The one-paragraph version. We take every price a book shows, remove the book's cut so the numbers add to 100%, build our own probability for the same event from a simulation of the whole game, compare the two, and then — the part almost nobody does — check whether your legs are secretly the same bet before we multiply them together. What comes out is one number: the cents on the dollar you're paying above or below fair.
A slip we grade positive can lose. A slip we grade negative can win, and often will. Nothing on this page predicts a game. It prices one. If you want the version of this that promises winners, there are about four hundred accounts on X ready to sell it to you.
STEP 01Strip the vig
A betting line is not a probability. It's a probability plus the house's cut, and the cut is baked in so smoothly that most people never subtract it. Take a standard two-way prop:
Those two outcomes cover everything that can happen, so honest probabilities have to sum to 100%. The extra 5.30 points is overround. Removing it is called devigging, and how you remove it matters more than people expect.
Three ways to do it, and why we don't use the easy one
The obvious approach is to scale both numbers down proportionally until they sum to 1. That's the multiplicative method, and it's what most free calculators use:
On a near-coinflip the three answers agree to within half a point and it doesn't matter. On a longshot they diverge badly, and that's where parlays live. Bettors systematically overpay for big prices — the favorite-longshot bias — so the book's cut isn't spread evenly across both sides. It's loaded onto the longshot.
| Method | True prob | Fair price | Book's edge |
|---|---|---|---|
| Multiplicative | 26.1% | +283 | 6.1% |
| Shin | 25.2% | +297 | 7.9% |
| Power | 24.4% | +310 | 9.6% |
Three and a half points of probability sounds small. On a four-leg parlay, a three-point error on one leg moves the whole ticket's fair price by more than 200 cents. We use the power method on any leg priced longer than +150, multiplicative below it, and we show you which one we used.
One book's line is one book's opinion. The consensus of a sharp market — weighted toward the books that move first and take the biggest bets — is a much better starting probability than the number sitting in your app. Half of the "edges" you'll find in a slip aren't edges at all, they're your book being slow.
STEP 02Build a distribution, not a projection
Here's the mistake that sinks most homemade prop models. Someone projects Bijan Robinson for 58 rushing yards, sees a line of 60+, and concludes the over is a loser. That reasoning is wrong even when the projection is right.
"60+ rushing yards" is not a question about the average. It's a question about the tail: how often does the top half of this player's outcome distribution clear a number. Two backs projected for the same 58 yards can have wildly different answers.
How the number actually gets built
We don't model the prop. We model the game and then read the prop off it. For each matchup we run 50,000 simulations, and each one plays out roughly like this:
- Possessions. Draw a plausible number of drives from the pace distribution of these two teams, adjusted for the projected score margin — trailing teams run more plays.
- Play mix. On each drive, sample run or pass from the offense's tendency, conditioned on down, distance, and current game state.
- Usage. Assign the touch to a player from their route-participation and carry-share distribution, not their season average — shares move with personnel.
- Outcome. Draw yards from that player's own yardage distribution against this defense's allowed distribution, then check for the goal-line, red-zone, and turnover branches.
- Tally. At the end of the sim, record everything: every player's yards, receptions, touchdowns, and the final score.
The payoff is that every prop, every total, and every side comes out of the same 50,000 games. So when we ask "how often did Drake London get 5 catches and the game went over 44.5," we don't have to estimate it. We just count the sims where both happened. That count is the entire reason the next step works.
STEP 03Join the legs
Parlay math taught to a beginner says multiply the probabilities. That is correct if — and only if — the legs are independent. Legs from the same game are never independent. Legs from the same drive are barely even separate events.
Notice which way it went. Correlation made this ticket more likely, not less, because three of the four legs want the same thing: Atlanta throwing the ball a lot in a game that stays competitive. That's good news for you and it's why same-game parlays exist at all. It is also exactly why books shade the price. The full treatment of this is its own page — Same bet twice — because it's where most of the money is won and lost.
STEP 04Price the ticket
Now it's arithmetic. Convert the offered American odds on each leg into decimal, multiply them for the payout multiple, and compare the payout to the probability we counted.
Read that last line carefully, because it's the only number on the site that matters. It does not say the ticket loses. It says that if you played this exact ticket at this exact price a thousand times, you'd end up with about 98 cents for every dollar you put in. Two cents on the dollar is a fairly gentle beating by parlay standards — the median four-leg slip we examine holds 11.4%. But it's still the wrong side of zero, and it's fixable.
STEP 05Choose the cut
The intuitive move is to drop the leg with the worst standalone edge. That intuition is wrong often enough to matter, because a leg in a parlay isn't judged on its own probability. It's judged on its probability given that everything else on the ticket already hit.
A leg belongs on the ticket if P(leg | all other legs hit) × its decimal price > 1.
Everything below is that one comparison, run four times.
| Leg | Solo | Given the rest hit | Price | Marginal | |
|---|---|---|---|---|---|
| Bijan 60+ rush | 58.5% | 61.2% | 1.8696 | +14.4% | KEEP |
| ATL/CAR over 44.5 | 54.5% | 57.8% | 1.9091 | +10.4% | KEEP |
| Pitts anytime TD | 23.5% | 26.7% | 3.6000 | −3.9% | TOLERATE |
| London 5+ rec | 50.5% | 52.9% | 1.7692 | −6.4% | CUT |
Look at what correlation did for Drake London. His standalone probability is a coin flip at 50.5%, but conditional on Atlanta going over the total and Kyle Pitts finding the end zone, he gets to 52.9% — the ticket's other legs have already told us this was a passing game. Correlation gave him 2.4 free points. He still isn't worth −130. That's the cut.
The cut ladder
Once you accept the rule, an obvious question follows: if cutting one bad leg helps, why not cut two? Usually it does help. We show you the whole ladder and let you pick where to stop, because a bettor who wants a lottery ticket on Sunday is making a legitimate choice and we're not going to pretend otherwise.
| Version | Price | Hits | Pays | EV |
|---|---|---|---|---|
| All four legs | +2173 | 4.31% | $454.68 | −2.0% |
| Cut London | +1185 | 8.15% | $257.00 | +4.7% |
| Cut London & Pitts | +257 | 32.5% | $71.40 | +16.0% |
| Bijan alone | -115 | 58.5% | $37.39 | +9.4% |
The two-leg version is the best bet on the board and it pays $71 instead of $455. We know which one you're going to play. We'd just rather you play it knowing what it costs, and the honest framing is that the difference between the four-leg and the three-leg is $1.34 of expected value on a $20 ticket — real, repeatable, and free to take.
STEP 06What we can't do
Every model page on the internet ends with a victory lap. Here's the other thing instead.
- We're behind on news. Our board rebuilds every four minutes. A beat writer's tweet about a limited practice will move a market before we've re-simulated anything. On inactive-report Sundays, trust the market over us for the first fifteen minutes.
- Thin markets break the devig. A Tuesday MACtion receiving prop offered at two books with a 14% overround has no reliable fair price. We flag those THIN and we'd rather you skipped them.
- Our correlations are estimates. They come from simulation plus seven seasons of play-by-play. A pair we've seen a thousand times is solid. A pair involving a rookie tight end in a new scheme is a guess with error bars, and the error bars don't show up in a clean-looking number like +0.31.
- We don't know your book. Limits, boosts, SGP restrictions and how badly you'll get squared away for winning are all outside the model.
- Being right is not the same as winning. A +4.7% edge means that over a very long run you keep about five cents of every dollar. The distribution of outcomes around that is enormous. You can be correct all season and finish down.
Overround — how much more than 100% a market's implied probabilities sum to. Devig — removing it. Power method — a devig that raises each probability to a shared exponent, which takes more of the cut off the longshot. Marginal EV — what one leg adds or subtracts from a whole ticket. CLV — closing line value, whether your price beat the market's final one.