American odds convert to implied probability with one formula: negative odds divide the absolute value by (value + 100); positive odds divide 100 by (value + 100). At -110 you break even at 52.38%; at +150 you break even at 40%. The gap between your model and that number is your edge.
American odds tell you the breakeven win rate baked into every price. Convert them to implied probability, compare to your model, and you have a decision framework that works on any sport, any market, any book.
What is the formula for American odds to implied probability?
Two formulas cover every price you will see:
- Negative odds (-N): implied = N / (N + 100)
- Positive odds (+N): implied = 100 / (N + 100)
At -110: 110 / 210 = 52.38%. At -105: 105 / 205 = 51.22%. At +150: 100 / 250 = 40%. At +200: 100 / 300 = 33.33%. The math is the same whether you are looking at an NFL spread, a college moneyline, or a player prop.
Why does the same price appear on wildly different spreads?
Our live feed shows home spreads from -1.5 to -54.5 this week, all priced at -110 on the favorite side .
Dallas Cowboys -1.5 over New York Yankees and Arkansas-Pine Bluff at Missouri -54.5 both imply 52.38% at -110. The spread is the handicap — the market's estimate of the talent gap. The price is the vig — the book's cut. Converting the price tells you the breakeven rate; the spread tells you what the market thinks the true gap is.
Which live lines show the conversion in action?
NFL: key-number spreads at standard juice
Seattle Seahawks -4.5 over Tennessee Titans . Atlanta Braves -1.5 over Milwaukee Brewers . Both are -110, both imply 52.38%. If your model projects Seattle covers 57%, your edge is 4.62 percentage points. That is a real edge — provided your model is calibrated. Our edge calculator walks through the subtraction.
College football: blowout spreads at the same price
San Jose State at USC -38 and -38.5 . New Mexico State at Florida State -31 and -31.5 . Massachusetts at Rutgers -30.5 . Bethune-Cookman at UCF -42.5 . Eastern Illinois at Minnesota -43.5 . All priced at -110, all implying 52.38%.
The converter gives you the same number for all of them. Your job is to decide whether a 50-point handicap leaves enough signal for your model to beat 52.38%. The sharp money usually says no — variance in garbage time swamps the edge. Our edge guide covers why blowouts are no-play zones.
Competitive college games: where the converter finds edge
North Carolina at TCU -7.5 and -8 . NC State at Virginia -5.5 . Jacksonville State at North Dakota State -7 . These are competitive games where the market has converged but the outcome is uncertain. A half-point disagreement with the consensus at -7.5 creates meaningful edge because the variance is manageable and the hold is standard.
How does the vig hide inside the conversion?
Add the implied probabilities of both sides. At -110 / -110 you get 52.38% + 52.38% = 104.76%. The 4.76% over 100% is the hold. At -105 / -115 you get 51.22% + 53.49% = 104.71% — similar hold, different distribution. Line shopping for -105 instead of -110 saves you 1.16 percentage points of implied probability. That is the easiest edge in betting. Our vig explainer proves it with a two-book comparison.
What is the difference between American, decimal, and fractional odds?
All three formats represent the same payout structure:
- American (-110): risk $110 to win $100 → implied 52.38%
- Decimal (1.909): risk $1 to win $0.909 → implied 1 / 1.909 = 52.38%
- Fractional (10/11): risk 11 to win 10 → implied 11 / (11 + 10) = 52.38%
The implied probability is identical. The format is just notation. Convert once to implied probability and you can compare any book, any format, any market. Our odds reading guide covers all three.
How do I build this into my betting workflow?
- Pull the market price. Use our NFL picks feed or college picks feed — they show consensus line, price, and implied probability for every game.
- Convert to implied probability. Apply the formula above or use the converter in the feed.
- Generate your probability. Build a model in the no-code handbook or the NFL model explainer.
- Subtract implied from yours. That is your edge in percentage points.
- Apply a threshold. Only bet when edge clears your personal hurdle after vig.
- Log the closing line value. Every bet gets a CLV comparison. Average CLV is your report card. The CLV explainer shows why this matters more than win rate.
What are the most common conversion mistakes?
- Forgetting the vig. Converting -110 to 52.38% is correct. Converting -110 to 50% ignores the book's cut and overstates your edge by 2.38 points.
- Using the wrong formula for positive odds. +150 is 100 / 250 = 40%, not 150 / 250 = 60%. The numerator is always 100 for positive odds.
- Converting the spread instead of the price. -7.5 is the handicap. -110 is the price. Only the price converts to implied probability.
- Assuming implied probability equals true probability. It equals breakeven probability. The market's true probability is lower by half the hold. At -110 / -110 the market's true probability is roughly 50% each side; the 2.38% each side is the tax.
What number proves your conversion workflow works?
After a large sample of logged bets, your average closing line value tells the truth. Positive average CLV = winning process. Negative average CLV = losing process. The converter is just arithmetic; CLV is the receipt. Pair the converter with the tracking columns in the bet-tracking guide, the sizing framework in the Kelly criterion explainer, and the hold math in the vig guide. Four tools, one workflow: convert, compare, size, track.
If your average CLV over 200 bets drops below -0.5 points, your process is losing to the market — stop betting until the model recalibrates.
Bet responsibly — set limits, never chase losses.
Breakeven win % at common American odds
The win rate you need to break even at each price. Pick odds shorter than -150 and you must win >60% just to stay flat — a hurdle most casual handicappers never sustain.
Model calibration: predicted vs observed
Predicted win probability bucket vs the empirical win rate inside that bucket on the test set. Points on the y=x reference line are perfectly calibrated; points below mean the model is overconfident in that bucket.
