Kelly Criterion | Betting Strategy Analytics

Kelly Criterion

Stake a calculated fraction of bankroll from edge and odds.

StakingMedium riskBacktestableRule v1 (kelly engine v1)

How it works

Uses f = (b·p − q) / b with optional fractional Kelly. Requires decimal odds and a user-entered win probability in MVP (TODO: confirm probability source).

Canonical rules

Plain-language summary of the published rule version (same snapshot stored on every backtest).

Stake fraction f = (b·p − q) / b with fractional Kelly scaling. If f ≤ 0, no bet. Requires odds and a user-entered win probability (TODO(PRD-§13.2)).

Push / void: Push returns the stake and does not change progression; the next stake equals the current stake.

View stored rule JSON

{
  "rule": "Stake fraction f = (b·p − q) / b with fractional Kelly scaling. If f ≤ 0, no bet. Requires odds and a user-entered win probability (TODO(PRD-§13.2)).",
  "pushRule": "Push returns the stake and does not change progression; the next stake equals the current stake."
}

Default parameters

{
  "unit": 1,
  "bankroll": 100,
  "winProbability": 0.55,
  "fractionalKelly": 1
}

Risk profile

If you lost 5 bets in a row from a fresh cycle (ignoring caps):

Total staked0

Next stake0

Pros

Mathematically grounded sizing
Skips negative-edge bets

Cons

Sensitive to probability estimate
Requires odds and bankroll

Mini-simulator

Enter a W / L / P sequence to see stakes using the same engine code as backtests.

Outcome sequence

Use W (win), L (loss), or P (push), separated by spaces or commas.

Decimal odds

Step	Outcome	Stake
1	LOSS	10
2	LOSS	10
3	WIN	10
4	LOSS	10
5	WIN	10
6	WIN	10

Past results don't predict future results. No staking strategy overcomes the bookmaker margin in the long run.