-handedness wind decomposition, lineup-slot PA PMF, and a Shin-style devigged market consensus. The v3 residual layer (logistic regression on 127K settled HR-prop closing lines from 2025+2026) corrects book-specific shading and park-specific HR residuals before the consensus median. Picks are capped at +800 American — the longshot tail is where uncapped lanes consistently lose ~15% ROI; the cap gates that out.
v2 calibrated chain without the v3 residual layer applied. Same Statcast quality + park + pitcher inputs and Shin devig as v3, but the no-vig consensus is taken straight from the per-book devig instead of being run through the v3 residual model first. Cron forks at the calibration step and tracks v2 + v3 in parallel; the long-run delta in ROI / hit-rate / CLV reads out the v3 residual's contribution directly.
Original HR Hunter model, unchanged from launch. Multiplicative chain over batter Statcast quality, three-year park HR factor, pitcher HR/9 regressed to a league prior, weather (temp, wind direction, humidity), and platoon split. Picks are capped at +800 American so the longshot tail (where the multiplicative chain compounds optimism worst) doesn't blur the read. Volume is high; tracked over ~10 days at this writing.
Points, rebounds, assists, threes, and PRA props tilted by the assigned referee crew's multi-season foul / FTA / pace multipliers (55% chief, 22.5% each remaining ref). Pair-wise devig across every US book; graceful fallback to neutral math when crew assignments are still TBA.
SOG lines tilted by team linemate quality, team shot-for rate, and opponent shot-suppression. Player→team mapping pulled from live NHL rosters; team profiles seeded from 5v5 xGF%, SF/60, SA/60. Tilt weighting: 50% linemate, 25% own SOG-for, 25% opponent SA.
Total-saves props tilted by starter rest days, schedule density, and opponent shot-for rate. Back-to-back starters get an over-tilt; rested starters facing high-volume opponents get the strongest signal. Tilt weighting: 55% opp SOG, 30% rest, 15% trip density.