Real-Money Experiment · Polymarket

The Moneyline Theory

Our model's validated edge is the underdog +1.5 runline — but that cover typically costs −150 to −190 juice. The theory we're testing: the same late sharp money that predicts covers also predicts outright wins often enough that the plus-money moneyline is the better hold. So we're betting the ML with real money, flat $5 from a $20 bankroll, at real Polymarket CLOB prices — and tracking both answers side by side.

$21.90
Bankroll (started $20)
+$6.90
Net P&L · 7 bets
4-3
ML record

The scoreboard: what we bet vs. what the model validates

Moneyline (what we bet)

4-3 · +$6.90 on $35 staked

Real fills, real money. Wins pay plus-money (+104 to +122 so far), so one winner buys two losers.

+1.5 runline (the validated edge)

4-3 · −$3.03 hypothetical at the same $5

Same games, priced at Polymarket's real pre-pitch +1.5 ask (our automated snapshot feed). This is the control arm.

Every bet, on the record

DateBetEntryStakeML result+1.5 same game (control)
06-12Phillies ML
Phillies @ Brewers
32¢ (+212)$5LOST −$5.10No cover +104 → −$5.10
06-12Marlins ML
Marlins @ Pirates
46¢ (+117)$5WON +$5.96Covered -156 → +$3.25
06-11Rangers ML
Rangers @ Royals
49¢ (+104)$5WON +$5.20Covered -186 → +$2.69
06-10Orioles ML
Mariners @ Orioles
49¢ (+104)$5WON +$5.20Covered -163 → +$3.06
06-10Blue Jays ML
Phillies @ Blue Jays
45¢ (+122)$5LOST −$5.00No cover -133 → −$5.00
06-09Twins ML
Twins @ Tigers
46¢ (+117)$5LOST −$5.00No cover -178 → −$5.00
06-09Rays ML
Red Sox @ Rays
47¢ (+113)$5WON +$5.64Covered -163 → +$3.06

Entries are actual Polymarket CLOB fills (verifiable on-chain). The +1.5 control uses the last pre-first-pitch Polymarket ask captured by our 30-minute snapshot feed — no after-the-fact prices, no best-odds-screenshot inflation.

Off-script bets (full disclosure): 1 discretionary in-game bet outside the system, −$5.00 total — counted in the bankroll above, excluded from the experiment arms. The system did not pick these; a human did.

Honesty box

This is an experiment journal, not a track record — the sample is tiny and $5 stakes are intentionally toy-sized while the model's live sample matures. Plays come from the same signals published nightly in The Grind; the table lists settled bets only — this page is a results journal, not a tout sheet. Our EV gate currently grades most of these prices MARGINAL — the point estimate clears, the 95% confidence band doesn't. That's exactly why the stakes are small.

← Back to The Grind