Effective bets calculator
Correlated bets behave like fewer independent bets than you placed. If several bets ride on the same game or the same player, a bad night hits them together, so they should be sized as one larger position. This page counts how many bets you are really making.
Count your own effective bets
Enter how many bets you hold open and the average correlation between them. You get how many independent bets that adds up to.
Stake spread across correlated bets is not spread. Effective bets is the count a book is really holding: n divided by one plus n minus one times rho, where rho is the average pairwise correlation between positions. Independent bets give n_eff equal to n. Everything else gives less.
The correlation used as a starting point here is a prior, not a measurement: rho of 0.15. Beside it, 0.45 is what a synthetic ledger built at that level measures back: an example, since no real book was measured. Same bet counts, two columns.
| Bets | Effective at prior rho 0.15 | Effective at example rho 0.45 |
|---|---|---|
| 2 | 1.7 | 1.4 |
| 4 | 2.8 | 1.7 |
| 8 | 3.9 | 1.9 |
| 12 | 4.5 | 2.0 |
| 20 | 5.2 | 2.1 |
Read the example column down. From 2 bets to 20 bets, the effective count crawls from 1.4 to 2.1. Past a handful of correlated positions, adding another buys no diversification. It only adds stake.
That is what a staking rule has to see. A max single of 5% on a bankroll of 20000 caps a lone bet at 1000. The same rule, applied to a book already loaded with correlated exposure, caps it at 409.96. Competitors print the first figure and call the portfolio diversified. They are counting tickets, and tickets are not independent.
One row, worked
Take 8 bets with an average pairwise correlation of 0.45. The bottom of the fraction is one plus 7 times 0.45, which is 4.15, and 8 over 4.15 is 1.93. Eight tickets carry the risk of about 2 independent bets. At the prior of 0.15 the same 8 come to 3.90.
The correlation does the work. At zero, every bet counts in full. As it climbs towards one, any number of bets behaves like a single bet, because they all win or lose together.
Where correlation comes from
Several bets in one game: a side, the total and a player's points all move with how the game goes.
One team across several markets, or several games that share a cause, such as weather on a day of outdoor totals.
Not from an arbitrage. The two legs of an arbitrage are opposite sides of one result, which is what makes the profit fixed rather than risky.
When to count effective bets
Before adding one more position to a slate that is already loaded on one game or one team.
When sizing with Kelly. Kelly assumes each bet stands alone; on correlated bets the stakes add up to more risk than the fraction meant.
When reading a record. A month of correlated wins is fewer independent results than the ticket count says, so it proves less.
Free. No account and no subscription. Needs macOS 14 or later on a Mac with Apple silicon (M1 or later). 34.9 MB, notarised by Apple.
Where the number comes from
The worked example was computed by the same engine that prices bets, at the moment this page was built. None of it was typed in. The build fails if a figure appears here and not in the engine's own output. The calculator runs that engine, release 0.1.62, in your browser, and the build checks that it returns this example’s answers for this example’s inputs.