Just Cover Lab
Running total
How each bet type is defined
Every bet starts the same way: we take Pinnacle's odds (the sharpest market in the world), strip out their built-in cut to get a fair price, and compare it to the apps people actually bet on. The bet types differ in which edges they take and how much they bet.
Overall
Every model bet at standard size, all in one paper bankroll that started at $5,000: straight bets, player props (counted at $50 per unit, which is 1% of the bankroll) and the ratings gap bets. This is the honest "everything" number.
Straight bets
Spreads, totals and moneylines where an app's price is at least 3% better than the fair price (4% if the bet is placed 8+ hours before the game). Sized with a quarter of the Kelly formula and never more than 3% of the bankroll on one bet. Part of Overall.
Props
Player props (points, rebounds, assists, threes, receiving yards, catches and so on) at real apps, with the same 3%+ edge rule against Pinnacle's fair price. NFL props are Unders only, because Overs lost money in testing. Graded at a flat 1 unit each so a +150 and a −150 count the same.
Aggressive
The same qualifying bets sized bigger, on a separate $5,000 paper bankroll: half Kelly (vs a quarter), up to 5% on one bet, 4% on one game and 30% of the bankroll open at once. It also takes props (quarter Kelly, max 1.5%) and teasers/parlays built only from legs with 3%+ edge. Bigger swings both ways: this is the "what if we pressed" test.
Best Bets
The highest-quality subset, on its own $5,000 paper bankroll: edge between 2% and 10% (huge edges are often bad data), odds between −250 and +250, no fresh injury news, weather flags or unknown app limits, and at most 5 per day. Same sizing rules as straight bets.
Ratings gap bets
Side tests of our own power ratings: when our line differs from the market by a set number of points (for example 5+ in college basketball), we log a paper bet. Shown in Overall and in "Everything else we've tried."
Everything else we've tried
Side experiments, shown win or lose. Losing ideas stay on the board.
| Bet type | Record | Open | Staked | Profit | ROI |
|---|
Every bet
Every model bet, newest first, at the price and app we'd have bet. CLV = how much better our price was than the closing line.
| Date | Pick | App | Odds | EV | Stake | Result | Profit |
|---|
Props by sport and side
Flat 1 unit each. NFL props are Unders only since Nov 18, because Overs lost.
| Group | Record | Units | ROI |
|---|
Every prop
| Date | Player | Prop | App | Odds | EV | Final | Result | Units |
|---|
Lab Notes
Ideas we've tested against real data, whatever the answer was. Got a theory? Pitch it to @JustCoverLab on X (x.com/JustCoverLab) and we'll test it.
NBA back-to-backs and travel
4,138 games. Teams on a back-to-back are about 2 points worse, but the market already prices that in. Betting against tired teams didn't beat the line.
Only bet 3%+ edges
Bets under 3% EV lost money (−2.4%). Bets at 3% or more returned +15.1%. The model now skips anything under 3%, and needs 4% for bets placed 8+ hours early.
Bet close to game time
Bets placed under 3 hours before start beat the closing line 87% of the time, vs 46% for bets 8–24 hours out. Late news moves lines toward us when we wait.
Props when a teammate is newly out
NBA player props right after a teammate is ruled out hit 62% (fair odds said 50%), +29.6% ROI. Only 90 props from about 15 games, so it's being tracked, not trusted yet.
NFL player prop Overs
NFL Overs returned −27% while Unders returned +25%. The model now bets NFL props Unders only.
Free tools
The same ratings and calculators the model uses, free for everyone, with data refreshed daily.
How the model works
- Start from the sharpest price. Pinnacle's odds are the best estimate of a game's true chances. We remove their built-in cut to get a fair price.
- Shop every app. We compare that fair price to the odds at the apps regular people use.
- Only bet real edges. A bet qualifies when an app's price is at least 3% better than fair.
- Size bets small. Stakes are a fraction of the Kelly formula, capped at 4% of the bankroll per game, so no single game can sink us.
- Grade the price, not just the result. The real scorecard is closing-line value: did we get a better number than where the line closed? That says whether we're right long before win-loss does.
- Test everything in public. New ideas get backtested, and the results get posted in Lab Notes, even when the idea fails.
House rules
- Every pick is timestamped before the game starts. Losses stay on the board.
- These are paper bets on a $5,000 test bankroll, logged at real prices. No real money is shown here.
- Fun bets come from a set monthly budget. No 50-to-1 parlays, no chasing.
- Not betting advice. Past results don't guarantee anything.