Value Betting: The Only Thing That Matters Long-Term
If you can't articulate where your edge is coming from, you don't have one. Value betting is the discipline of only placing wagers where the price you're getting beats the true probability of the outcome — and ignoring everything else.
1. What value betting actually means
A value bet is a wager where the implied probability of the price you're getting is lower than the true probability of the event occurring. That's it. No mention of who you think will win. No mention of which way the line moved. Just: is the price wrong, in your favor?
Sportsbooks publish prices that imply a probability. A line of –150 implies the team will win 60% of the time. If you genuinely believe that team will win 65% of the time, that's a value bet — the book has mispriced the side relative to your estimate. The 5% gap is your edge.
The reverse is also true. If you love a team and think they'll win, but the book has priced them at –250 (71.4% implied), and you only think they're a 65% favorite, you have a negative-edge bet on a side you think will win. The team is the right pick. The bet is wrong.
You're not picking winners. You're identifying prices the market got wrong. Whether the bet hits is variance — whether the price was wrong is skill.
2. The math: expected value
Expected value (EV) is the average return per dollar wagered if you could repeat the bet infinitely under the same conditions. It's the single number that separates good bets from bad ones, and it has a clean formula:
EV = (P(win) × profit) − (P(lose) × stake)
Take a concrete example. The Bills are –140 home favorites against the Dolphins. You've modeled the game and estimate Buffalo wins 62% of the time. The book's –140 line implies 58.3%.
140 / (140 + 100) = 0.583 → 58.3%
# EV on a $100 bet at -140 if true win % is 62%
(0.62 × $71.43) − (0.38 × $100)
= $44.29 − $38.00
= +$6.29 per $100 → +6.3% EV
That's the bet. You're not guaranteed to win — you'll lose it 38% of the time — but every $100 you put on it earns about $6.29 in long-run expectation. Make 500 of those over the course of a season and the math becomes very loud.
If your model and the book disagree by less than the vig (typically ~2.4% per side at –110/–110), it's almost certainly not a real edge — it's noise in your model. Demand a margin that meaningfully beats the juice.
3. Why winning bets and good bets aren't the same
This is the part recreational bettors never internalize. A losing bet can be a great bet. A winning bet can be a terrible bet. The outcome of any single wager tells you almost nothing about whether the decision was correct.
Consider two bettors in Week 7:
- Bettor A bets a –300 favorite at fair value. The implied probability is 75%; they think the team wins 75%. The favorite wins. Bettor A celebrates. Their bet had 0% EV — they got lucky on a coin flip with no edge.
- Bettor B bets a +180 underdog. The implied probability is 35.7%; they think the dog wins 42%. The dog loses. Bettor B is annoyed. Their bet had +12.7% EV — they made the right decision and ran into variance.
Over 1,000 bets, Bettor B is wildly profitable and Bettor A is breakeven at best (probably worse, once vig is accounted for). The scoreboard at the end of one game is meaningless. The scoreboard at the end of a season starts to matter. The scoreboard across multiple seasons is the only one that reflects skill.
Evaluating your process by your weekly P&L. A great week can hide a terrible process; a brutal week can mask a sharp one. Track your closing line value instead — it's the leading indicator your win/loss record is the lagging indicator of.
4. Where edge actually comes from
If value betting is just "buy when the book is wrong," the obvious question is: why would the book ever be wrong, and how would you know before they do? There are three honest answers.
Better information or a better model
Books set lines using a combination of power ratings, market signal, and adjustments for injuries, weather, rest, and travel. If you have a genuinely better forecasting model — accounting for things the market underweights, like specific O-line/D-line matchups, coaching tendencies in narrow situations, or weather effects on totals — you'll occasionally see prices the market is mispricing. This is the hardest source of edge to build and the easiest to overestimate. Most people who think they have a "better model" are picking up noise.
Market inefficiency on soft books
Not every sportsbook is sharp. Recreational-focused books (the ones with heavy marketing budgets and generous promos) often lag the sharper market by minutes or even hours, especially on midweek and early-week lines. When sharp money has already moved a game from –3 to –3.5 elsewhere, a soft book still sitting at –3 is offering you a price the market itself has already invalidated. That gap is edge.
Line shopping and reduced juice
The most reliable and replicable source of edge isn't predictive at all — it's logistical. By having accounts at multiple books and always taking the best available price, you raise the average price you pay per bet. Going from –110 to –105 on the same side raises your breakeven win rate from 52.4% to 51.2% — a structural edge that doesn't require you to be smarter than the market. See the Line Shopping guide for the math on how this compounds.
5. Sample size and the patience problem
Even with a real edge, variance dominates the short run. To get a sense of how brutal this is, here's roughly what's required before you can statistically distinguish a sharp bettor from a coin flip:
An NFL regular season has 272 games. If you bet every one against the spread, you only generate ~272 bets per year. A bettor with a genuine 3% ROI edge can still finish a 272-bet season down money. It's not unusual — it's expected with that sample size. The math is unforgiving here, and it's why so many bettors with real edge quit before their results materialize, and so many bettors with no edge stay in because they ran hot for a season.
This is also why CLV (closing line value) is so important. It tells you whether your process is sharp now, without waiting two seasons for the win-loss record to confirm it.
6. Applying this on Sunday
Value betting in practice isn't a single skill — it's a workflow. The bettors who do it well share a few habits:
- Generate your own number first. Whether through a model, a power rating, or careful situational analysis, write down what you think the line should be before looking at where the market is. If you check the line first, you're anchored to it and can't see the gap clearly.
- Compare your number to the no-vig market price. The book's posted line includes juice. Strip it out so you're comparing your number to the market's actual implied probability, not its retail price.
- Bet only when the gap meaningfully beats the noise. A 0.5-point disagreement on a spread is within your model's margin of error. A 2-point disagreement, with a clear reason behind it, is worth taking seriously.
- Size proportional to edge. A 3% EV play and a 0.5% EV play should not be the same bet size. Kelly Criterion is one principled way to translate edge into stake — see the Fixed Unit vs Kelly guide for the tradeoffs.
- Track your CLV. Every bet, every week. If your average bet beats the closing line, your process is sharp regardless of your weekly results.
Stop asking "who's going to win?" Start asking "what's the right price, and am I beating it?" Edge is the only thing that compounds. Everything else is entertainment.