What implied probability really is
Implied probability is the win rate hidden inside a sportsbook price. When a book offers -150 on the Lakers, the number is not just a payout instruction; it is a probability statement. -150 means "we think the Lakers win 60% of the time, and we are willing to charge a small commission to take action on that view." The implied probability is that 60%. It is the only number that allows comparison across formats, books, and markets — the price tag in a currency every bettor can read.
Every quantitative betting decision begins by converting the displayed odds to implied probability. From there, the bettor compares to their own model, to other books' implied probabilities, and to the sharpest no-vig benchmark in the market. The conversion is mechanical; the interpretation is where edges live or die.
The conversion formulas
Three odds formats, three formulas that produce the same number. American odds: for negatives, implied = (–odds) / (–odds + 100); for positives, implied = 100 / (odds + 100). Decimal odds: implied = 1 / decimal. Fractional odds X/Y: implied = Y / (X + Y). A -150 American is 1.667 decimal is 4/6 fractional — all three yield 60.0% implied probability. The American convention is the only one where the math changes based on sign; decimal and fractional are continuous. This is why most pricing software, models, and academic literature use decimal: one formula, no edge cases.
The reverse conversion — probability back to odds — is the same arithmetic in reverse. A 60% probability converts to 1 / 0.60 = 1.667 decimal, which is -150 American (because 1.667 - 1 = 0.667, and 1 / 0.667 = 1.50, indicating a 1.5-to-1 risk-to-win ratio, or -150). Bettors with a probability estimate use this reverse conversion to identify what price would be a fair break-even bet on their own forecast.
The overround — why prices sum to more than 100%
A two-way market with no vig would sum to exactly 100% implied probability. A real sportsbook market never does. -110/-110 sums to 52.4% + 52.4% = 104.76%. The 4.76% excess is the overround — the structural margin the book builds into every market. On heavy favorites the overround can hide further: a -500/+350 line implies 83.3% + 22.2% = 105.5%, larger than the headline because asymmetric markets carry more risk premium for the book.
The overround is the book's commission, but it is not always 4.55% (the canonical -110/-110 figure). Pinnacle runs NFL spreads at 2.4-3.5% overround. DraftKings and FanDuel run the same market at 4.3-5.2%. Same-game parlays carry 15-30%. Futures and prop markets can run 130-140% (one side of the market carries 30-40% margin built in). The implied probability formula treats all of these markets identically; the bettor must remember that the displayed implied probability includes whatever margin the book has chosen to charge.
Stripping the vig — three methods, three answers
To recover the book's underlying probability estimate, you remove the overround. There are three standard methods, and they disagree on lopsided lines. Take a +130 / -150 market — the prices imply 43.5% and 60.0%, total 103.5%.
| Method | Underdog (+130) | Favorite (-150) | Best on |
|---|---|---|---|
| Multiplicative (standard) | 42.0% | 58.0% | Close-to-even markets, -300 to +300 |
| Shin (1992) | 42.3% | 57.7% | Asymmetric markets, heavy favorites |
| Power method (Halicioglu) | 41.8% | 58.2% | Extreme lines, longshots and heavy favorites |
On this moderately lopsided market the three methods agree within half a percentage point. On a -800 / +550 market, the same three methods produce true probabilities for the underdog of 13.4% (multiplicative), 11.8% (Shin), and 12.6% (power) — a 1.6-point spread that translates to meaningful EV differences across volume. Sharp bettors rarely use the multiplicative method on heavy favorites; they default to Shin or power because the underlying assumption (proportional redistribution) breaks down when the action is one-sided.
Why the three methods disagree
The multiplicative method assumes that the book's margin is added uniformly to both sides — that a 3.5% overround is 1.75% on each side regardless of price. This is not how bookmakers actually price. Shin's model (Hyun Song Shin, 1992) assumes a fraction of bettors have inside information and the book widens the favorite side more than the underdog side to defend against informed action. The power method finds the exponent that forces implied probabilities to sum to 1, which empirically tracks closing-line behavior on long-tail markets better than the alternatives. None of the three is "correct" — they are different theoretical assumptions about how bookmakers actually load their margins, and the right method depends on the market structure.
For retail bettors, the practical approach is to use multiplicative on standard two-way markets (-300 to +300), and switch to Shin or power for futures, props, longshots, and heavy favorites where the multiplicative method is known to mis-estimate the true probability.
Implied probability across market types
Two-way moneylines and totals are the cleanest application of implied probability — strip the vig, get the no-vig number, compare to your model. Three-way markets (soccer 1X2, hockey 60-minute lines including draws) require summing three implied probabilities and dividing each by the total — the math is identical but the inputs are three numbers. Props with binary outcomes (Will player X score a touchdown?) use the same two-way framework. Futures and outrights with 20-30 possible winners require the power method or a more sophisticated normalization because the multiplicative method spreads the margin disproportionately across the longshots.
The most common retail mistake is reading the implied probability of one side of a prop and treating it as the book's estimate. A +180 prop on a touchdown scorer implies 35.7% — but the same book might price the opposite (no touchdown) at -250, implying 71.4%. Sum: 107.1%. The 7.1% overround is hidden because most retail bettors only look at one side of the prop screen. The no-vig price for the scorer is 33.3%, not 35.7% — a meaningful 2.4-point difference that can flip an apparent edge into a losing bet.
Calibration — checking whether the implied probability is honest
A well-calibrated market is one where prices implying 30% win 30% of the time, prices implying 70% win 70% of the time, and so on across the curve. Academic studies of Pinnacle closing lines consistently show calibration within 1-2 percentage points across the entire probability range — the closing line is the most accurate public probability estimate available. Retail US books are less well-calibrated, particularly on lopsided markets where public action distorts pricing. The bettor's edge against retail books is precisely this calibration gap.
The standard calibration test bins outcomes by implied probability (0-5%, 5-10%, etc.) and compares the expected win rate of each bin to the actual win rate observed across a season. Sharp models produce calibration plots that hug the 45-degree diagonal; recreational systems show systematic miscalibration on favorites or longshots. The same diagnostic applied to a bettor's own forecasts reveals whether their model is over-confident, under-confident, or directionally biased.
Favorite-longshot bias in implied probability
The favorite-longshot bias is the single most-documented inefficiency in betting markets. Across horse racing (where it was first measured), sportsbook markets, and even financial prediction markets, prices on longshots systematically over-represent their true probability, and prices on favorites slightly under-represent theirs. A horse priced at 100-to-1 (implied 1.0%) historically wins about 0.6-0.8% of the time. A -300 favorite (implied 75%) wins closer to 77%. The bias is small but consistent and amounts to a structural retail edge that has survived decades of academic publication, presumably because the bias is rooted in human psychology that markets cannot fully arbitrage away.
The practical implication: implied probability on extreme prices is not the same as fair probability. A naive no-vig strip on a +1500 longshot still overstates the true win chance. Bettors who model these markets correctly bet against the longshot side, knowing the true probability is even lower than the no-vig number suggests. The opposite is true on heavy favorites — sharps bet the heavy chalk more often than the public realizes, because the favorite is slightly under-priced relative to truth.
Using implied probability to find the true line
The workflow that turns implied probability from theory into edge has four steps. One: pick a sharp reference book — Pinnacle for international, Circa or Bookmaker for US. Two: convert the sharp book's two-sided prices to implied probability, sum, and divide each by the total. This is the no-vig fair line. Three: convert recreational US book prices on the same market to implied probability. Four: any recreational price whose implied probability is meaningfully below the no-vig benchmark is a positive-EV bet — by the amount of the gap, minus an adjustment for whether the sharp book's line has moved since you sampled it.
This is the entire foundation of low-variance retail betting. It does not require a proprietary model; it requires only line-shopping discipline, accurate conversion, and patience. The edges are small per bet (1-3% on the average mispricing) but they compound across volume. The bettors who beat the closing line consistently are almost always doing this exact workflow, not running secret simulations.
Practical checklist for using implied probability
- Convert every price to implied probability before you compare anything. The number on the screen is meaningless until it is a percentage.
- Always strip the vig. A standalone implied probability is the book's price; the no-vig is the book's estimate of truth.
- Use multiplicative on close markets, Shin or power on extremes. The method matters when the line is lopsided.
- Anchor on the sharpest book. Pinnacle or Circa no-vig is your true line; recreational books are the trading surface.
- Account for favorite-longshot bias at the tails. No-vig on extreme prices is still optimistic on longshots.
- Calibrate your own forecasts. If your model says 70% but those bets only win 60%, your forecasts — not the market — are wrong.
Sources & further reading
- Shin, Hyun Song. "Prices of state-contingent claims with insider traders, and the favorite-longshot bias." Economic Journal, 1992 — the foundational paper for the Shin vig-stripping method.
- Snowberg, Erik & Wolfers, Justin. "Explaining the favorite-longshot bias: Is it risk-love or misperceptions?" Journal of Political Economy, 2010 — the modern empirical treatment of FLB.
- Štrumbelj, Erik. "On determining probability forecasts from betting odds." International Journal of Forecasting, 2014 — comparison of multiplicative, Shin, and power methods across a large dataset.
- Levitt, Steven D. "Why are gambling markets organised so differently from financial markets?" Economic Journal, 2004 — bookmaker pricing structure and its implications for implied probability.
- Pinnacle Betting Resources — public documentation on margin calculation, no-vig price construction, and the Pinnacle closing line as a probability benchmark.