Singles, Accumulators and Ticket Risk
Compare six ticket case studies, equal budgets, correlated builders and shared exposure.
Decision processes, not winning systems
Understand the trade-off behind every ticket.
A useful strategy specifies the evidence, exact market, acceptable price, exposure and review method before the result is known. Work through the cases, then use the lab to see how a decision behaves when a leg loses.
- 01Compare a decision
- 02Build a ticket
- 03Test a failure
- 04Review the record
18+ learning material. Every case study, price and probability in these workshops is fictional unless explicitly labelled otherwise. These exercises do not place bets, reserve account funds, or establish that a strategy is profitable. Set affordable time and money limits; taking no bet is a valid decision.
Six decision case studies
Start with the question, not the payout.
Research-led singles
You assess Fixture A’s home win at 55%. The available price is 1.80. You like the team, but the price needs a 55.56% chance just to break even.
Reason through the decision
Record “pass at 1.80” rather than taking a shorter price because the selection sounds convincing. At 1.95, the same unaltered 55% estimate would imply +7.25% net return per stake before costs; at 1.80 it implies −1%.
Stress-test the assumption
The 55% estimate may be wrong. A later line-up change also invalidates the assumption that the estimate remained unchanged. Keep a timestamp and an evidence note; do not reinterpret the old forecast after the result.
Read the source: Smarkets: expected-value calculation and uncertain estimates ↗
Interactive workshop · fictional fixtures
Build it. Break it. See what settles.
Keep the total budget fixed, change the ticket structure, then make one leg lose or become void. The numbers show exactly which combinations survive. This is an arithmetic sandbox, not an account or bookmaker connection.
One bet combines all selected fixtures. Every active leg must win. A losing leg loses this ticket; a void leg is removed at odds 1.00 in this simplified exercise.
- Number of separate bets
- 1
- Stake per bet
- £6.00
- Scenario returns so far
- £0.00
- Final scenario net
- -£6.00
| Combination | Stake | All-win odds | Assumed all-win chance | Scenario settlement | Returned |
|---|---|---|---|---|---|
| A + B + C | £6.00 | 6.237 | 16.50% | Lost | £0.00 |
Total committed in this exercise: £6.00. If all selections win at the entered odds, total return is £37.42, including stake. A return below £6.00 is still a net loss. Calculations keep full precision; displayed pennies are rounded. A real operator may round stakes and returns differently.
Same selections. Same budget. Different exposure.
| Structure | Bets | All-win return | Return in your scenario | Final net | Pre-result model EV |
|---|---|---|---|---|---|
| Separate singles | 3 | £11.10 | £7.80 | £1.80 | £0.06 |
| One accumulator | 1 | £37.42 | £0.00 | -£6.00 | £0.17 |
| All doubles system | 3 | £20.43 | £7.56 | £1.56 | £0.12 |
| Trixie: doubles + treble | 4 | £24.68 | £5.67 | -£0.33 | £0.13 |
Assumed chances and pre-result model EV use your entered probabilities and the independence assumption, before any outcomes are known. They exclude fees and assume binary win/loss settlement with no voids. Changing a scenario outcome affects its settlement, not that original estimate. The EV column is not a probability of profit or a predicted payout.
Why are probabilities not being added together?
For an all-win treble, the fictional independent-event assumption gives 0.55 × 0.50 × 0.60 = 16.5%. This is not the chance that a system returns a profit. Different combinations overlap, so adding their win probabilities double-counts shared outcomes. Different fixtures also need not be independent: related team news, tournament incentives or model errors can affect several selections.
The outcome switches illustrate a scenario; they are not a simulation of future results. Voids here remove that leg at 1.00. Same-match builders, promotional terms and abandoned events can have different operator rules.
Read the source: bet365: Trixie structure, three doubles and one treble ↗
Same-match workshop
Three conditions can describe only one event.
A builder joins conditions from the same match. Check their logic before thinking about the payout. Some overlap, some contradict one another, and some add no extra requirement. Multiplying the displayed single-market odds does not price that relationship.
- First condition
- Satisfied
- Second condition
- Satisfied
- Combined result
- Both satisfied
- Illustrative return
- £6.40
The entered quote is fictional and unverified. An actual supported builder requires the bookmaker's combined price and product-specific rules; a void condition can affect an entire builder differently from an ordinary accumulator.
Read the source: Betfair: same-match builders and product rules ↗
Probability workshop
Watch the chance shrink as the legs grow.
Imagine each leg has the same true chance and every leg is independent. This deliberately simplified model isolates what adding requirements does. A lower combined chance does not, by itself, establish bad value: you must compare it with the available combined price and the uncertainty in the estimates.
42.19% chance every leg wins
(75.0% as a decimal) raised to 3. Under these assumptions, the losing-ticket chance is 57.81%. At eight legs and 75% each, only about 10% of tickets land; that does not mean every tenth ticket wins.
Exposure workshop
A staking rule cannot repair the wrong price.
A budget is a limit on what you can lose, not a target you must spend. In these exercises, the sizes are arbitrary teaching inputs. No fixed percentage is a universally safe stake, and an evaluation’s virtual-account rules are not personal betting advice.
| Illustration | After five losses | What it reveals |
|---|---|---|
| Flat £2 from a £100 teaching balance | £90 remains; £10 lost | The next £2 is now 2.22% of the remaining balance, rather than 2%. |
| 2% of remaining balance each time | About £90.39 remains; £9.61 lost | Stakes decrease with the balance. That changes exposure, not forecast quality. |
| Double a £2 stake after each loss | Stakes £2, £4, £8, £16, £32; £62 lost | The next requested stake is £64. It exceeds the £38 remaining. |
Why we do not turn Kelly into an automatic staking recommendation
For a binary bet at decimal odds d with a perfectly known chance p, the full-Kelly fraction is (p × d − 1) / (d − 1). At 2.00 with a true 55% chance that gives 10%. If the real chance is 48%, the premise of a positive edge disappears. Model error, correlated tickets, liquidity and finite samples make applying that fraction much less straightforward. Reducing the fraction lowers exposure but cannot make a false probability estimate true.
Stop when your predetermined time or money limit is reached. Increasing stakes to get back to a previous balance changes the amount exposed, not the chance of the next result. If control becomes difficult, use limits, timeouts or self-exclusion rather than another staking formula.
Read the source: Betfair Automation Hub: staking methods and assumptions ↗
Read the source: Gambling Commission: loss chasing and other harm indicators ↗
Post-match review
Judge the decision and the outcome separately.
The same process can produce a winning and a losing ticket. Keep enough detail to find which part failed: the underlying forecast, the accepted price, the ticket construction, the stake, or settlement. A result without a pre-match record makes those explanations easy to invent later.
| Record before kickoff | Add after settlement | Question it answers |
|---|---|---|
| Fixture, market, line, period, timestamp | Official result and operator settlement | Did the exact forecast win? |
| Accepted odds, stake, expected probability range | Return, fees and net profit | Was the ticket profitable? |
| Team news, source links, uncertainty, pass criteria | Which evidence was missing or changed? | Was the reasoning defensible at the time? |
| Other tickets with the same team or leg | Combined daily exposure and drawdown | Did one assumption dominate the account? |
A simple forward test fixes rules before a new period starts. Keep every eligible selection, including misses, no-pick decisions and prices you could actually obtain. Test a proposed change on a later period; repeatedly searching past results for a winning league, weekday or price band can produce a pattern that disappears on new data. Shared fixtures also reduce how independent your apparent sample really is.