Expected Value in Sports Betting: The Only Metric That Matters

Expected value (EV) in sports betting is the average profit or loss per bet over a large sample, calculated as EV = (win probability × profit if win) − (loss probability × stake). A bet with positive expected value (+EV) is profitable on average; a bet with negative expected value (-EV) loses money on average. The entire sportsbook business model is built on offering -EV bets to the public via the vigorish. Beating the market requires finding +EV opportunities where your probability estimate exceeds the market's implied probability by enough to overcome the vig. Monte Carlo simulation produces the probability estimates; EV quantifies the edge; and the Kelly Criterion determines how much to bet on that edge.

What Is Expected Value?

Expected value is one of the most fundamental concepts in probability theory, formalized by Christiaan Huygens in 1657 and later refined by Pierre-Simon Laplace. It represents the long-run average outcome of a random event repeated many times. In gambling, it tells you exactly how much you will make or lose per bet if you make the same bet thousands of times.

The formula is deceptively simple:

The key word is "true probability." The sportsbook's odds embed an implied probability, but that implied probability includes the bookmaker's margin (vig). The true probability is what you believe the actual chances are, independent of what the sportsbook charges. The gap between the true probability and the implied probability is where EV lives.

EV Calculations: Worked Examples

Example 1: Standard -110 Spread Bet

You bet $110 on a spread at standard -110 odds. If you win, your profit is $100. If you lose, you lose $110.

If the true win probability is 50% (no edge):

If the true win probability is 53% (small edge):

If the true win probability is 57% (strong edge):

Example 2: Moneyline Underdog at +200

You bet $100 on an underdog at +200. If you win, your profit is $200. If you lose, you lose $100.

The implied probability at +200 is 100/300 = 33.3%. If your model says the true probability is 38%:

This example illustrates why win rate alone is meaningless. A 38% win rate looks terrible. But at +200 odds, it is extremely profitable. Conversely, a 65% win rate at -300 odds is negative EV: (0.65 × $33.33) − (0.35 × $100) = $21.67 − $35.00 = -$13.33 per bet.

Example 3: Over/Under Total at -108

You bet $108 on the over at -108 odds. Profit if win: $100. Your Monte Carlo model estimates a 56% probability of the over hitting:

Why EV Is the Only Metric That Matters

Win Rate Is a Vanity Metric

The sports betting community is obsessed with win rate. Tipsters advertise "62% win rate!" as proof of skill. But win rate is meaningless without context on the odds. Consider two bettors over 1,000 bets:

Bettor A wins 62% of bets and loses $7,000. Bettor B wins only 44% and makes $5,600. Win rate told a completely misleading story. EV told the truth. This is not a hypothetical edge case; it is the normal state of affairs in sports betting. Heavy favorites win more often but pay less. Underdogs lose more often but pay more. Only EV captures the interaction between probability and payoff that determines profitability.

The Law of Large Numbers

The law of large numbers guarantees that over a sufficient sample size, your actual profit will converge to your expected value. A bet with +$5 EV will not win $5 on every play. It might win $100 this time, lose $110 next time, win $100 three times in a row, then lose five in a row. But over thousands of bets, the average profit will approach $5 per bet with increasing precision.

This is the foundational principle of professional sports betting: identify +EV bets, bet them at the right size, and let the law of large numbers do the rest. Short-term variance is noise. Long-term EV is signal. Professional bettors do not get excited about individual wins or upset about individual losses. They care about whether the bet was +EV at the time it was placed, because that is what determines long-term outcomes.

The Vig Tax

Every sportsbook bet carries a built-in negative EV for the bettor: the vigorish (vig or juice). Standard American odds of -110 on both sides of a spread imply that each side has a 52.38% chance of winning (110/210 = 52.38%). But since both sides cannot have a >50% probability simultaneously, the total implied probability is 104.76%. That extra 4.76% is the vig — the bookmaker's profit margin.

To find a +EV bet, your edge must be large enough to overcome the vig. At -110 odds, you need a true probability of at least 52.38% to break even. Anything above that is +EV; anything below is -EV. The vig sets the minimum edge required for profitability, which is why beating the sportsbooks requires a genuine informational or analytical advantage, not just "knowing sports."

How Monte Carlo Simulation Produces EV Estimates

The EV formula requires one critical input: the true win probability. But nobody knows the true probability of a sporting event. The best we can do is estimate it, and the quality of that estimate determines whether our EV calculations are reliable.

Monte Carlo simulation is the most rigorous method for estimating probabilities because it models the full range of possible outcomes rather than producing a single prediction. Here is how it connects to EV:

1. Model the game: Define probability distributions for every key variable (team efficiency, pace, shooting, turnovers, injuries, etc.)
2. Simulate 10,000 times: Run the game with randomized inputs drawn from those distributions
3. Count outcomes: Win probability = (number of simulations Team A won) / 10,000. Spread coverage = (number covering the spread) / 10,000.
4. Compare to market: The sportsbook's -110 line implies 52.4%. If the model says 57%, the edge is approximately 4.6 percentage points.
5. Calculate EV: EV = (0.57 × $100) − (0.43 × $110) = +$9.70 per $110 bet
6. Size the bet: Feed the probability and odds into the Kelly Criterion to determine the optimal wager size

The precision of the Monte Carlo estimate is critical. With 10,000 simulations, the standard error of a 57% estimate is approximately 0.5 percentage points. The true probability is 57% plus or minus 1%, with 95% confidence. That precision is sufficient to distinguish a 57% probability (strongly +EV) from a 53% probability (barely +EV) with reasonable confidence.

But the Monte Carlo estimate is still an estimate, not ground truth. This is why Bayesian calibration is essential: it adjusts raw model probabilities based on historical accuracy. If the model's 57% predictions historically win only 54% of the time, the calibrated probability should be 54%, and the EV calculation should use 54%. Using uncalibrated probabilities overstates EV and leads to overconfident bet sizing.

Edge vs. EV: Related but Different

Edge and EV are closely related but measure different things:

A 5-percentage-point edge translates to different EV amounts depending on the odds. A 5% edge at -110 produces about +4.5% ROI. A 5% edge at +200 produces about +15% ROI. The same probability edge is far more valuable on underdog bets because the payoff multiplier is larger.

This is one reason why contrarian underdog betting can be so profitable when done rigorously. The market tends to slightly overprice favorites (due to public bias toward betting on the better team), and any mispricing on an underdog translates to disproportionately high EV because of the larger payoff. Monte Carlo models that can detect even small probability edges on +150 to +300 underdogs can generate substantial expected profit.

How to Find +EV Bets

Method 1: Model-Based Edge Detection

Build or use a quantitative model (like Monte Carlo simulation) to estimate probabilities for each game. Compare your estimates to the sportsbook's implied probabilities. When your model's probability exceeds the break-even threshold, you have a +EV opportunity.

This is the approach Olympus Bets uses. The advantage is that model-based edges can be systematic, repeatable, and applied across hundreds of games per season. The disadvantage is that the model must be well-calibrated; an overconfident model will identify phantom edges that do not actually exist.

Method 2: Market-Based Arbitrage

Different sportsbooks often post different odds on the same game. When the gap is large enough, you can find +EV by comparing one book's line to the market consensus. If the consensus implies 55% and one book is offering odds that imply only 50%, betting at that book is +EV even without a model.

This is essentially what line shopping accomplishes. Platforms like Unabated and OddsShark facilitate this by displaying odds across dozens of sportsbooks simultaneously. The edges are typically small (1-3%) but consistent and require no modeling expertise.

Method 3: Closing Line Value

Closing line value (CLV) is both a method and a validation tool. If you consistently bet on sides that see the line move in your direction before the game starts, you are likely capturing +EV. The closing line, established by the full weight of sharp and public money, is the market's most efficient probability estimate. Beating the close means you were right before the market caught up.

Common EV Mistakes

Mistake 1: Using Win Rate as a Proxy for EV

As demonstrated above, a high win rate can coexist with negative EV, and a low win rate can coexist with positive EV. Always calculate EV directly from probabilities and odds. Never judge a strategy by win rate alone.

Mistake 2: Ignoring the Vig in EV Calculations

Some bettors calculate EV using the sportsbook's implied probability as the "true" probability, which always produces EV = 0 or slightly negative (the vig). The whole point of EV analysis is to use your own probability estimate, which may differ from the market's. If you use the market's probability, you are just confirming that the vig exists, which is not useful.

Mistake 3: Chasing Large Edges Without Calibration

A model that says 75% when the market says 60% shows a massive edge. But is the model right? Without calibration, you cannot know. Large edges are more likely to be model errors than genuine opportunities. Empirically, the highest-edge picks across all sports betting models tend to perform worst — this is the overconfidence inversion that plagues uncalibrated models. Always be more skeptical of larger edges, not less.

Mistake 4: Calculating EV Without Accounting for Correlation

If you place five bets on the same night's NBA overs, those bets are not independent. A systematic factor (like all games being high-scoring due to scheduling or back-to-backs) affects all five. Your total EV is not simply 5 × individual EV because the outcomes are correlated. Risk-adjusted EV accounts for this by discounting correlated bets.

Mistake 5: Confusing EV with Certainty

A +EV bet is not guaranteed to win. A $10 EV bet might win $100 or lose $110 on any individual play. The +EV designation means that if you made this exact bet thousands of times, you would average $10 profit per bet. On any single bet, you are still subject to variance. This is why bankroll management and proper bet sizing are inseparable from EV analysis.

EV and Kelly Criterion: The Complete System

EV tells you whether to bet. The Kelly Criterion tells you how much. Together, they form a complete betting system:

1. Estimate probabilities using Monte Carlo simulation
2. Calibrate probabilities using Bayesian methods to correct overconfidence
3. Calculate EV for each potential bet
4. Filter: only proceed with bets that have positive EV exceeding a minimum threshold
5. Size bets using Kelly: larger bets for larger edges, smaller for smaller
6. Track CLV to validate that the edges were real

Steps 1-3 are the analytical layer: finding the edge. Steps 4-5 are the execution layer: exploiting the edge optimally. Step 6 is the validation layer: confirming the edge existed. Skip any step and the system degrades. Find edges but size bets randomly? You leave money on the table. Size bets perfectly but on phantom edges? You lose faster. This interconnected system is what separates quantitative sports betting from recreational gambling.

At Olympus Bets, every projection includes its calculated edge, EV, and Kelly-optimal unit size. This transparency allows you to see not just which side to bet but why and how much. The interactive Kelly calculator lets you plug in your own probability estimates and odds to compute EV and optimal bet size for any scenario.

Practical EV Thresholds

Not every +EV bet is worth making. Transaction costs (time, account management, the risk of being limited by sportsbooks) create a practical floor below which +EV bets are not worth the effort.

The 2-5% range is where most legitimate +EV opportunities live in major sports markets. Finding consistent 5-10% edges requires either excellent modeling, market access to soft lines, or both. Edges above 10% on major market sides (not props or exotics) should be viewed with skepticism and verified through multiple independent methods before sizing up.

Further Reading

• Monte Carlo Simulation Guide — how probability estimates are generated
• Kelly Criterion Guide — optimal bet sizing once EV is calculated
• Kelly vs Flat Betting — why edge-proportional sizing outperforms flat betting
• Bayesian Calibration Guide — ensuring probability estimates are accurate
• Closing Line Value — validating that your edges are real
• Monte Carlo vs Poisson vs Elo — comparing prediction model approaches
• Kelly Criterion Calculator — interactive EV and sizing tool
• Our Methodology — full technical overview