A-01 v1.2 · Understanding probability and the house edge
In use · lockedThese are the twenty questions exactly as the Q2 2026 cohort sat them, in the order they were shown. This is the instrument the before-and-after comparison is built on, so it is kept here word for word rather than rebuilt from a description.
This version is locked — everything on this page is read-only
There is no Edit, no reorder, no add and no remove on a locked version, for anyone. 64 people sat baseline on v1.2 and 23 of them have not sat endline yet; changing a single word now would put those 23 on a different ruler from the one they started on. Rewording goes into the v1.3 draft instead, and v1.3 only ever applies to a new cohort.
Version record
A-01 · v1.2| Written by | Sam Okafor · content admin |
|---|---|
| Signed off by | Dr Alice Brennan · clinical reviewer · 24 March 2026 |
| Locked by | Priya Raman · impact analyst · 31 March 2026, 09:14 |
| Locked for | Q2 2026 · 64 people |
| Replaces | v1.1 |
| Sittings recorded | 105 · 64 baseline, 41 endline |
| Unlocks | Never. A version used to collect data stays locked permanently, including for the system administrator. |
Assessments pay no tokens
A baseline or endline set carries no reward field at all — not set to zero, absent. If sitting one paid tokens, people would sit it repeatedly for the tokens and the measurement would be worthless.
There are no badges, no ranking, and no countdown on any of the twenty questions below. It is a measurement, not a game.
Why the correct answers are visible here
The author and the reviewer have to be able to check them. They are never sent to the device while somebody is sitting the assessment, and the person sitting it is not shown what they got right afterwards — telling them at baseline would teach the answers and inflate the endline score.
The 20 questions, in the order they were shown
Two of the rows below are queued for rewording in the v1.3 draft. That changes nothing here: v1.2 keeps the wording the Q2 cohort actually saw.
| # | Code | Question and answer options | What it measures | Correct | Marks | Against v1.1 |
|---|---|---|---|---|---|---|
| 1 | B-01 | A market is priced at 2.00. What probability does that price imply? A. 20% · B. 50% · C. 100% · D. It depends on the sport |
Odds and probability | B · 50% | 1 | Unchanged |
| 2 | B-02 | Odds of 1.75 imply roughly what chance of that outcome happening? A. 17.5% · B. 43% · C. 57% · D. 75% |
Odds and probability | C · 57.1% | 1 | Unchanged |
| 3 | B-03 | Odds of 1.85 correspond to roughly what implied probability? A. 18.5% · B. 46% · C. 54% · D. 85% |
Odds and probability | C · 54% | 1 | Unchanged |
| 4 | B-04 | Two sides of the same match are priced at 2.10 and 1.75. Add the two implied
probabilities together. What is the total? A. Exactly 100% · B. About 105% · C. About 95% · D. It cannot be worked out |
Recognising the margin | B · 104.8% | 1 | Reworded |
| 5 | B-05 | In the previous question the total came to more than 100%. What is the extra part
called? A. A rounding error · B. The displayed overround · C. A government tax on the stake · D. A bonus returned to winners Queued for rewording in v1.3 |
Recognising the margin | B | 1 | Unchanged |
| 6 | B-06 | You stake 100 over and over on a market with a house edge of 4.5%. On average, how
much of each 100 comes back? A. 104.50 · B. 100 · C. 95.50 · D. It depends on how well you pick |
House edge as a rate of loss | C · 95.50 | 1 | Unchanged |
| 7 | B-07 | What is the house edge, and where does it sit in the odds you are shown? A. A fee added when you withdraw · B. Built into the price itself, before any bet is placed · C. A charge taken only from winning bets · D. A rate set by the government |
Recognising the house edge | B | 1 | Unchanged |
| 8 | B-08 | Which of these changes the house edge that applies to one person's bets? A. Betting more often · B. Betting larger amounts · C. Studying the sport closely · D. None of these |
House edge is fixed, not earned down | D | 1 | Unchanged |
| 9 | B-09 | A displayed price implies 47.6%, while normalising all prices in that market gives
a 45.5% estimate. What does the gap represent? A. The person betting · B. The operator · C. Neither side · D. Whoever bets first |
Margin embedded in the displayed price; the normalised figure is not a true-chance forecast | B | 1 | Unchanged |
| 10 | B-10 | A coin has come up heads five times in a row. What is the chance of heads on the
sixth toss? A. Less than half · B. Exactly half · C. More than half · D. It cannot be worked out |
Independent events | B | 1 | Unchanged |
| 11 | B-11 | After nine reds in a row, how do the chances of black on the tenth spin
change? A. They rise · B. They fall · C. They do not change · D. They depend on the table |
Gambler's fallacy | C · No change | 1 | Unchanged |
| 12 | B-12 | A gaming machine has not paid out all evening. What does that tell you about the
next spin? A. A payout is due · B. Nothing at all · C. The machine is faulty · D. The odds have improved |
Gambler's fallacy | B | 1 | Reworded |
| 13 | B-13 | “I have a staking system that beats the odds over time.” What is wrong with
that? A. Nothing, if it is followed strictly · B. It needs a larger starting balance · C. No staking pattern changes the price built into each bet · D. It only works on some sports |
Staking systems | C | 1 | Unchanged |
| 14 | B-14 | If you double your stake after every loss, what happens to your long-run
expectation? A. It turns positive · B. It stays negative · C. It becomes zero · D. It depends on the length of the streak |
Chasing losses and negative expectation | B | 1 | Unchanged |
| 15 | B-15 | Someone is down 500 and stakes 500 more to “get back to even”. Which statement is
true? A. The chance of getting back is better than usual · B. The 500 already lost has no effect on the price of the next bet · C. Odds improve after a losing run · D. Losses even out over time |
Chasing losses | B | 1 | Unchanged |
| 16 | B-16 | Over a long series of bets on a market carrying a margin, what happens to the total
staked? A. It grows slowly · B. It stays about level · C. It shrinks at a steady rate · D. It moves at random with no direction Queued for rewording in v1.3 |
Long-run expectation | C | 1 | Unchanged |
| 17 | B-17 | Which of these best describes “expected value”? A. The most likely single result · B. The average result per bet over the long run · C. The largest win available · D. The amount the operator pays out |
Long-run expectation | B | 1 | Unchanged |
| 18 | B-18 | A “risk-free bet” refunds your stake as a bonus that must be turned over five times
before it can be withdrawn. What is the effect? A. The bet genuinely carries no risk · B. The refund is worth less than the cash that was staked · C. The odds improve · D. The margin disappears |
Reading a promotion | B | 1 | Reworded |
| 19 | B-19 | An offer promises “guaranteed profits” from following tips. What does that tell
you? A. The tipster has information others do not · B. Nothing can be guaranteed against a priced market, so treat it as a warning sign · C. It is legal, so it is safe · D. It works on small stakes only |
Reading a promotion | B | 1 | Unchanged |
| 20 | B-20 | What is the only reliable way to reduce what a priced market costs someone over
time? A. Betting at shorter odds · B. Betting more often to spread the risk · C. Staking less, or not at all · D. Waiting for a losing run to end |
Applying the maths | C | 1 | Unchanged |
What changed between v1.1 and v1.2
v1.1 was an internal trial in Q1 2026 with 12 sittings and was never used for a reported figure, so these three rewordings could be made before the instrument was locked. Once 64 people had sat v1.2, that window closed.
| Code | v1.1 wording | v1.2 wording | Why it was changed |
|---|---|---|---|
| B-04 | “Two sides of the same match are priced at 2.10 and 1.75. Added together, is the total more or less than 100%?” | “Two sides of the same match are priced at 2.10 and 1.75. Add the two implied probabilities together. What is the total?” | A two-way choice was answered correctly by half the trial group without any working. The four-option form measures whether they can actually do the sum. |
| B-12 | “A poker machine at your local club has not paid out all evening…” | “A gaming machine has not paid out all evening…” | Nothing in tibbi names a real venue. The construct being measured is independence between spins, and the venue detail added nothing to it. |
| B-18 | “A ‘risk-free bet’ refunds your stake as a bonus with a 5× turnover requirement. What is the effect?” | “A ‘risk-free bet’ refunds your stake as a bonus that must be turned over five times before it can be withdrawn. What is the effect?” | A third of the trial group did not recognise “turnover requirement”, so the question was measuring vocabulary rather than understanding. |
Any of these three changes would be refused on v1.2 today
Rewording changes how hard a question is, even when the subject is untouched. That is why the same edits, applied now, would have to go into the v1.3 draft and wait for the next cohort. There is no exception for a typo — a typo is still a change to what the question asks.