SorteCalc
How to Calculate Sports Betting Expected Value From Odds and Your Own Probability
← BlogSports betting expected value

How to Calculate Sports Betting Expected Value From Odds and Your Own Probability

2026-09-11 · 22 min · 4,296 words

You are staring at your phone, watching the odds flash from +140 to +155 on the Timberwolves to cover the spread. Your gut says yes. Your buddy who watched every game this season says yes. But your bankroll is down 40% this month because you have been saying yes to bets that feel right but mathematically bleed money. The difference between the bettors who last and the ones who reload their accounts every Friday is not luck, and it is not "knowing the sport." It is the discipline to calculate Expected Value (EV) before risking a dollar.

EV is the mathematical average result of a bet if you could place it thousands of times. A positive EV (+EV) means you will profit in the long run. A negative EV (-EV) means the sportsbook has priced the bet so that you will lose money over time, even if you win tonight. This article will teach you to calculate EV precisely, but more importantly, it will teach you how to generate the one input the sportsbook cannot give you: your own accurate probability of an event occurring. Without that, the formula is just arithmetic on garbage.

Why Expected Value is the Only Metric That Matters

Imagine I offer you a wager on a fair coin flip. You pay me $1 to play. If it lands heads, I pay you $1.90. If it lands tails, you lose your dollar. You might take this bet once for fun. You might even win. But if you play this game every day for a year, you will lose five cents per flip on average. The EV is negative.

Professional bettors do not try to predict winners. They try to find pricing errors. When a sportsbook offers +200 on a team (implying a 33.3% chance of winning), but your analysis shows the team actually wins 40% of the time, you have found a gap. That gap is value. Every other metric—win percentage, hot streaks, "feeling confident"—is noise unless it translates into an EV calculation that shows you are getting paid more than the risk requires.

The goal of this guide is to teach you to think like the person setting the line, not the person betting into it. You will learn to reverse-engineer the sportsbook's implied probability, strip out their commission (the vig), and replace it with your own true probability derived from data or market inefficiencies.

The Three Ingredients of Any Bet's Value

Every EV calculation requires three inputs. Two are objective facts you can look up. One is a subjective estimate that will make or break your profitability.

The Sportsbook Odds: This is the price. American odds (+150, -110), decimal odds (2.50), or fractional odds (3/1) all express the same thing: how much the book will pay you relative to your stake. The odds also contain the book's profit margin, which is why you must never accept them at face value as the "true" chance of winning.

Your Probability Estimate: This is your assessment of how likely the outcome actually is. If the odds imply a 40% chance, but you know the star quarterback is playing with a hidden injury and the backup has a 0-4 record in away games, your probability might be 32%. This is where you build your edge. The sportsbook provides the price; you provide the analysis.

Your Stake: This scales the result. A +EV of 2% on a $10 bet is 20 cents. On a $1,000 bet, it is $20. The stake does not change whether the bet is +EV or -EV, but it determines the magnitude of your long-term results. Bet too much on a +EV wager and variance will wipe you out before probability math saves you. Bet too little and you leave profit on the table.

The Expected Value Formula: A Step-by-Step Calculation

The formula for Expected Value is straightforward:

EV = (Probability of Winning × Potential Profit) − (Probability of Losing × Amount Staked)

Let us walk through a concrete example using decimal odds, as they are the cleanest for this math.

Step 1: Find the odds. The Miami Heat are listed at 2.50 to win outright. This means for every $1 you bet, you get $2.50 back if they win (your $1 stake plus $1.50 profit).

Step 2: Determine your true probability. After analyzing matchups, recent defensive ratings, and injury reports, you believe the Heat win this game 45% of the time. This means they lose 55% of the time.

Step 3: Calculate potential profit. On a $100 bet at 2.50, your profit is $150 (you get back $250 total, minus your $100 stake).

Step 4: Plug into the formula.

  • Win component: 0.45 (probability) × $150 (profit) = $67.50
  • Loss component: 0.55 (probability) × $100 (stake) = $55.00
  • EV = $67.50 − $55.00 = +$12.50

This means that if you could make this exact bet a thousand times, you would expect to average a $12.50 profit per $100 wagered. The bet is +EV.

Now flip the scenario. Suppose you estimate the Heat's true win probability is only 35%. The calculation becomes (0.35 × $150) − (0.65 × $100) = $52.50 − $65 = −$12.50. You should not touch this bet, even if you "like" the Heat.

A five-step flowchart showing the decision process for sports betting expected value: Locate Odds, Estimate True Probability, Calculate Implied Probability, Compare Values to Identify Edge, and Place Bet (+EV only).
The Expected Value calculation involves a systematic five-step process, starting with identifying odds and culminating in placing a bet only when a positive expected value is determined.

Decoding the Language of Odds: American, Decimal, and Fractional

Odds are just different dialects of the same language: probability. You must be fluent in converting between them because different sportsbooks default to different formats, and you need to see the implied probability instantly.

Decimal Odds (2.50, 1.80): The global standard. Multiply your stake by this number to see your total return (stake + profit). To find implied probability, divide 1 by the decimal odds. A 2.50 line implies a 40% chance (1 / 2.50 = 0.40).

American Odds (+150, -200): The moneyline format. Positive numbers show profit on a $100 bet (+150 means win $150). Negative numbers show how much you must bet to win $100 (−200 means bet $200 to win $100).

  • To convert positive American to decimal: (American / 100) + 1. So +150 becomes (150/100) + 1 = 2.50.
  • To convert negative American to decimal: (100 / abs(American)) + 1. So −200 becomes (100/200) + 1 = 1.50.

Fractional Odds (3/2, 1/2): Common in UK racing. The first number is profit; the second is stake. 3/2 means $3 profit for every $2 staked. To convert to decimal, divide the first number by the second and add 1: (3/2) + 1 = 2.50.

Format Example Formula to Convert to Decimal Implied Probability Profit on $100 Stake
American (Positive) +150 (150 / 100) + 1 40.0% $150
American (Negative) -200 (100 / 200) + 1 66.7% $50
Decimal 2.50 2.50 40.0% $150
Fractional 3/2 (3 / 2) + 1 40.0% $150
Fractional 1/2 (1 / 2) + 1 66.7% $50

Once you have decimal odds, you can calculate implied probability instantly. This is your starting point. If you cannot beat the implied probability with your own estimate, you are paying the sportsbook's rent.

The Art and Science of Estimating True Probability

The EV formula is useless if your probability input is wrong. This is where bettors separate themselves from Excel jockeys. You need a systematic way to generate probabilities that are more accurate than the market's.

Method 1: Market De-Vigging

The sportsbook builds a profit margin (vig) into the odds. On a standard two-outcome bet like Over/Under 47.5 points, you might see both sides at −110. The implied probability of −110 is 52.4%. Add both sides: 52.4% + 52.4% = 104.8%. The extra 4.8% is the vig.

To find the market's "true" probability without vig, divide each implied probability by the total book percentage (1.048). So 52.4% / 1.048 = 50%. The market thinks there is a coin-flip chance on each side. If your analysis suggests the Over hits 54% of the time, you have found a 4% edge.

Method 2: Bottom-Up Statistical Modeling

Build your own probability from data. For player props, calculate the mean and standard deviation of a player's recent performance. For soccer, use Poisson distribution to estimate the likelihood of specific scorelines based on expected goals (xG) models. For NFL, use Elo ratings adjusted for injuries, weather, and travel.

The danger here is overfitting. If you build a model that perfectly predicts last season but fails this season, you have curve-fitted noise. You must backtest against historical data you did not train on.

To see how this works in practice, walk through a player prop example. Suppose you are pricing the Over on 2.5 three-pointers made for a guard. The market posts both sides at −110, so the de-vigged baseline is 50%. You build your own estimate from four quantified inputs:

  • Base frequency. Suppose your tracking shows the player has hit this line in 55% of games this season, but that is a small sample. You regress it toward a long-term mean of 45%, giving you a weighted base rate of 50%.
  • Opponent adjustment. The opponent is missing its primary wing defender. Your past matchup data suggests this adds two percentage points to the hit rate. 50% → 52%.
  • Injury and usage. A ball-dominant teammate is out, which funnels more shot attempts to your player. You estimate a two percentage point bump. 52% → 54%.
  • Pace. The projected game pace is neutral relative to the player’s average, so you add zero. 54% → 54%.

Your bottom-up estimate is 54%. Against the market’s 50%, that looks like an edge. But now you stress-test the uncertainty. If your opponent adjustment is really only one point because the replacement defends well, and your usage bump is closer to one point because the offense stagnates, your estimate drops to 52%. If your base rate is polluted by variance and the true mean is 48%, the estimate drops to 51%. Your plausible band is roughly 51% to 57%.

At −110, the break-even is 52.38%. If your estimate is 54% but your error band is ±3%, the floor of that band sits at 51% — below the break-even. That means you need a better price than −110 before betting. At +100, the break-even drops to 50%, so even your conservative estimate clears the hurdle with room to spare. The formula gives you the number; the error band tells you whether the number is strong enough to act on.

Method 3: Information Arbitrage

Sometimes you know something the market has not priced in yet. The lineup drops and the backup goalkeeper is starting, but the odds have not moved. You have 90 seconds to calculate the new true probability and bet before the line moves. This requires pre-built models that can ingest new data instantly.

Method Description Pros Cons Skill Required
Market De-Vigging Remove bookmaker margin from odds to find consensus probability Fast; leverages collective market intelligence Assumes market efficiency; limited to clear two-way markets Low to Medium
Basic Statistical Analysis Recent form, injuries, head-to-head records Accessible; intuitive Vulnerable to recency bias; misses interaction effects Medium
Advanced Modeling Poisson, Elo, machine learning on large datasets Captures subtle edges; scalable Requires coding/data science; risk of overfitting High

Most professionals use a hybrid. They start with the de-vigged market probability, then adjust it based on their proprietary data. If you are starting out, master de-vigging first. It teaches you what the market knows before you try to outsmart it.

Implied Probability: A Shortcut to Finding Value

Implied probability is the translation of odds into percentages. It tells you the break-even point. If a bet has an implied probability of 40%, you must believe the true probability is 41% or higher to make it +EV.

The calculation is simple for decimal odds: Implied Probability = 1 / Decimal Odds.

For American odds:

  • Positive (+150): 100 / (150 + 100) = 40%
  • Negative (−200): 200 / (200 + 100) = 66.7%

The value betting rule is binary: Your Estimated Probability > Implied Probability = +EV.

This reframes the hunt. You are not looking for "who will win." You are looking for situations where the market underestimates the likelihood of an outcome. If the implied probability is 25% (decimal 4.0) and you calculate the true probability at 30%, you have a massive 5% edge. Over 100 bets at $100 each, that edge compounds significantly.

A side-by-side bar chart comparing 'Implied Probability' at 40% with 'Your Estimated Probability' at 45%. The 5% difference between the two bars is highlighted and labeled as 'Positive EV Zone / Your Edge'.
Identifying your 'edge' in sports betting means finding situations where your estimated probability of an outcome is higher than the implied probability derived from the odds.

Where to Hunt for +EV Opportunities

+EV bets do not sit in the main lobby of a sportsbook. They hide in inefficiencies. You will not find them by betting the same NFL sides as everyone else at opening lines that have been sharpened by millions of dollars.

Line Shopping: Sportsbooks price risk differently. Book A might have the Knicks at +140. Book B has them at +155. Both cannot be correct. If you calculate the true probability at 43%, +140 is −EV (implied 41.7%), but +155 is +EV (implied 39.2%). Having accounts at five or six books lets you pick the price, not just the pick. Tools like SorteCalc's calculators can help you quickly compare these margins.

Niche Markets: Major markets (NFL point spreads, Premier League moneylines) are efficient. The sharp money has already corrected the lines. But player props on a Tuesday night MAC football game? The bookmaker might be using a simple algorithm or outdated depth charts. Specialize in one small sport or league. Know the third-string running back's injury status before the oddsmaker updates the model.

Timing: Bet early if you have information the market will later discover. If you model weather better than the books and see a 70% chance of high winds that will suppress passing, bet the under before the line drops. Conversely, bet late if you are reacting to confirmed lineup news (like a late scratch) that has not propagated to all books yet.

Promotions and Arbitrage: Occasionally, books offer "no-vig" lines or odds boosts that create temporary +EV situations. These are rare and often have stake limits, but they are worth taking when the math works.

The Gambler's Fallacy: Common Mistakes That Erode Your Edge

Calculating EV is the easy part. Executing the strategy without self-sabotage is what destroys bankrolls.

Overconfidence in Your Model: You ran the numbers. Your model says the Celtics have a 65% chance to win, but the market implies 55%. You bet 5% of your bankroll. The Celtics lose. You bet 10% on the next "sure thing" to make it back. This is bankruptcy in three moves. A +EV bet can lose 40% of the time if your edge is small. Variance is real, and it hurts.

Ignoring Variance: Expected Value is a long-term average. In the short term, you are at the mercy of standard deviation. A bettor with a 2% edge on 100 bets will still lose money roughly 38% of the time due to variance. You must size your bets so that a 10-bet losing streak (which will happen) does not force you to stop.

Chasing Losses: The math does not care that you are "due" for a win. Each bet is independent. If you deviate from your EV-based staking plan because you are tilted after a bad beat, you are no longer a value bettor. You are a gambler.

Mistaking Activity for Edge: Betting every game on the Saturday slate because you are bored is not a strategy. If you cannot calculate a specific edge for a wager, you are donating to the sportsbook. Wait for the lines where your number and the book's number diverge significantly.

Advanced Applications: EV in Parlays, Teasers, and Live Betting

Once you master single-game EV, you encounter exotic bets where the math gets punitive or complex.

Parlays: The sportsbook does not just charge vig on each leg; it compounds. Start by separating the sportsbook’s break-even probability from the event’s true probability. A −110 line requires you to win 52.38% of the time just to break even, but a fair two-outcome bet might have a true 50% chance on each side. Now parlay two fair 50/50 legs. Your true probability of hitting both is 0.50 × 0.50 = 25%. The fair payout for a 25% chance is +300 (decimal 4.0). A standard two-leg parlay using −110 lines pays roughly +264 (decimal 3.64). Against fair 50/50 legs, the EV is (0.25 × $264) − (0.75 × $100) = $66 − $75 = −$9 per $100 wagered. The entire shortfall comes from the vig embedded in each leg.

Can a parlay ever be +EV? Only if the individual legs are +EV on their own. Suppose you estimate Leg A’s true win probability at 56% despite the −110 price, and Leg B’s at 55%. Your true parlay hit rate is 0.56 × 0.55 = 30.8%. The book still pays +264, which implies a 27.5% break-even rate. Your EV is (0.308 × $264) − (0.692 × $100) = $81.31 − $69.20 = +$12.11 per $100. The parlay magnifies your edge, but it also magnifies variance. You will hit less than one in three, so bankroll management becomes critical.

The exception is the correlated parlay. If you bet a team to win and the same team to cover the first half spread, these events are positively correlated. Winning the game usually requires leading at halftime. If the book does not adjust the odds to account for this correlation (many do not allow these bets or adjust them heavily), you can find +EV. But this requires understanding the covariance between the events.

Teasers: These let you buy points on spreads. A two-team teaser (buying 6 points in football) must capture value by crossing key numbers (3 and 7). Teasing a team from +2.5 to +8.5 is valuable because you capture both the 3 and the 7. Teasing from +7.5 to +13.5 is less valuable because you skip the 7. Calculate the EV of the teaser by comparing the new implied probability against your true probability. Most teasers are −EV unless strategically placed around key numbers.

Live Betting: The odds move in real-time. A goal is scored, and the underdog's odds jump from 4.0 to 6.0. If your model says the true probability is now 20% (implied odds 5.0), but the book offers 6.0 (implied 16.7%), you have found value. But you have 15 seconds to calculate this and place the bet before the next play starts. Live betting EV requires automation or pre-calculated lookup tables. You cannot run a full Poisson model during a commercial break.

Closing Line Value: The Scoreboard for Your Probability Estimates

Here is the uncomfortable problem everything so far creates. With a genuine 3% edge, you can place 200 bets and still be down money. Variance does that. So if short-term results cannot tell you whether your probability estimates are any good, what can?

The answer is Closing Line Value (CLV), and it is the metric professional bettors actually watch. The logic runs like this: a betting line opens rough and gets sharper as money arrives. By kickoff, the odds reflect injury news, weather, lineup confirmations, and the positions of the sharpest bettors in the world. The closing line at the most efficient sportsbooks is the closest thing to a true probability the market produces. If you consistently bet at better prices than the close, your estimates are beating the collective intelligence of the market. If you consistently bet worse, they are not — no matter what your last ten results say.

The measurement is simple. Say you bet Arsenal at 2.10 on Tuesday (implied 47.6%). By Saturday, the line has shortened to 1.95 (implied 51.3%). You beat the close by 3.7 percentage points. That is real CLV, and there is a shortcut for converting it directly into expected value: EV ≈ (your odds ÷ closing odds) − 1. Here, (2.10 ÷ 1.95) − 1 = +7.7%. You locked in a price roughly 7.7% better than the market's final answer.

Now the practice. Log every bet in a spreadsheet with two columns beyond the basics: the odds you took, and the closing odds. After 300 or more bets, average your CLV. Read it this way:

  • Positive average CLV (even +1–2%) while your bankroll is down: your process works. Keep going. The results will follow.
  • Negative average CLV while your bankroll is up: you are lucky, not good. This is the most dangerous position in betting, because the wins are training you to keep doing something that will eventually collect its debt.
  • Negative CLV and down: stop and rebuild your estimation process before staking another unit.

Two caveats keep CLV honest. First, news luck cuts both ways. Sometimes you bet and a star striker is ruled out an hour later, handing you a beat-the-close you did nothing to earn. Over hundreds of bets, random news moves for and against you average out to roughly zero — which is exactly why persistent CLV over a large sample is signal, while one dramatic beat-the-close is not. Second, in thin markets your own bet can move the closing line, which flatters your numbers. CLV means the most in liquid markets where your stake is a rounding error.

One practical warning: sportsbooks track CLV too. Consistently beating the close is the single clearest fingerprint of a winning bettor, and soft books respond by limiting your stakes. If your accounts keep getting restricted, take it as confirmation your method works.

Frequently Asked Questions

Is a +EV bet guaranteed to win?

No. A +EV bet is a bet where the odds are in your favor, but the outcome is still probabilistic. If you have a 5% edge on a bet that wins 45% of the time (and loses 55%), you will still lose more often than you win on individual wagers. However, over a large sample size—hundreds or thousands of bets—the positive expected value will manifest as profit. Think of it like insurance companies: they price policies so that on average they collect more in premiums than they pay out, even though they cannot predict which specific houses will burn down.

How much should I bet on a +EV opportunity?

This is a question of bankroll management, not just EV. The Kelly Criterion is the standard answer: it calculates the optimal fraction of your bankroll to bet based on your edge and the odds. The formula is (BP − Q) / B, where B is the decimal odds minus 1, P is your probability of winning, and Q is your probability of losing (1 − P). If this gives you 0.02, you bet 2% of your bankroll. In practice, most bettors use "fractional Kelly" (betting half or a quarter of the full Kelly amount) to reduce variance. Betting the full Kelly on a sport with high variance (like baseball or hockey) can lead to devastating drawdowns. Tools like SorteCalc's Kelly Criterion calculator can run this math instantly for you.

What is a 'good' EV to bet on?

Most professional bettors will take any positive EV they can verify, typically in the +1% to +5% range. A 2% edge is solid; a 5% edge is excellent. Anything above 10% is suspicious—you have likely miscalculated either the probability or misunderstood the market conditions (like a player injury you missed). Consistency matters more than size. Making one hundred +2% EV bets is better than making ten +10% EV bets because the variance smooths out over volume.

How is the sportsbook's 'vig' or 'juice' related to EV?

The vig is the commission the sportsbook builds into the odds. On a standard −110 bet, you must risk $110 to win $100. The implied probability is 52.4%, but the true probability of a fair coin flip is 50%. That extra 2.4% is the vig. To find a +EV bet, your estimated probability must be high enough to overcome this margin. The break-even point for −110 is 52.38%, so anything above that is technically +EV. If you estimate a team wins 53% of the time, the calculation is (0.53 × $100) − (0.47 × $110) = $53 − $51.70 = +$1.30 per $110 staked. That is positive, but it is a razor-thin edge. After variance and the possibility that your 53% estimate is slightly off, you would want a significantly better price or a higher confidence level before risking meaningful capital.

Can I just use an online EV calculator?

Yes, and you should for speed. However, you must understand the mechanics. If you plug a number into a calculator and it says +EV, but you got your probability from a hunch, the calculator is validating garbage. Understanding the formula lets you sanity-check the tool's output and, more importantly, forces you to scrutinize the quality of your probability input. The calculator is only as good as the assumptions you feed it.

Does this concept apply to casino games like roulette or blackjack?

Yes, but with a critical distinction. In roulette, the probabilities are fixed and known. A single-zero wheel has a 2.7% house edge on every spin. The EV is always negative for the player; you cannot "out-analyze" physics. In blackjack, you can achieve +EV through card counting because you are tracking the composition of the remaining deck, effectively changing the probabilities in real-time. In sports betting, the probabilities are unknown and constantly shifting, which is why it is a beatable game for those who can estimate them more accurately than the market.

Sources