The operational definition of value
A bet has positive expected value (+EV) if your estimate of the true probability of the bet hitting is higher than the implied probability of the offered price.
Worked example:
- You estimate Team A has 55% probability of winning
- The sportsbook offers Team A at -110 (52.4% implied probability)
- Your estimate is 2.6 percentage points higher than implied
- This bet is +EV by 2.6 / 100 × payout multiplier
The expected value formula: EV = (P × payout) - ((1 - P) × stake) where P is your true probability estimate.
Use our no-vig calculator to convert American odds to fair-probability. Use our Kelly Criterion calculator to size bets based on your edge.
Technique 1: Line shopping (the easiest)
The same bet at multiple sportsbooks usually has slightly different prices. Take the best price.
Example: NFL Sunday late-window. Five operators show the same game with these prices:
- DraftKings: -110
- FanDuel: -110
- BetMGM: -108
- Caesars: -112
- bet365: -105
Bet365's -105 is 5 cents of free value vs the consensus. Across 500 bets where you consistently take the best line, you capture an average 1-3 cents of additional CLV per bet — that's the easiest, most reliable edge available to a US bettor.
Multi-account requirement. To line-shop you need accounts at 4-6 operators. Most US states allow this; check your state guide for licensed operators.
Technique 2: Model output vs market price
If you have a quantitative model that produces probability estimates for games, compare model output to market prices. Bets where your model probability exceeds the market's implied probability by more than the operator's vig margin (~5%) are +EV.
Building a credible model is hard. Common approaches:
- Regression on team stats: Predict score margin from rest, travel, recent form, opponent strength. Produces a fair-spread estimate.
- Pythagorean expectation: Predict win % from points-for / points-against ratio. Works for NFL, NBA, MLB.
- Player-level Bayesian models: For props (prediction market data + player rest + opponent defensive rating + recent form).
The catch: your model has to be better than the consensus market model, which is built by quants at the operator with access to far more data than you. The realistic strategy: build a model that targets specific niches (e.g. WNBA player props, second-half NCAA totals) where the market is less sharp.
Technique 3: Public-bias fade (selectively)
When 75%+ of public action is on one side, the operator may have moved the line beyond fair-value to attract balancing action. The opposite side then offers value.
Caveat: most modern sportsbooks adjust for public bias before posting initial prices. The "fade the public" strategy that worked in 2010-2015 has largely been arbitraged out. Specific situations still offer edge:
- Heavy public favorite as small home dog: Operators sometimes overcompensate. The favorite-as-dog scenario is occasionally +EV.
- Late line moves on no news: If a line moves significantly without injury news, sharp money is on the side that benefits. You can fade the public side that the line moves against.
- National-narrative games: Primetime NFL, marquee NBA games often have public-money distortions. Compare against earlier (Tuesday) lines for the same game.
What "value" is NOT
Avoid these false-value patterns:
- "This team has won 5 in a row, they have to win again." Streaks are not predictive; the market has already priced them in.
- "The line moved 2 points since Monday so there must be value left." If the line has already moved, the value has been captured by whoever moved it. You're getting the post-move price.
- "I have a strong gut feeling about this bet." Strong feelings are correlated with bets that have already moved through your decision-making for non-quant reasons. They're not evidence of edge.
- "This pick service has 60% wins this month." Survivorship bias. Pick services that hit 60% lose 40% of the time at -110, which is exactly break-even minus the service's monthly fee. The math doesn't work.
How to verify you're actually finding value
The verification is closing line value over a large sample. If you think you're finding value but your average CLV is negative, you're not finding value — you're finding variance.
See our CLV guide for the full tracking methodology. The short version:
- Track every bet (not just memorable ones) for 200+ bets minimum
- For each bet, record both the price you got and the closing price
- Average your CLV across all bets
- If average CLV is positive over 200+ bets, your value-detection is working
- If average CLV is zero or negative, your value-detection is biased and you should adjust strategy
Tools to use
Our no-vig calculator strips operator margin to give you fair-probability per side. Our parlay calculator compounds American odds across legs. Our Kelly Criterion calculator sizes bets given your edge estimate.