Poisson Calculator: Model Prop Bets and Counting Markets
The Poisson distribution models how often a count event occurs given a known average rate. Soccer goals, hockey goals, strikeouts, and three-point attempts all fit reasonably well. Enter an expected average and a target count and the calculator returns the probability of hitting exactly, at least, or at most that number, plus fair American odds for the at-most outcome.
What makes Poisson the right tool for counting bets
Poisson works when events are independent, occur at a roughly stable average rate, and are discrete counts. Soccer goals fit: each shot sequence is somewhat independent, teams maintain consistent average goal rates over large samples, and goals come in whole numbers.
Strong fits: soccer goals per game, hockey goals per game, MLB strikeouts per start, NBA three-point attempts per game, NFL field goal attempts per game. Each has a stable lambda and discrete outcomes.
Weak fits: NFL total points (game script creates clustering), basketball point totals (foul trouble and pace swings cause high variance), any event where outcomes are strongly correlated within a sequence.
How lambda drives the probability outputs
Lambda is the expected average count for the specific window you are modeling. A player who averages 2.7 three-point attempts per game has lambda = 2.7 for a full-game prop. Lambda for the first half would be roughly 1.35.
The calculator returns three probabilities: exactly k, at least k, and at most k, where k is your Proposition input. Exactly is the probability of that precise count. At least covers all counts equal to or above k. At most covers all counts equal to or below k.
The fair American odds output uses the at-most probability and converts it to the equivalent American line. This gives you a vig-free benchmark for the under side of a prop.
- Exactly k: Poisson mass function at the specific count
- At least k: 1 minus cumulative probability at k minus 1
- At most k: cumulative probability through k
Step by step: using this calculator
Lambda should match the exact game window, not a raw season average applied uncritically. A pitcher averaging 7 strikeouts per start is a reasonable lambda for a full-game prop, but adjust downward for high-contact lineups or pitch count limits.
The Proposition must be a whole number. For a 3+ prop, enter 3 and read At Least. For an under-4.5 total, enter 4 and read At Most.
- Enter Expected Average - the estimated mean count for the window.
- Enter Proposition - the target count as a whole number.
- Read the Exactly, At Least, and At Most percentages.
- Read At Most (Fair Odds) for the under-side American equivalent.
Worked example: player averaging 2.7 threes, prop at 4+
Lambda = 2.7, Proposition = 4. P(X >= 4) = 1 - P(X <= 3). P(0) = 6.7%, P(1) = 18.1%, P(2) = 24.5%, P(3) = 22.0%. P(X <= 3) = 71.3%. P(X >= 4) = 28.7%. Fair American for the over: roughly +249.
If the sportsbook posts this prop at +280, the implied probability is 26.3% versus the model's 28.7%. EV on $100 at +280: (0.287 x $280) - (0.713 x $100) = $80.36 - $71.30 = +$9.06. That is solid edge.
Cross-check: if the book has -130 on the same 4+ prop, implied probability is 56.5% versus model 28.7%. That is dramatically -EV. The difference between +280 and -130 on the same prop is the kind of pricing gap that separates sharp books from recreational ones.
Exactly 4
17.2%
Over 3.5
28.6%
Fair odds
+250
Poisson distribution by lambda
Slide lambda - the player or team average - to watch the probability mass shift.
When Poisson breaks down
Football is the worst fit. NFL scoring comes in lumps and game script creates strong clustering - if a team falls behind by 20 in the second half, they pass more and score more garbage-time points. Independence assumption is violated.
Player props affected by usage changes are also problematic. A center forward's shot count depends on minutes played, opponent scheme, and game state. If any of these shift from the lambda estimate, the model breaks.
Over-dispersion is the technical signal. If historical variance for the event is significantly higher than the mean, Poisson underestimates tail probabilities. Soccer total goals in leagues with high-variance teams sometimes show this pattern.
What most calculator pages skip
Most Poisson tools output raw probabilities without odds conversion. OddsGuard shows probability outputs alongside fair American odds in your local format, so you compare the model output directly to the price your book posts without manual conversion steps.