Expected Value, Explained: How to Actually Use EV in Betting
EV is the most name-dropped and least understood number in sports betting. Most people use it as a fancy word for "a bet I like." It isn't that. It's a specific comparison between your probability and the probability the sportsbook's price implies, and once you can calculate it you stop grading yourself on the one thing that never mattered: whether last night's bet cashed.
Your W-L Record Is a Terrible Scoreboard
You spend an hour on a matchup, bet a hitter's total bases, and he goes 0-for-4. Was the bet wrong?
You genuinely cannot tell from that. A hitter with a legitimately great matchup still fails most nights, because most plate appearances end in an out for everyone who has ever played. If you grade your handicapping by whether the last ticket cashed, you'll abandon good process after bad variance and double down on bad process after good variance. That's how bettors end up worse in October than they were in April, with a full season of "experience."
Expected value fixes this. It grades the bet, not the result — and you can measure it the moment you place the wager, before a pitch is thrown.
EV in One Sentence
Expected value is the average amount you'd win or lose per bet if you could place that exact bet an infinite number of times.
Roll a fair die and get paid the number that comes up. Any one roll pays 1 through 6, but the expected value is 3.5 — an amount no single roll ever pays. That's the whole idea. EV describes the bet, not the night. Any individual result is just one sample from a distribution.
Every Price Is a Probability in Disguise
When a book posts odds, it is publishing a probability estimate in an awkward unit. -150 means "this happens about 60% of the time." +120 means "about 45%." The conversion is mechanical:
| American odds | Implied probability |
|---|---|
| -200 | 66.7% |
| -150 | 60.0% |
| -110 | 52.4% |
| +100 | 50.0% |
| +120 | 45.5% |
| +180 | 35.7% |
So you and the sportsbook are making the same kind of claim about the same event. That makes you directly comparable, and the comparison is where every dollar of long-term profit comes from.
Kill the Vig First, or Your Edge Is Fake
Convert both sides of a market into implied probabilities and add them up. They won't total 100%. They'll total 104%, 107%, sometimes worse. That surplus is the vig, and it's the house's cut baked into both prices.
This matters enormously, because it means raw implied probabilities overstate both sides. If you compare your 55% projection to a raw implied 53.5% and declare a 1.5-point edge, you may have no edge at all — you were measuring the juice.
The market: over -115 (53.5% implied), under -105 (51.2% implied). Total: 104.7%. The extra 4.7% is the vig.
De-vig: divide each by the total. Fair over = 53.5 ÷ 104.7 = 51.1%. Fair under = 48.9%. Now they sum to 100%, and 51.1% is what the book actually thinks.
Your model: 55.0%. That's a real 3.9-point edge over the no-vig fair price.
The EV: at -115 you risk $1 to win $0.87. EV = (0.55 × 0.87) − (0.45 × 1) = 0.478 − 0.45 = +2.8% EV. Betting $100 into that spot is worth about $2.83 every time you fire it.
Notice how modest that is. A near-4-point probability edge — which is a genuinely good spot — turns into under 3% EV once you pay -115. This is why line shopping matters so much: the same 55% projection at -105 instead of -115 is +7.4% EV — more than double. You didn't handicap any better. You just stopped donating the difference.
+EV Bets Lose. Constantly. Forever.
Here's where people lose faith. That 55% bet above is +EV and it also loses 45% of the time. Four straight losers at 55% happens about 4% of the time, which across a season of hundreds of bets means it will happen to you over and over.
A losing week tells you close to nothing about whether your bets were good. You need hundreds of wagers before results say anything, and in the meantime the only honest feedback loop is whether your process is sound.
The mental model: a +EV bet is a slightly weighted coin, not a prediction. Nobody evaluates a weighted coin from four flips. The bettors who survive are the ones who can sit through a 3-9 stretch without touching a process that's working.
+EV Is Not the Same as Picking Winners
Two different skills get confused constantly:
- Picking winners maximizes hit rate. You do that by hammering heavy favorites at prices that already account for them — and you can go 70% and still lose money.
- Finding +EV maximizes the gap between your number and the book's. It often points at underdogs that lose more than they win.
A bet your model puts at 40% when the no-vig market says 30% is a big edge, and you'll still lose it three times in five. Anyone judging you on record thinks you're bad at this. The bankroll disagrees. Mixing those two scoreboards is the single most common reason people abandon a profitable approach.
How to Actually Use It
Set a minimum EV threshold
Your projection has error bars and so does the market's. Sub-1% EV is inside the noise of both, and chasing it floods your card with false positives. Most solid bettors require 3% EV or better to fire. Anything above 15% should make you audit your own inputs before it makes you excited — an edge that large usually means stale data, a wrong line, or news you haven't seen.
Bet flat units
EV math assumes consistent sizing. If you bet 1 unit on your small edges and 5 on the ones that "feel" right, the feelings — not the math — determine your season. Pick a unit at 1–2% of bankroll and hold it. Variance at 55% is brutal enough without letting stake size amplify it.
Track closing line value, not your record
The single best short-term proof that your bets are +EV is CLV: did you beat the closing line? If you're taking -110 on numbers that consistently close at -125, the market is agreeing with you after the fact, and profit follows even if this month's results don't. Log the line you took, the no-vig fair price, your EV, and the close. That record tells you the truth years before your W-L does.
Be careful with thin markets
A 9% edge against one obscure book with low limits is far weaker evidence than 4% against a sharp consensus. Thin markets carry lazy prices, and a huge edge against a single number nobody updated usually means the number is stale, not that you found gold.
What EV Doesn't Tell You
- It inherits every flaw in your model. EV is arithmetic on top of a projection. Bad projection, confidently wrong EV.
- It says nothing about tonight. It's a long-run average, and the long run is longer than most bankrolls' patience.
- It ignores correlation. Five +EV bets that all need the same game to go over aren't five edges. They're one edge wearing five hats — and a parlay of them is far riskier than the individual EVs suggest.
- It goes stale fast. An edge calculated against a price from two hours ago may be gone. Lines move for reasons, and the reason is usually information you don't have yet.
How PropPrizm Calculates It
The model produces a win probability for every player prop, and the board builds a no-vig consensus fair value from the live sportsbook prices — Kalshi included, since it's a real liquid market, with pick'em apps excluded because a symmetric pick'em de-vigs to a useless 50%. EV% is then computed against the actual price at the book you're shopping, so the number reflects what you'd really be laying.
Every projection gets graded once the game finalizes, and those grades feed the calibration record: when the model said 60%, how often did it hit? That check is what keeps the EV numbers worth reading instead of just worth printing.
Quick rule: if you can't state the no-vig fair probability, you can't say a bet is +EV. "I love this spot" with no number attached isn't handicapping — it's a feeling with vocabulary.
The Takeaway
EV isn't a rating of how likely a bet is to win. It's a measurement of disagreement between your number and the market's, priced in dollars. It will be right in aggregate and wrong constantly in the moment, which is exactly what a probability is supposed to do.
De-vig before you compare, demand a real threshold, size flat, and track CLV. The bettors who get paid from this concept are the ones who stopped asking whether the last one hit.
Want to see it live? Open any matchup on the dashboard and compare the model's probability to the no-vig consensus next to it — that gap, priced against the odds you're actually getting, is your EV. PropPrizm is for informational and entertainment purposes only and does not guarantee outcomes. Please bet responsibly. If you or someone you know has a gambling problem, call 1-800-GAMBLER.