Key takeaways
- A bet is +EV when your win probability times the payout exceeds the loss probability times your stake, judged against the no-vig fair line rather than the posted odds.
- Edge equals your model probability minus the no-vig fair probability, and anything under 1% should be skipped because line movement and misjudged key numbers eat thin edges.
- Kelly sizing turns edge into a stake fraction, but quarter to half Kelly is the standard compromise because full Kelly assumes your probabilities are exact, and they never are.
- Consistently beating the closing line is the strongest available confirmation that your screening process has a real edge, audited over hundreds of bets.
- Line shopping is the cheapest edge source: a 10-cent difference between books is a real, measurable edge on every bet and requires no model.
Expected value betting is staking only when your fair probability beats the implied probability in the odds, with the gap being your edge. This guide runs one real market end to end: convert the odds, strip the vig, compute edge percentage, size with Kelly, then screen it through a calculator you can reuse on your own numbers.
What is expected value in sports betting and why does it matter?
Expected value betting is the long-run profit you make per unit staked on a wager, averaged across all outcomes. A bet is +EV when your win probability times the payout exceeds the loss probability times your stake. Because bookmakers bake margin into their prices, the baseline for judging any price is the no-vig line, not the posted number; Pinnacle's betting resources explain how removing the vig produces fair odds (source). CLV matters here because closing lines are the sharpest public prices available, and beating them consistently correlates with long-run profit even when short-term results are noisy. That is the core point: a bet can lose and still be +EV, because EV is about the quality of the price, not the outcome of a single game. If you want definitions for related terms like RLM and steam, the glossary covers them.
What is the expected value formula for betting and what do its inputs mean?
EV = (win probability × profit per bet) − (loss probability × stake). Win probability is your model's number, not the book's implied probability. Profit is the payout net of stake; stake is what you risk. If EV is positive, the bet is +EV and worth screening further.
Edge as a percentage of stake
Divide EV by stake and you get edge per bet, the number that feeds Kelly. A 2% edge at American odds of +150 means EV = 0.02 × stake on every identical wager. Track it per market: our graded history at https://mvpcapping.com/records logs every pick with its closing line value so you can check whether stated edges actually cleared.
Why bookmaker margin makes raw odds lie
The vig inflates implied probability above 100%, so odds alone overstate the required win rate. Strip the margin to get fair odds, then compare your probability against that. Raw odds versus fair odds is the whole edge calculation.
How do you convert odds into implied and fair probability?
American and decimal odds to implied probability
Implied probability is just odds inverted. For decimal odds, divide 1 by the price: 2.50 implies 40%. For American odds, use 100 divided by the price when the price is negative, and 100 divided by (price + 100) when positive.
| Odds format | Price | Implied probability |
|---|---|---|
| American | -150 | 60.0% |
| American | +130 | 43.5% |
| Decimal | 1.91 | 52.4% |
| Decimal | 2.20 | 45.5% |
Removing the vig to get no-vig fair odds
A two-way market's implied probabilities sum to over 100%. Strip the margin by normalizing: divide each side's implied probability by the total. If -150 and +130 sum to 103.5%, fair probabilities are 58.0% and 42.0%, giving fair odds of about -138 and +117. The multiplicative method assumes the book's margin is proportional across outcomes; it is the standard first pass, though low-liquidity markets can skew it.
When to trust your model probability over the market
Your edge is your model probability minus the no-vig fair probability. If you have no model, treat the de-vigged market as the baseline and screen against it, as covered in our EV guides. Once you do have one, stake with Kelly sizing rather than flat units; the Kelly criterion formalizes why overbetting a real edge costs more than underbetting it.
How do you calculate expected value in sports betting step by step?
Step 1: pull the best available line
Line-shop across your books before you touch a number. A side at -110 at one book and -105 at another changes the entire EV calculation, so the price you screen must be the best price you can actually click.
Step 2: compute implied probability
Convert the American price. For -110: risk 110 to win 100, so implied probability = 110 / 210 = 52.38%. For -105 it is 105 / 205 = 51.22%. Our betting cheat sheet lists these conversions so you are not redoing division every slate.
Step 3: derive fair probability from the no-vig line or model
Take the no-vig fair line for the market. If the sharp price on this side is -110 fair, fair probability = 52.38%. If your own model outputs a probability, use that instead and skip the de-vig entirely.
Step 4: compute EV and edge
Here is the full worked example, and it is worth walking through slowly because the direction of the edge is where most bettors get turned around. Say you can get an NFL side at -105, but the no-vig fair line says the side should be -110. Your price implies 105 / 205 = 51.22%. Fair probability is 110 / 210 = 52.38%. Edge is fair minus implied: 52.38% − 51.22% = +1.16 percentage points. Dollar EV on a $500 stake: payout at -105 is $476.19, so EV = (0.5238 × $476.19) − (0.4762 × $500) = $249.43 − $238.10 = +$11.34 per bet. Full Kelly says stake f = (b·p − q) / b, where b = 100/105 = 0.952, which gives (0.952 × 0.5238 − 0.4762) / 0.952 = 2.37% of bankroll; most sharps bet half that. Run the same math on the reverse, -110 against a -105 fair line, and EV is −$11.08 on $500. Same market, opposite sign, purely a function of the price you secured.
Step 5: sanity-check against CLV
A bet that beats the fair line should also beat the closing line. If you consistently grab -105 on sides that close -110, you are logging positive CLV, which is the only long-run confirmation your screening process has a real edge. If your CLV is flat or negative over a few hundred bets, the EV inputs, not the variance, are the problem.
What makes a bet positive expected value and how do you find +EV opportunities?
A bet is +EV when your win probability times payout exceeds the loss probability times stake. In practice, that means your number beats the market's number, and the cheapest way to find those spots is line shopping.
Line shopping is the cheapest edge source
If one book posts -3 (-105) while the rest sit at -3 (-115), that 10 cents is a real, measurable edge on every bet. With accounts at multiple books, you capture it instantly; with one book, you never see it. No model required.
Steam, RLM, and betting into moving markets
Beating the close matters more than beating the open. Pinnacle's own analysis shows closing line value is the most reliable predictor of long-term betting profit. Steam moves fast, so the edge is getting your bet in before the market fully adjusts, not after.
Model vs market: when to override
Trust your model only when it disagrees with a stale or slow-moving line, not the sharp one. Every pick we post at MVP Capping is auto-graded with CLV tracked, so you can audit whether our numbers actually beat the close. Where they don't, that's visible too.
How do you use an expected value betting calculator to screen bets?
Inputs and outputs of the embedded calculator
The calculator takes three inputs: the book's odds, your fair win probability (from a no-vig line or your own model), and your stake. It outputs EV in dollars, edge percentage, and a full-Kelly fraction. Worked example: -110 odds, model probability 52.4%, $1,000 stake. Implied probability is 52.38%, fair is 52.4%, edge is roughly 0.04% — a pass. At 54% model probability, edge is about 1.6%, EV is roughly $16 on the stake, and Kelly suggests staking about 3.2% of bankroll.
Setting a minimum edge threshold
Skip anything under 1% edge. Thin edges get eaten by odds movement between screen and bet, and by half-point or hook risk you misjudged. Set the floor once, apply it to every market, and log the picks you rejected so you can audit whether the floor is calibrated.
Common calculator mistakes (wrong vig method, stale fair line)
Two errors wreck the output. First, deriving fair odds by removing vig from both sides equally when the vig distribution is lopsided — use a true no-vig calculation, not a naive 50/50 split. Second, screening against a fair line that's hours old; a stale close means your edge is fiction. Recalculate against the current line before staking, then track CLV on every bet to confirm the edge was real.
How should stake sizing change based on expected value?
Kelly from edge and decimal odds
Kelly turns the edge from your EV calc into a stake fraction: f = (b·p − q) / b, where b is decimal odds minus 1, p is your model probability, and q is its complement. Take the earlier example: fair odds of 2.20 (p = 45.45%) priced at 2.30 (b = 1.30). That gives f = (1.30 × 0.4545 − 0.5455) / 1.30 ≈ 3.5% of bankroll.
Fractional Kelly and variance
Full Kelly assumes your probabilities are exact. They never are, and overestimating an edge by a point or two turns a growth-optimal stake into severe drawdowns, which is why most practitioners bet a fraction of the full amount. The Kelly criterion literature documents how sensitive growth is to estimation error. Quarter to half Kelly is the standard compromise: the 3.5% full-Kelly stake becomes 0.9% to 1.75%.
Bankroll units vs dollar stakes
Express stakes in units so sizing scales with bankroll, not mood. On a $5,000 roll, quarter Kelly on that bet is roughly 0.45 units, or $45. Recompute after every bankroll change; a fixed-dollar stake silently drifts off the Kelly fraction as the roll moves.
Which EV tools are worth paying for, and where does MVP Capping fit?
Criteria: raw feed, CLV grading, price
Judge any screening tool on three things: does it expose the raw data feed (implied probabilities, no-vig lines, line movement) or just the output pick; does it track CLV on every graded result and let you audit the grading; and is the price justified against the bankroll you're actually staking.
How the tools compare
Sharp App, EdgeAI Sports, The Over, and Sharp Edge are direct competitors in the picks and analytics space, while Action Network sits adjacent as a broader betting media product. The recurring complaints across this category are pricing that demands a dedicated bankroll and interfaces built for intermediate-to-advanced users. Those are real costs: if the tool costs $50+ monthly and you stake small, your edge gets eaten before the first bet.
Where MVP Capping falls short
MVP Capping's Edge tier runs $29/mo, shows auto-graded picks you can audit before paying, tracks CLV on every pick, and publishes calibrated confidence scores with no deleted losses. It shares the category's weaknesses, though: heavy US focus with fewer international markets, and a learning curve if you're not already fluent in CLV and Kelly. If you bet exclusively internationally or run an automated arbitrage operation without analytical input, skip it.
FAQ
Is a bet with negative EV ever worth placing?
No, not if your goal is long-run profit. A negative EV bet means the price is worse than the fair probability, so every identical wager leaks money on average. A single negative EV bet can still win, which is why outcomes mislead, but repeated negative EV staking guarantees drawdown regardless of how confident you feel about one game.
How much bankroll do I need for positive EV betting to matter?
Enough that Kelly-sized stakes produce meaningful dollars without risking ruin. With a 1-2% edge and quarter Kelly, a $5,000 roll stakes roughly $45 per qualifying bet; a $500 roll stakes single digits, so edges get eaten by minimum bet sizes and line movement. Below a few thousand dollars, EV betting is practice, not income.
Does beating the closing line always mean I had a +EV bet?
Not always, but it is the strongest available signal. CLV compares your price to the sharpest public number at close, and consistent positive CLV correlates with long-run profit even when short-term results are noisy. A single bet that beat the close can still lose, and occasional CLV wins can be luck; the audit runs over hundreds of bets.
Can I use an expected value calculator for parlays and props?
Yes, with more caution. For parlays, multiply the leg probabilities for your fair win probability and compare against the parlay payout, remembering compounded vig makes most parlays heavily negative EV. For props, fair probabilities are harder to derive because low-liquidity markets can skew the de-vig, so screen with a wider minimum edge and trust the numbers less.
What edge percentage is big enough to bet?
At least 1%. Thin edges get eaten by odds movement between your screen and your click, plus half-point or hook risk you may have misjudged. Set a 1% floor once, apply it to every market, and log the picks you reject so you can audit whether the threshold is calibrated. Edge plus Kelly fraction determines the stake.