← Back to the picks

How to read the picks

Every number we publish, explained. Two-minute video version on our YouTube channel.

The pick card (Today's Optimal Plays)

Element	What it means
Team name	The team the model backs in this game.
RUN LINE +1.5	The pick with a 1.5-run cushion: it cashes if the team wins, or loses by 1. This is the model's product — every pick it has ever made is a run-line underdog. A margin bet, not a totals (over/under) bet.
MONEYLINE +103	The same team to simply win the game, at those odds. Tracked for transparency; the run line is where the edge lives.
% EV	Expected value: the model's estimated long-run profit per dollar on that market at current odds. The outlined market is the better-value side of the two.
WIN PROB	The model's probability that the run line bet cashes.
R edge	Runs of cushion beyond the model's expectation. Example: line +1.5 while the model expects the team to lose by only 0.5 → 2.0 runs of edge.
STRONG / LOCK	Confidence tier from win probability. Higher tier = higher model conviction, not a guarantee — the season tier records are on the site.
10% bet	Suggested stake as a percent of a bankroll (Kelly-derived sizing). Illustrative, not financial advice.

Pick History (how results are graded)

Column	What it means
PICK + LINE	The team taken, with its run line. +1.5: cashes on a win or a 1-run loss. +1: a 1-run loss is a push (stake returned).
EDGE	The R-edge the card showed when the pick was made.
SCORE	Final score, away-home.
RL	Run-line result: W / L / P.
ML	The same pick graded straight-up (did the team win the game).

The record strips

SEASON TO DATE is the full public record since the model went live — wins, losses, pushes, and profit for both markets, at $100/pick flat stakes. LAST 30 DAYS is a rolling window of the same picks: totals can shrink day to day as old picks age out — that's the window moving, not history being edited.

A note on the profit numbers. Moneyline profit is graded at the real odds captured when each pick was made. Run-line profit is currently graded at a flat standard price, because the actual run-line prices weren't recorded — real prices on +1.5 dogs are often steeper (-140 to -200), so treat the run-line dollar figure as illustrative, not as what a bettor would have banked. The W-L-P record itself is exact either way. We're working on capturing real run-line prices going forward.

Every pick is captured before first pitch and graded automatically after the final. Wins and losses both stay up. That's the deal.

FAQ

Why does the model only ever take underdogs?

It compares its predicted margin to the run line and bets when the cushion beats the prediction by enough. In baseball, expected margins are tiny — even huge mismatches project to winning by barely more than a run — so laying −1.5 almost never clears the bar, while a close dog getting +1.5 often does. The dog-only pattern is the math, not a rule.

Why is the moneyline record worse?

Underdogs lose outright more often than they lose by 2+. The picks are chosen for the run line; the moneyline column is the same picks graded a harder way, shown for honesty.

Why so few picks per day?

The model only fires when the edge clears a threshold — typically 2–3 games a day out of a full slate. No edge, no pick.