Correlated parlays: why the chance of both matters
When outcomes depend on the same game conditions, multiplying individual probabilities can misstate the chance of the combined ticket. The missing quantity is a conditional probability.
Use the conditional multiplication rule
The chance of A and B is P(A) × P(B given A). Independence is the special case in which P(B given A) equals P(B). A shared game script, player role or weather condition can make these quantities differ.
A statement that two legs are “positively correlated” is not enough to price the ticket. You need the size of the dependence at the exact thresholds, as well as the offered combined payout.
Build a coherent hypothetical joint table
| A result / B result | B wins | B loses | Total |
|---|---|---|---|
| A wins | 36% | 24% | 60% |
| A loses | 14% | 26% | 40% |
| Total | 50% | 50% | 100% |
A wins 60% of the time and B wins 50%. Their independent product would be 30%, but the table’s joint chance is 36%. Here P(B given A) is 36% ÷ 60% = 60%, so the conditional rule gives 60% × 60% = 36%.
The same marginals can support very different tickets
Keeping A at 60% and B at 50%, another coherent table can put only 20% on both winning, 40% on A alone, 30% on B alone and 10% on neither. The individual chances are unchanged while the joint chance falls from 36% to 20%.
This is why two single-leg forecasts do not identify a multi-leg probability. For these marginals, the joint probability can mathematically range from 10% to 50%. Those bounds are not a useful point estimate by themselves.
Combine a justified joint chance with the actual quote
At an illustrative combined decimal price of 3.00 and a $10 stake, using a 36% joint chance gives EV of $10 × (0.36 × 3 − 1) = +$0.80. Using 20% gives −$4.00. Neither input is a verified estimate for a real ticket.
The linked EV tool can perform this final binary calculation if you supply an independently justified joint probability. It does not infer correlation, build the joint table or handle push/void branches. Enter the actual combined quote, not a product of singles that the operator does not offer.
Try the numbers yourself.
Open the calculator with this guide’s example inputs. Change one assumption at a time and see what changes.
Open expected value calculatorCommon questions
Does positive correlation always make a parlay good value?
No. The offered payout may already reflect the dependence or be less favorable than your justified joint estimate requires.
Can simulation frequency be treated as a verified probability?
Not automatically. Simulation inherits its inputs and dependence assumptions. Calibration and validation are separate requirements.
Sources and example scope
The conditional-probability rule is standard probability theory. The joint tables and EV cases are original hypothetical examples, with no claimed sports edge.