Expected Value in Sports Betting, Explained
A rigorous, numbers-first walkthrough of how professional bettors quantify edge. Three worked examples — moneyline, spread, totals — the formula derivation, the mistakes that kill casual bettors, and how EV feeds directly into Kelly staking.
Quick Calculator
Expected value is the mathematical backbone of every sustainable betting operation in the world. It strips away the noise of any single bet's outcome and answers one question: if I placed this exact bet infinitely many times, would I end up ahead or behind, and by how much per bet? If the answer is positive, the bet is a good decision even if it loses tonight. If the answer is negative, the bet is a bad decision even if it wins tonight.
Understanding EV is what separates gamblers from bettors. Anyone can pick winners occasionally. Only bettors who consistently identify bets with positive expected value relative to the market price generate positive returns over a career.
1. The Core EV Formula
EV per unit stake = (p × b) − (1 − p) × 1 where p = your estimated true win probability (0..1) b = net decimal profit per unit = (decimal odds − 1) Equivalent form: EV = p × decimal − 1 Dollar EV on a stake S: $EV = S × EV per unit Break-even probability: p* = 1 ÷ decimal
The derivation is elementary. When the bet wins (with probability p) you gain b units of profit. When it loses (with probability 1 − p) you lose 1 unit. The probability-weighted average of those two outcomes is your expected value. If p × b exceeds (1 − p), the bet is +EV. If not, it's −EV.
The decisive input is p, your probability estimate. The book gives you decimal odds; you must supply your own p. If your p equals the no-vig fair p implied by the market, your EV is slightly negative (you pay the vig). If your p exceeds the no-vig fair p, you have positive edge.
2. Worked Example — Moneyline
Celtics at Heat. Sportsbook prices the Heat at +135 (decimal 2.35). Your model gives the Heat a 45% win probability.
p = 0.45 decimal = 2.35 b = 1.35 EV = (0.45 × 1.35) − (0.55 × 1) = 0.6075 − 0.55 = +0.0575 per unit ROI = +5.75% $EV on $100 stake = +$5.75 Break-even p = 1 / 2.35 = 42.55%
Your 45% is 2.45 percentage points above the 42.55% break-even. That gap is your edge. A +5.75% EV moneyline is a strong bet — place it.
3. Worked Example — Point Spread
Chiefs −3 priced at −115 (decimal 1.870). Your model gives Chiefs a 55% chance of covering.
p = 0.55 decimal = 1.870 b = 0.870 EV = (0.55 × 0.870) − (0.45 × 1) = 0.4785 − 0.45 = +0.0285 per unit ROI = +2.85% $EV on $100 stake = +$2.85 Break-even p = 1 / 1.870 = 53.48%
Edge is only 1.52 percentage points (55 − 53.48). Real, positive, and playable — but smaller, so stake smaller per Kelly sizing.
4. Worked Example — Totals (Over/Under)
Under 8.5 priced at −120 (decimal 1.833). Your model gives Under a 53% probability.
p = 0.53 decimal = 1.833 b = 0.833 EV = (0.53 × 0.833) − (0.47 × 1) = 0.4415 − 0.47 = −0.0285 per unit ROI = −2.85% $EV on $100 stake = −$2.85 Break-even p = 1 / 1.833 = 54.55%
Your 53% falls below the 54.55% break-even. Even though you "like the under," at this price it is −EV by 1.55 points. Pass the bet. You would need either a better price or a stronger model estimate.
5. The Five EV Mistakes That Bankrupt Bettors
Ignoring the vig
At −110 both sides your estimate must exceed 52.38% just to break even. Estimating "I think they win" and not running the math puts you at −4.76% EV by default.
Overestimating your edge
"I think they win 65%" on a market implying 55% is usually wrong. Sharp closing lines are efficient; edges above 5% are rare. Regress your gut estimates aggressively toward the market.
Recency bias
Last weekend's 3-game win streak is not evidence your p-estimates are calibrated. Track 200+ bets before trusting your hit rate vs your predicted rate.
Chasing parlays
Parlay EV compounds the vig. Four −110 legs in a parlay typically produces book holds of 18–25%. Even +EV singles become deeply −EV when parlayed.
Ignoring line movement = ignoring CLV
If you repeatedly bet a line and the market moves away from you, your p was below fair. If the line moves toward you, you probably had +EV. Closing line value is the best live proxy for your real EV.
6. From EV to Stake Size — The Kelly Bridge
EV tells you whether to bet. Kelly staking tells you how much. The full Kelly fraction is:
f* = EV per unit ÷ (decimal − 1) = ((p × b) − (1 − p)) ÷ b Example: p=0.45, decimal=2.35, b=1.35, EV=+0.0575 f* = 0.0575 ÷ 1.35 = 4.26% Quarter Kelly (conservative) = 4.26% ÷ 4 = 1.07% of bankroll Half Kelly = 2.13% of bankroll
Most serious bettors stake fractional Kelly (0.25× to 0.5×) because probability estimates always contain error. A 5% EV estimate might really be 2% EV; full Kelly on 5% would massively over-stake. Fractional Kelly trades a bit of expected growth for a lot of variance reduction and ruin protection.
7. Frequently Asked Questions
What is the minimum EV worth betting?
In practice, anything above 1–2% EV that you believe is real (not modelling noise) is worth staking — at a Kelly-appropriate size. Below 1% EV the transaction friction (time, book limits, variance) often outweighs the edge.
How many bets do I need before my EV shows up in ROI?
Variance dominates early. With a 3% EV and typical −110 pricing, 1,000 bets is the rough order of magnitude before observed ROI is likely within ±1% of your true EV. Under 200 bets, outcomes are dominated by luck.
Can EV be greater than 100%?
On a single outcome bet, no — EV per unit caps at (decimal − 1). A price of 3.00 with p = 1 gives EV = +2.00 per unit, i.e., 200% ROI. But on combinations (parlays, exotics), theoretical EVs can exceed 100% when you have enormous edges, though these are typically book errors that get voided.
Is EV the same as edge?
Almost, but not exactly. Edge is usually defined as (your p) − (fair no-vig p) expressed as percentage points. EV is the dollar result per unit. Roughly, EV% ≈ edge × (decimal − 1), but the exact relationship depends on odds.
How do I estimate p accurately?
Build a model (statistical, historical, or market-based). Back-test against closing lines. Use multiple sharp books' no-vig consensus as a prior. Pure intuition without a model or edge mechanism almost always produces p-estimates worse than the market.
Ready to run your own numbers? Use the EV calculator, then size the bet with the Kelly calculator, and track outcomes against closing line value.
Responsible gambling notice. This article is educational only. Nothing here is a betting recommendation. Sports betting carries real financial risk and is inherently variance-heavy. Never stake money you cannot afford to lose. If you or someone you know is struggling with gambling, help is available at BeGambleAware.org or 1-800-GAMBLER (US). Must be of legal betting age.