37 days of juice

The headline

Since 2026-06-21 we've captured 4,432 picks and graded 3,788 of them. The ledger reads 1944-1765-79, +194.41u at flat 1u stakes, +5.24% ROI, with 83.5% of CLV-measurable picks beating the close.

Five weeks is a small sample and we're going to keep saying so. What five weeks is good for is the shape of the thing: which markets the edges live in, and how much price alone is worth.

Split by market type

• Props: 3314 graded, 1620-1618-76, +21.66u, +0.67% ROI, +0.03% avg CLV
• Moneyline: 216 graded, 116-97-3, +12.62u, +5.92% ROI, +2.93% avg CLV
• Spread: 187 graded, 173-14-0, +160.12u, +85.63% ROI, +3.27% avg CLV
• Total: 71 graded, 35-36-0, +0.01u, +0.01% ROI, +2.65% avg CLV

The spread between Spread (+85.63%) and Total (+0.01%) is wide, and at these sample sizes most of that spread is noise rather than signal. The useful read isn't "bet more spread" — it's that our edge estimates behave differently in thin markets than in liquid ones, which is exactly what you'd expect when fewer books quote a number.

What −110/−110 actually costs

Take the standard American two-way price. −110 implies 52.38% to win. Two sides at −110 sum to 104.76%, so the book is holding about 4.55% of the market. In practice you need to win 52.38% of your −110 bets just to break even; at a true 50/50 coin flip you lose about 4.5 cents on every dollar cycled.

That's the number every "+EV" claim has to clear first. A 2% edge sounds thin until you notice it's being measured against a house take more than twice that size — the whole job is finding the market where the price is wrong by more than the vig, and then actually getting that price.

Line shopping is the cheapest version of this. The same side at −105 instead of −115 moves your break-even from 53.49% to 51.22%. You didn't get smarter; you just stopped donating two points of edge per bet.

Where the price advantage shows up

• CLV-measurable picks: 734
• Beat the close: 83.5%
• Average CLV: +3.20%
• Median CLV: +3.21%

Beating the close doesn't pay rent this month, but it's the part of the record that stabilises first. If the number you took keeps ending up better than the number the market settles at, the edge was real — the results just haven't caught up yet.

The honest caveats

644 captured picks are still ungraded, 576 of them past the point where a normal game would have settled — mostly leagues with no reliable score feed. Anything you read above is graded plays only, and we'd rather show you the hole than quietly close it.