Expected Value Betting Formula Explained

The math behind every smart bet, using cross-market signals to show where the formula finds real edges and where it honestly finds none

GhostLeg, 2026-08-10, 7 min

Most bets you will ever place have negative expected value. The sportsbook built that in before you placed your first dollar. The expected value betting formula tells you exactly how much; more importantly, it tells you when the market has made a mistake you can exploit.

Here is the formula upfront, because every EV explainer buries it halfway through the article:

EV = (P_win × Payout) − (P_lose × Stake)

Where P_win is your estimated probability the bet wins, Payout is the profit on a winning bet, P_lose is 1 − P_win, and Stake is the amount wagered (usually normalized to $1). If EV is positive, the bet is mathematically worth taking over a large sample. If it is negative, you are donating to the house.

That is the whole formula. What the calculator-heavy corners of the internet skip is the harder question: how do you get an accurate P_win estimate, one that actually differs from what the market implies? That is where the edge lives, and that is what this post is about.

What the Expected Value Sports Betting Formula Is Actually Measuring

Every sportsbook line carries an implied probability. A moneyline of −110 implies the book thinks the outcome has a 52.4% chance of happening (before devigging). A +200 underdog implies 33.3%. These are not neutral estimates: they include the vig, the house's built-in margin that ensures the book profits regardless of the outcome.

So when you run the EV formula, you are comparing two probability estimates:

1. The market's implied probability: the probability baked into the price you are being offered
2. Your assessed probability: what you actually think the true win probability is

If your estimate is higher than the market's, the bet has positive expected value. If it is lower, you have negative EV. The size of the gap times the payout is your expected edge per dollar wagered.

This is why understanding odds vs. probability is a prerequisite for applying the EV formula correctly. You cannot know whether you have an edge until you translate the American odds into an implied probability you can compare to your own estimate.

How to Calculate Expected Value in Sports Betting: MLB Example

Let's use a concrete example with real numbers from GhostLeg's intel system, including its unglamorous conclusion.

On August 4, 2026, the Toronto Blue Jays visited the Houston Astros. Pinnacle (the sharpest price-setter in the market) implied the Astros at 59.8% to win (post-vig). That put the Blue Jays as the underdog, offered at +120. The raw implied probability of a Blue Jays win at +120: 100 / (120 + 100) = 45.5%, and that figure still contains the book's margin on that side of the line.

Here is what the three main pricing venues showed for the Astros' win probability, side by side (each venue's Blue Jays number is simply the other side of the same two-way market: 41.0% on Kalshi, 42.5% on Polymarket):

Our pre-game intel read verbatim: "Toronto Blue Jays at Houston Astros: Sportsbooks, Kalshi, and Polymarket are all aligned within 2pp — efficient market."

Prediction markets operate with lower vig (1-2% vs. the standard 4-5% sportsbook hold), so their implied probabilities tend to be cleaner estimates of true probability. Both prediction markets priced the Blue Jays a touch higher than the sportsbook's devigged 40.2%, but well below the 45.5% the +120 price implies at face value. Most of that gap between 40-42.5% and 45.5% is not edge. It is the vig share baked into the offered price.

Here is the EV calculation at the sportsbook's offered price:

• Market odds offered: +120 (Blue Jays moneyline)
• Raw implied win probability at that price: 45.5% (= 100/220)
• Prediction-market estimate for the Blue Jays: 41.0-42.5% (Kalshi/Polymarket)
• Payout per $1 wagered: $1.20 profit on a win

Using the more generous prediction-market estimate of 42.5% as the probability anchor:

EV = (0.425 × $1.20) − (0.575 × $1.00)
EV = $0.510 − $0.575
EV = −$0.065

Negative six and a half cents per dollar. That is the honest math: when the venues agree with each other, the underdog price's apparent cushion is mostly the house's margin, and the formula says so. In a market our own intel labeled efficient, there was no +EV bet here at the offered price. Cross-market divergence under 2-3pp is a "watchlist" signal, not a "fire the cannons" signal.

Now the epilogue, because it teaches the most important lesson in this post: GhostLeg's signal stack still leaned Blue Jays on this game (composite score 56, top pick at +120), and the Blue Jays won 3-1. A negative-EV price cashed. One result like that proves nothing in either direction; the formula's verdict and a single outcome live on different timescales, which is exactly what the variance section below is about. The formula's output also depends entirely on which P_win estimate you trust: a bettor whose model genuinely had the Blue Jays above 45.5% saw a +EV bet where the prediction markets saw none.

This is what the GhostLeg MLB intel dashboard and model intelligence surface are doing continuously: converting the spread between what the sportsbook implies and what cross-market data suggests into a concrete EV picture for each game. You can pull these numbers programmatically through the Data API for your own models.

The Blue Jays example illustrates the core insight: the formula is five symbols; the work is in sourcing a probability estimate worth trusting and knowing when the market has left a gap worth acting on.

What Our Signals Show at Scale

One game is an anecdote. The published academic literature on sports betting efficiency, including Pinnacle's own research series, shows that even sharp bettors typically sustain win rates of 53–55% at −110 over thousands of bets. Beating that threshold requires either better data, better discipline, or both. Most recreational bettors never clear 50%.

That benchmark is useful context for what the EV formula is actually up against. A 1-2pp edge in win probability sounds modest. At scale (hundreds of bets, correct Kelly sizing) it compounds into meaningful expected return. At single-bet scale, it is invisible in the noise.

What the signal data does show is that certain conditions correlate with markets being more reliably mispriced: Pinnacle-led line movement (indicating informed sharp money), persistent cross-market divergence (sportsbook vs. prediction markets), and high composite scores across multiple signal types. You can access the underlying signals powering this analysis through the GhostLeg intel dashboard or pull the raw market data programmatically via the Data API.

Why Most Bets Are Negative EV by Default

Before you can find +EV, you need to understand why the baseline is negative. This goes back to the vig, a structure the house installs in every line.

A standard -110/-110 line on both sides means each team's implied probability is 52.4%. Added together: 104.8%. The extra 4.8% is the overround: the mathematical certainty that over a large enough sample, the book profits. To have positive EV on a −110 bet, your true win probability needs to exceed 52.4%. To break even over time, you need to be right more often than the market's implied probability says you should be.

This is why same-game parlays and standard parlays tend to be deeply negative EV; each leg compounds the house margin, so a 4-leg parlay at -110/-110 prices gives the book roughly 18% hold before the first result is settled.

The only path to positive EV is a systematic edge in probability estimation. You need to identify spots where the market's implied probability is wrong. Wrong in a direction that benefits you.

Where Positive EV Comes From

There are a few documented sources of +EV in sports betting markets:

Sharp-money flow (reverse line movement). When the public is hammering one side but the line moves the other way, sharp money is on the minority side. The market-making books trust sharp money more than public volume. Reverse line movement does not guarantee a +EV bet, but it is one of the more reliable signals that the market's implied probability on the public side is too high.

Cross-market divergence. When sportsbook pricing diverges from prediction market pricing (Kalshi, Polymarket), the gap often reflects a market inefficiency. Prediction markets carry lower vig than sportsbooks, with a 1-2% fee structure vs. the standard 4-5% sportsbook hold, which means their implied probabilities are often cleaner estimates of true probability. GhostLeg tracks this divergence live across the intel surfaces.

Line velocity and steam. When a Pinnacle line moves sharply in a short window, what the industry calls a "steam move," it almost always reflects a large, informed bet. Catching the line before it fully settles is a common way sharp bettors find temporary +EV.

Model-identified inefficiencies. Quantitative models that integrate multiple data streams (form, rest, travel, market microstructure) can produce probability estimates that diverge meaningfully from the market. The GhostLeg model runs these estimates continuously; the output shows up as the composite score visible in the model intelligence dashboard and available to API consumers through the Data API for programmatic integration.

Player props as a +EV surface. Prop bets, individual player stat markets, are one of the more reliably mispriced surfaces in retail sportsbooks. Books set hundreds of props per game and cannot price every one as efficiently as a moneyline. When a player's recent form diverges meaningfully from their posted line, and sharp money confirms the discrepancy, the EV formula can find genuine edges that the moneyline market has already arbitraged away.

EV and Closing Line Value Are Cousins

If you have read the closing line value explainer, you already understand the empirical version of this. CLV is what happens when you measure whether the price you got at bet time was better or worse than where the market settled at game time. If you consistently beat the closing line, it is strong evidence that your probability estimates are systematically better than the market's, which is the same thing as saying your bets have positive expected value.

EV is the forward-looking calculation: "does this bet have positive expected return?" CLV is the backward-looking measurement: "did it?" They are asking the same question from opposite ends of the timeline. Most successful bettors use both: EV to decide what to bet, CLV to measure whether their process is actually working.

What the Formula Cannot Tell You

Expected value math is impeccable over large samples. It says nothing about any individual bet.

A bet with a genuine +EV edge still loses whenever the unlikely outcome lands. You can run 20 consecutive +EV bets and post a losing record; that is not a failed bet strategy, that is variance. Profitable sports betting is a long-game problem, and the EV formula is a tool for long-game thinking.

This is why bankroll management is the natural companion to EV math. The Kelly criterion uses your EV estimate to tell you exactly how much of your bankroll to allocate to any given bet. Bet too large and you go broke before the edge materializes. Bet too small and you leave returns on the table. EV gives you the edge. Kelly sizing turns that edge into a sustainable strategy.

The other caveat: your EV calculation is only as good as your probability estimate. If your P_win is wrong, the formula gives you a confident wrong answer. This is why the aggregate data matters. You cannot validate a probability estimate on one bet or twenty bets. You need hundreds of graded outcomes, tracked over time, across sports, with honest accounting for sample size, before you can trust that your estimates beat the market.

That is the infrastructure GhostLeg has built: graded legs, model calibration, cross-market signals, and closing-line tracking all in one place. The intel surfaces track it live at the dashboard. The underlying data feeds are available for API integration at /data-api for anyone building their own models or research workflows.

The Only Formula That Matters in Sports Betting

Every professional betting operation on the planet runs on the expected value betting formula. It is not a trick or a shortcut; ignoring it is simply a choice to bet blind. Most bettors avoid running it honestly because it requires admitting you do not know P_win with any certainty. The market's implied probability is usually a pretty good starting estimate.

The edge comes from the exceptions. From the spots where the market is systematically mispricing an outcome because of public bias, late-breaking information, or microstructure dynamics the books have not fully reflected yet. Finding those spots consistently, and sizing correctly when you do, separates recreational betting from something approaching a sustainable process.

The formula is five symbols. The edge is in the inputs.

For entertainment purposes only. Past performance does not indicate future results.