How it works

Every few minutes we pull moneyline and totals odds from dozens of US sportsbooks, for both full-game and First-5-innings (F5) markets. We split the books into two camps:

• Sharp books — Pinnacle, BetOnline, LowVig, Matchbook, Betfair Exchange. These books take big action from professional bettors and move their lines based on respected money. Pinnacle is the anchor — we convert its price to a no-vig fair probability and use it as truth.
• Square books — DraftKings, FanDuel, BetMGM, Caesars. These shape lines around the public, which means their numbers often lag where pros think the price should be.

When sharp and square books disagree — or when any book's price drifts away from Pinnacle's fair line — that's where the edge lives.

Every signal runs on all four MLB markets:

• Full-Game Moneyline — pick the winner.
• First-5 Moneyline (F5) — pick the team leading after 5 innings. No bullpen risk, isolates the starting pitcher.
• Full-Game Totals — Over/Under on total runs.
• First-5 Totals (F5) — Over/Under on runs through 5 innings.

Every line — full-game ML, F5 ML, full-game total, and F5 total — is compared to where it opened, to consensus on the other side, and to Pinnacle's no-vig fair price. A pick fires when one of these primary signals trips and the Pinnacle Expected Value (EV) gate passes:

• Underdog value — opener was an underdog (+120 or longer) and the sharp price tightened by 10+ cents. Our most profitable signal historically.
• Pinnacle EV edge — the market price beats Pinnacle's de-vigged fair price by 3% or more. Pure math, no narrative needed.
• Reverse Price Move — sharp books moved the price one way while square books moved the other. Proxy for sharp money fading the public. Note: true RLM needs paid bet-percentage data we don't subscribe to — we infer it from cross-book price action.
• Big Consensus Move — the entire market repriced 10+ cents in the same direction. Lower-conviction, usually news-driven.
• Pitcher Edge — starter z-scores on ERA, WHIP, and K/9 vs the opposing arm. Boosts confidence (and occasionally triggers on its own for F5) when the matchup decisively favors one side.

Stack-only boosters: Steam (3+ sharp books moving in sync) and Sharp Lead (sharp books leading square by a wide margin) no longer fire as standalone picks — in v7 they only raise confidence on top of a primary signal.

The EV gate: Every candidate is checked against Pinnacle's no-vig fair price. Moneyline picks need EV ≥ −3%, totals picks need EV ≥ +1.5%. Anything worse is killed, regardless of how strong the signal looks.

Mid-favorite guardrail: Favorites priced between −120 and −200 historically bled units, so we now require a 2+ signal stack or EV ≥ +1% before posting one.

Each pick is tagged with its primary signal, a confidence score recalibrated around Pinnacle EV, an EV badge, and the reasoning. The recommended price is the best price available across our tracked books at the moment the signal fired.

• Picks are locked at first pitch. After the game starts, the recommendation, odds, signal, and reasoning cannot be changed. Only the result and grading fields update.
• Settled picks become public — every premium pick is visible to anyone after the game ends, so there's no way to hide losers.
• Profit is tracked in units at the recommended price, with each pick sized by ¼ Kelly off Pinnacle's fair line (½ Kelly optional). The unit number is what you'd have if you'd followed along at our price and sizing.
• Moneyline grading — full-game ML settles on the final winner; F5 ML settles on whoever leads after the top + bottom of the 5th, pulled from the MLB Stats API linescore.
• Totals grading — full-game O/U settles on total runs vs. the captured line; F5 totals settle on runs through the end of the 5th.
• CLV (closing-line value) compares the recommended ML price to the closing line. Beating the close consistently is the standard measure of long-term edge. Totals CLV isn't tracked yet (requires multi-variable line+price history).
• If a game is postponed, called before 5 innings (for F5), or otherwise voided, the pick counts as a push (0u).

Flat 1u betting — one unit on every pick regardless of edge — has lost money across our published record. The drag comes from no-edge tickets: betting them at full size bleeds. So we don't do it. Every recommended stake is Kelly-sized off Pinnacle's fair line.

• ¼ Kelly (default) — quarter-Kelly sized off Pinnacle EV. Bigger edge = bigger stake, capped at your max-stake setting in Settings. When the model says pass, no stake is recommended.
• ½ Kelly — half-Kelly for those who want bigger swings. Same ROI% as ¼ Kelly by definition; it simply doubles every stake.

On the same picks, Kelly returns roughly +12% ROI where flat 1u is negative. Flip ¼ / ½ Kelly with the toggle on the picks page; the whole site (board, track record, signal grids) re-sizes in lockstep. This is a tracker, not a sportsbook — set a $/unit in Settings to see dollars alongside units.