Expected value betting: how to find an edge and judge it honestly

What is expected value in sports betting?

It is the long-run average profit of a bet, per dollar staked. A +EV bet pays more, on average, than its odds say it should, and a -EV bet pays less.

A bet at +150 pays decimal 2.50. If its true win chance is 42%, then 0.42 x 2.50 - 1 = 5.0%, or $5.00 per $100 on average. At 38% the same price is -5.0%. Break-even depends on the price: at -110 you need to win 52.38% of the time, because a -110 market has 4.76% of vig built in. EV does not say you will win this bet. It says what the bet is worth on average.

How do you calculate EV?

Convert the price to decimal, estimate the true win chance, and compute p x d - 1. Everything difficult is in the second step.

1. Convert the offered price to decimal odds (the odds converter does this).
2. Estimate the fair win chance from prices with the vig removed (next section).
3. Multiply: p x d - 1. A positive result is +EV; multiply by the stake for dollars.
4. Shade the chance down for uncertainty and compute again, so you see a conservative EV.

A tie or push needs one more term: EV = p x (d - 1) - q, where q is the chance of losing and the push returns your stake. SlateProof's engine reports both the EV at its best estimate and a conservative EV at a lower, shaded probability, and an edge is a model claim, never a certainty.

Where does the fair probability come from?

From other venues' prices, with each one's margin removed. A fair probability is not what a sportsbook posts: it is that price with the vig taken out, blended across reference books that have earned trust.

Removing the vig has several methods. For a two-way market priced at -150 and +130 the implied chances add up to 103.48%, a vig of 3.48%. The favorite's fair chance depends on how the margin is spread:

SlateProof uses the power method by default and offers all four in the no-vig calculator. Several rules keep the estimate honest: the book you are testing never sets its own fair value; books that copy each other share one vote; daily-fantasy apps never count; a one-sided or milestone market is never de-vigged because it needs the complete outcome set; and a price no reference quotes directly can be extrapolated from the same player's other lines only with a fitted distribution, labeled as such and with wider uncertainty. A price with no qualifying reference is a coverage note, not an opportunity.

How much EV is enough?

There is no universal threshold. The right minimum depends on how certain the fair price is, how long the market has been quoted and how much margin hides in it. A large apparent edge is more often a stale or mismatched price than a gift.

Uncertainty is the reason to shade. Take a +100 bet (decimal 2.0) with a best-estimate chance of 55%: the EV is 10.0%. If the estimate has a logit-scale uncertainty of 0.10 and you take one standard deviation off, the chance drops to 52.51% and the EV to 5.0%. SlateProof's screening policy sets different minimums for game lines and props and requires the shaded EV to hold up, and for promos it ignores legs whose apparent EV is above +10%, because beyond that an error is likelier than a real edge.

Does closing line value prove an edge?

No. Closing line value (CLV) is the difference between the price you took and the price at game time, usually the sharpest estimate of the true chance. Beating it often is a good sign, but it is not profit.

SlateProof's own audit shows why. In a pre-registered test of 214,000 MLB prop bets at DraftKings and FanDuel, closing value measured against the reference books was positive while the return on those bets was negative, and no fair-value, timing or modeling variant beat the books' own prices. So SlateProof measures CLV and realized return together, with intervals, and promotes a strategy only when both agree. Its market-edge lane is on paper only for that reason. A close is also not always comparable: a different threshold or a missing close is never substituted.

How many bets before the result means anything?

Many more than most bettors place. A bet's result is dominated by luck, so a small edge needs a large sample to show.

Take a bet at +110 (decimal 2.10) with a true 5% edge. The profit on a one-unit bet has a standard deviation of about 1.05, so the average result is two standard errors above zero only after about 1,764 bets (n = (2 x sd / edge)^2). Staking size matters too. Full Kelly, (p x d - 1) / (d - 1), for a 55% chance at decimal 2.0 is 10% of your bankroll, which is aggressive when p is uncertain. SlateProof's paper policy stakes a quarter of Kelly at the conservative probability and caps it at 2% of the bankroll (a quarter Kelly here is 2.5%). Singles are better than ladders: the margin compounds with every leg.

Will sportsbooks limit +EV bettors?

Often, yes. Sportsbooks restrict accounts that win consistently, which can mean lower maximum stakes or reduced promo eligibility, and they often cap prop-market stakes to begin with.

Exchanges have less reason to restrict a winner because they match traders with each other, but their limit is depth and their prices carry a fee. Do not assume any venue will take unlimited stakes at the price you saw.

Is +EV betting legal?

Yes, wherever the bets themselves are legal. +EV betting is ordinary betting with a price check, so the only legal question is whether each venue is legal where you are when you bet.

Check the state pages for the venues available in your state, with the date we checked them. The venue's own app has the final say, and this is not legal advice. A bet with positive EV can still lose, and you should bet only what you can afford to lose.

+EV, arbitrage or promo hedging: which fits?

The honest answer from SlateProof's tests is that promotions carry the measurable value, arbitrage is small and fast, and market edges are unproven.

Keep going

Take the vig out of a market with the no-vig calculator, convert formats with the odds converter, and read about locked profits in the arbitrage guide and the sportsbook promos guide. The methodology page documents how fair value is built and tested.