Normal Distribution explained for betting

In one sentence

The normal distribution is the symmetric bell curve described by an average and a spread, useful whenever a quantity is the sum of many small independent pieces.

How it works

Add up lots of small, independent contributions and the total tends to form a bell shape. That is the central limit theorem, and it is why the normal distribution appears everywhere. A season's profit is many bets added together; a basketball points total is many possessions; a cricket innings is many balls.

The curve needs just two numbers: the mean μ (the centre) and the standard deviation σ (how wide it is). About 68% of outcomes land within one σ of the mean, and about 95% within two.

For betting, you estimate μ and σ, then read off the chance of finishing above or below a line. For evaluation, it tells you the likely range of profit after n bets.

The maths

f(x) = \frac{1}{\sigma\sqrt{2\pi}}\, e^{-\frac{(x-\mu)^2}{2\sigma^2}}

z = \frac{x - \mu}{\sigma}, \qquad P(X > x) = 1 - \Phi(z)

x is the value you care about, such as a profit target or a totals line.
μ is the expected value and σ the standard deviation.
z is how many standard deviations x sits from the mean.
Φ is the standard normal cumulative probability, found in any stats library.

In words: convert the line to a z-score and look up how much of the curve sits beyond it.

Worked betting example

Football, Over/Under 2.5 goals on Betfair, illustrative figures. You plan 400 bets of £10 over a season at an average price of 2.00, and each bet truly wins 53% of the time. Commission is 2%, so a winner pays £9.80.

Mean per bet: μ = 0.53 × £9.80 − 0.47 × £10 = £0.494.
Spread per bet: σ = £19.80 × √(0.53 × 0.47) = £9.882.
Season: mean = 400 × £0.494 = £197.60, standard deviation = £9.882 × √400 = £197.64.
P(season in loss): z = (0 − 197.60) ÷ 197.64 ≈ −1.00, so about 15.9%.
P(season profit above £500): z = (500 − 197.60) ÷ 197.64 ≈ 1.53, so about 6.3%.
A 95% range for the season is about −£190 to +£585.

The 53% is the input that matters. At 52% the season mean falls to £118.40 and the chance of a losing season rises to about 27.5%.

Where it's good

Totals and margins in high-scoring sports: cricket innings, basketball, NFL, rugby.
Estimating the likely range of profit over a large number of bets.
Approximating the binomial or Poisson when counts are large.
Quick z-score checks on whether a result is unusual.

Limitations and pitfalls

Sport outcomes are often skewed and bounded. Innings totals have a long left tail from collapses; the normal treats both sides equally.
Low-count sports such as football are discrete and lumpy; use Poisson, not normal.
Tails are thin. Extreme events (a collapse, a blowout, a big losing week) happen more often than the normal suggests.
Key numbers matter in some sports (NFL margins of 3 and 7), which a smooth curve ignores.
σ is as important as μ and harder to estimate; underestimate it and you will overbet the favourite side of every line.

How to build it

scipy.stats.norm (cdf, sf, ppf) in Python; pnorm and qnorm in R.
Data: your bet log (stakes, prices, results), or historical totals by venue and conditions; residuals from a regression give σ.
Tip: fit μ with a regression (venue, teams, conditions) and use the residual spread as σ, rather than the raw spread of all totals.

Central limit theorem: why the bell curve appears.
Log-normal: for quantities that cannot go negative and are skewed.
Skellam: margins in low-scoring sports.
Value at risk: using the spread of profits for risk.