1.5Repeated trials and "retry until decided" experiments
The idea
Some experiments are repeated until they give a decisive answer. The retries do not matter for who wins eventually; the first decisive round is what counts.
Think of it like this: Flip a coin until it is not a tie: the ties are just do-overs.
Why it exists
These show up as "repeat until" or "wins on the third trial" questions, and they have a short answer once you see the retries cancel.
How it works
- Per round: Y with probability a, N with probability b, retry with probability c = 1 − a − b.
- P(the final result is Y) = a/(a + b): the retries cancel out.
- The round that decides is geometric: P(decided exactly in round k) = c^(k−1) (1 − c).
- For "someone wins on the k-th trial": (retries)^(k−1) × (decisive).
- Check by summing the geometric series: a(1 + c + c² + …) = a/(1 − c).
Worked example: GATE 2016: coin flipped twice, TH gives Y, HH and HT give N, TT retries
a = 1/4, b = 2/4, c = 1/4. P(Y) = (1/4)/(1/4 + 2/4) = 1/3 ≈ 0.33. In the lab, choose the first preset and read "P(final output is Y)".
Now play with it
- →Choose the 2018 preset (two dice, retry on a tie) and read the row "decided in round 3": it is 5/216 ≈ 0.023.
- →Slide the retry probability up: the expected number of rounds grows, but P(Y) stays the same.
| round k | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|
| P(decided exactly in round k) | 0.75 | 0.1875 | 0.0469 | 0.0117 | 0.0029 |
| P(decided with outcome Y in round k) | 0.25 | 0.0625 | 0.0156 | 0.0039 | 0.001 |
Two ways to the same answer: sum the geometric series a(1 + c + c² + …) = a/(1 − c) = a/(a + b), or argue that ties/retries change nothing, so condition on the round being decisive. GATE 2016: (1/4)/(1/4 + 1/2) = 1/3. GATE 2018: the third trial decides, so (1/6)²·(5/6) = 5/216 ≈ 0.023.
Check yourself
Each round: Y with probability 0.2, N with probability 0.3, otherwise retry. What is P(final output is Y)?
Exam notes (the dense version)
If an experiment is repeated until it ends in a decisive result, tie/retry rounds just restart it. With per-round probabilities a (outcome Y), b (outcome N) and c (retry), P(final = Y) = a/(a + b) and the round that decides is geometric: P(decided in round k) = ck−1(1 − c). For "wins on the k-th trial", multiply k − 1 retries by one decisive round.
Previous year questions: Counting-Based Probability & Events (17)
PYQs are grouped by topic, so this is the full set for Counting-Based Probability & Events; try the ones that match this subtopic. Open all 17 on the topic page.