Expected Value: The Only Betting Concept That Matters

Expected value is the average amount a bet wins or loses if you could place it an infinite number of times. It is the only number that determines whether a betting approach makes money, and it is computed with one line of arithmetic:

EV = (p × profit) − ((1 − p) × stake)

Where p is the true probability of winning, profit is what you collect on a win, and stake is what you lose otherwise. Positive EV means the bet is worth more than it costs. Negative EV means the opposite, no matter how the bet turns out.

A worked example, twice

Take a $100 bet at +150. The profit on a win is $150. Break-even sits at 100 ÷ 250 = 40% — below 40% the bet loses money, above it the bet makes money.

Now vary your estimate of p and watch the value move:

The same bet, at the same price, is worth +$12.50 or −$5.00 depending entirely on a number you supplied. Nothing about the sportsbook changed between those two rows.

That is the shape of the whole problem. The formula is arithmetic a ten-year-old can do. The input is a probability estimate that has to be better than the estimate produced by a market with far more information than you have.

The break-even probability is the dividing line

Every price contains the exact win rate at which EV equals zero. That number is the price’s implied probability, and computing it is covered in full in implied probability explained.

The practical version of EV is one comparison: your number against that column. Everything else is bookkeeping.

Where does p come from?

This is the question that separates people who understand EV from people who can recite it.

There are two honest answers, and one dishonest one that shows up constantly.

The dishonest answer is that p comes from your read on the game. It does not. Your read on the game is an opinion with no calibration attached — you have never checked whether the things you rate at 60% happen 60% of the time, because almost nobody has. Plugging an uncalibrated feeling into a precise formula produces precise nonsense.

The first honest answer is that p comes from the market itself, de-vigged. The no-vig line is a probability estimate produced by everyone with money in the market, priced continuously, and it is very hard to beat on major leagues. If you use the market’s own number as your p, EV comes out slightly negative on every bet — which is the correct answer for most bets most of the time.

The second honest answer is that p comes from a model, or from information the price has not absorbed yet. That is a real path, and it is far narrower than it sounds. A model has to beat the aggregate of every other model plus the vig. Information has to be both true and faster than the market’s reaction, which is a matter of seconds in liquid markets and is precisely what makes line movement worth watching.

How the vig moves the goalposts

A coin flip priced at +100 on both sides would be a zero-EV bet forever. Nobody offers that. The standard price of −110 on both sides means each side must win 52.38% of the time to break even, and the two sides together imply 104.76% of probability.

That surplus is the fee, explained mechanically in what is the vig. Its effect on EV is direct: you are not trying to be right more than half the time, you are trying to be right more than 52.38% of the time, and the extra 2.38 points is a toll on every ticket.

Against a market that is well calibrated, betting at −110 has an expected return of 95.45 cents on the dollar. That is the default outcome of betting without an edge, and it does not require you to be bad at picking games.

Positive EV is not the same as winning

Here is the part that costs people their bankrolls: a real edge produces losing months, and no amount of being right about EV prevents it.

Take a bettor with a genuine 2% edge — a return of $2 per $100 staked, which is a strong long-run result. At −110 that corresponds to winning 53.43% of bets, against a break-even of 52.38%. One extra win per hundred, roughly.

Flat-staking $100 per bet, here is what that looks like:

The standard deviation arithmetic, so you can check it: each bet returns +$90.91 or −$100, a spread of $190.91. With p = 0.5343, the per-bet standard deviation is 190.91 × √(0.5343 × 0.4657) = $95.23. Over n bets that scales by √n, so 100 bets gives 95.23 × 10 = $952 and 1,000 gives 95.23 × 31.62 = $3,011.

Read the table again. After a hundred $100 bets, a genuinely skilled bettor expects to be up $200 and is down about two times in five. After a thousand bets — more than most recreational bettors place in three years — a real edge is still underwater a quarter of the time.

Why “I won, so it was a good bet” is wrong

Results-oriented thinking is judging a decision by its outcome when the outcome was mostly noise. A +900 longshot that lands was probably still a bad bet. A −EV parlay that hit was still −EV, for reasons the parlay math sets out in detail.

The reverse error is more expensive: abandoning a sound approach after a bad month, because the month felt like evidence. Given the standard deviations above, it almost never is.

This is why serious bettors track something other than profit. Measuring the price you got against the price at kickoff — closing line value — gives you a signal about decision quality with far less noise in it, months before your profit-and-loss column means anything at all.

The two honest paths to positive EV

There are exactly two, and they are not equally available.

1. A better estimate than the market’s

Building a model that prices games more accurately than the aggregate market, then betting only where the two disagree by more than the vig. This is a real profession and a real skill.

It is also rare, and it gets rarer every year in major markets. The people doing it are competing against each other, they employ statisticians, and they get their bets down before the number moves. In big liquid markets the honest prior for any individual is that their model is worse than the closing line.

Where it is more achievable is in thin markets — lower leagues, obscure props, anything where the book is pricing from a template rather than from a deep market. Those are also the markets with the lowest limits and the fastest account restrictions, which is not a coincidence.

2. A better price than the one in front of you

This is the accessible one, and it requires no forecasting ability at all.

• Shopping. Two books with different numbers on the same game are two different EVs. Taking +3.5 where another book shows +3, or −105 where another shows −115, moves the break-even rate in your favor by 2.27 percentage points on price alone — every time, with no opinion required.
• Promotions. Risk-free bets, boosts and bonuses have a computable dollar value, and some of them are worth more than the negative EV of the qualifying bet. This is arithmetic, not handicapping, and the terms are where the value gets clawed back.
• Timing. Prices at open are less accurate than prices at close. Getting an early number that the market later moves toward is value, whether or not the bet wins.

None of these require you to know anything about the sport. All of them are limited by how much you can get down before a book decides you are not the customer it wants.

EV per bet vs EV per hour

Professional bettors care about EV per unit of time and capital, not EV per bet. A 4% edge on a $50 maximum stake is $2, and if finding it took forty minutes it was not worth finding.

For recreational bettors the calculation runs the other way, and it is worth being explicit about it. If you are betting for entertainment, your real metric is cost per hour of enjoyment — and by that measure, a small stake on a market you will watch for three hours is cheap, while a large stake on a market you will refresh anxiously is expensive at any EV. The stake size question that follows from this is the subject of bankroll management for sports betting, and it matters more than any single price you will ever get.

The bottom line on EV

If you cannot articulate why you have an edge — the specific thing you know or model that the price does not reflect — then you do not have one, and your expected value is the vig, negative and steady.

That is not an argument against betting. It is an argument against lying to yourself about what betting is. A −EV bet placed knowingly, at a stake you have decided you can lose, is entertainment with a price tag, and the price tag is roughly 4.55% of everything you stake at −110. Buying that knowingly is a defensible choice. Buying it while telling yourself it is an investment is how people end up in trouble.