Every live leg is required
Under a straight all-to-win accumulator, one losing live leg makes the combined bet lose. A void or push may shorten the multiple, but the current rules decide that state. Betfair's Sportsbook rules provide one operator example.
If four independent legs each have a 0.60 chance:
Joint success probability = 0.60^4 = 0.1296, or 12.96%
Under OpenStax's independence rule, the multiplication applies only to independent events. Even though each illustrative leg is more likely than not, the fourfold succeeds less than one time in eight on the stated model.
A transparent margin example
Suppose four independent events each have a fair probability of 0.60. Fair decimal odds for each would be approximately:
1 / 0.60 = 1.6667
If each leg is instead offered at 1.60, the combined offered price is:
1.60^4 = 6.5536
The raw break-even probability of the accumulator is:
1 / 6.5536 = 0.1526, or about 15.26%
But the illustrative true joint probability remains 12.96%. Expected return per unit staked is:
0.1296 * 6.5536 = 0.8493 units
Illustrative expected net result = 0.8493 - 1 = -0.1507 units, or about -15.07%
OpenStax's expected-value treatment supports probability-weighted payoffs. This example demonstrates compounding under assumed probabilities; it is not an estimate of any real operator or market.
Dependence and model error
OpenStax defines independence as one outcome not changing another's probability. Same-match legs, repeated teams, common weather, competition incentives and shared model inputs can violate that assumption.
Estimation error also compounds. If each marginal probability is slightly too high, their product can be materially too high. Calibration and out-of-sample checks matter more than a winning narrative for each leg.
Diagnostic table
| Symptom | Check |
|---|---|
| One leg repeatedly loses | Review leg-level prices, probabilities and market type |
| Losses cluster around one team or league | Check shared exposure and data quality |
| Same-match combinations look unusually attractive | Compare the operator product price with a joint model |
| Results depend on late voids or non-runners | Archive market-specific settlement rules |
| Stakes rise after losses | Stop the recovery cycle and return to a fixed budget |
Next step
Use How Many Legs Acca for the next part of this topic.
Diagnose the loss without rewriting the forecast
Classify a losing accumulator using information that existed at the decision time:
| Cause | Evidence to inspect |
|---|---|
| Ordinary joint-event failure | Pre-match probability assigned to the exact losing path |
| Price disadvantage | Fair price estimate versus accepted combined price |
| Dependence error | Conditional links omitted from the model |
| Data or lineup error | Snapshot time, availability and source history |
| Settlement difference | Market period, provider, void or participation rule |
| Process breach | Stake, leg count or market outside the written policy |
Do not label every close loss unlucky or every win well analysed. A calibrated low-probability event will lose most individual times, and a poor estimate can still win once. Review probability quality across a defined sample of comparable forecasts, while investigating calculation and settlement defects immediately.
The most useful output is a correction that can be tested: a revised market mapper, a dependence feature, a stricter data cutoff, or a documented decision rule. Adding another leg, changing stake after the result, or excluding inconvenient losses does not explain the original error.
Continue learning
- Next guide: Acca Builder Tools Compared
- Related guide: Accumulator Loyalty Programmes
Assumptions and limitations
The four-leg example assumes equal, independent and known probabilities solely to isolate the arithmetic. Real probabilities are uncertain, margins vary, and dependence can increase or decrease joint probability.
GambleAware
GamCare
Gambling Therapy
GAMSTOP