Every bet is a probability claim
A -150 favorite is not a 'lock'. It is a market claim that the team should win roughly 60% of the time. A +200 underdog is a market claim of about 33.3%. Your job as a bettor is to decide whether you disagree - and by how much.
Implied probability is the bridge between price and forecast. Without it, you are just guessing whether a number 'feels right'. Every price on a sportsbook can be reduced to a percentage chance, and every percentage can be reduced back to a price.
From price to probability - the formulas
Three identities cover every odds format. Memorize these and you can think in probabilities everywhere.
- Decimal odds → probability: p = 1 / decimal. Example: 2.00 → 50%; 1.50 → 66.7%; 4.00 → 25%.
- Negative American odds → probability: p = |odds| / (|odds| + 100). Example: -150 → 60%; -110 → 52.38%.
- Positive American odds → probability: p = 100 / (odds + 100). Example: +200 → 33.3%; +150 → 40%.
Implied vs no-vig (true) probability
Sportsbook prices include vig, so the implied probabilities on both sides of a market sum to more than 100%. On a standard -110 / -110 spread, each side implies 52.38%, summing to 104.76%. That 4.76% above 100% is the book's theoretical hold.
To get the no-vig probability the book is implying, divide each side by the sum. 52.38 / 104.76 ≈ 50%. That is roughly the market's estimate of fair value, stripped of margin. It is not your edge - it is the baseline you have to beat.
Why ranges beat point estimates
A serious bettor rarely says 'the Bears will win 58% of the time'. They say 'I think the Bears win somewhere between 55% and 62% of the time, and the market has them at 50%'. That range - and its overlap with the market - is where edges live.
Reasoning in ranges forces honesty about uncertainty. A tight range you can defend with evidence is more valuable than a precise number you cannot. If your range overlaps the no-vig market, you do not have an edge - you have an opinion.
The law of large numbers - and why it betrays you in short samples
Over thousands of bets, results converge to expected value. Over 50 bets, they often do not. A bettor with a 53% true win rate at -110 has a positive edge, but Monte Carlo simulations show roughly a 1-in-4 chance of being underwater after 200 bets. That is not a bug - it is variance doing its job.
This is why outcome judging is so dangerous. A single losing month, or even a losing year, does not falsify a bettor's process. Conversely, a winning week proves nothing. Sample size is the lens through which all betting results must be read.
Independent vs correlated probabilities
Single-bet probabilities multiply only if the events are independent. The probability of two independent 60% bets both winning is 0.60 × 0.60 = 36%. Three is 21.6%. This is why parlays look attractive but pay structurally less than the multiplied price suggests, especially with vig stacked into each leg.
Same-game parlays add a second wrinkle: legs from the same game are usually correlated (a team's moneyline and over often move together). Sportsbooks reprice SGPs to remove the correlation premium - typically taking far more vig than a normal parlay would.
Bayesian updating - how new information should change your number
Most bettors update too aggressively on small samples and not enough on structural news. Bayes' rule formalizes the correct weighting: posterior odds = prior odds × likelihood ratio. A team's preseason rating (prior) should move slowly after one game and quickly after a season-ending injury to a star.
A useful heuristic: weight a prior by its effective sample size. A 1,000-possession prior shifts only modestly after a 100-possession update - the prior carries 10× the weight. This is why analytics models that 'regress to the mean' early in a season outperform raw season-to-date numbers.
- Prior - your best estimate before the new data (preseason rating, market no-vig price, long-run base rate).
- Likelihood - how probable the new evidence is under each hypothesis (team is good vs team is average).
- Posterior - your updated estimate after combining prior and likelihood, weighted by their respective sample sizes.
- Effective sample size - the larger your prior's sample, the slower your posterior moves on each new data point.
Common probability mistakes that cost bettors money
Five reasoning errors recur in losing betting accounts. Each one has a name in the academic literature and each one is fixable once you can spot it.
- Gambler's fallacy - believing past independent outcomes affect future ones ('the over has hit 5 in a row, due to go under').
- Hot-hand misread - assuming streaks have predictive power in samples too small to distinguish from noise.
- Base rate neglect - ignoring how often an outcome occurs across the league and over-weighting a recent salient game.
- Conjunction fallacy - judging a specific scenario (parlay leg combo) as more likely than its broader category.
- Outcome bias (resulting) - grading a bet by whether it won, not by whether the price was +EV when placed.
Reference table - common American prices and their probabilities
A printable cheat sheet. The middle column is the implied probability (with vig). The right column is the no-vig probability assuming an equal and opposite price on the other side of a -110 / -110 style two-way market.
- -200 → 66.7% implied · ~63.6% no-vig (vs +170)
- -150 → 60.0% implied · ~57.1% no-vig (vs +130)
- -120 → 54.5% implied · ~52.2% no-vig (vs +100)
- -110 → 52.4% implied · 50.0% no-vig (vs -110)
- +100 → 50.0% implied · ~47.6% no-vig (vs -120)
- +150 → 40.0% implied · ~42.9% no-vig (vs -170)
- +200 → 33.3% implied · ~36.4% no-vig (vs -240)
- +300 → 25.0% implied · ~27.3% no-vig (vs -370)
Frequently asked questions
What does implied probability mean in sports betting? It is the break-even win rate the price quotes. For decimal odds it is simply 1 / decimal; for American odds use the formulas above.
How do you calculate the true probability of a bet? You cannot, exactly. You can estimate the no-vig probability the market implies, and you can build your own estimate from data, models, or domain expertise. The bet is +EV when your estimate is meaningfully higher than the no-vig price.
How many bets does it take to know if I have an edge? Statistically, hundreds to thousands. CLV (closing line value) is a much faster signal - consistent CLV is observable in dozens of bets, results in hundreds.
Why don't both sides of a market add up to 100%? The overage is the sportsbook's hold. A perfectly fair, zero-margin two-way market would sum to exactly 100%.