Which repeats support reuse?
Check the circuit version and each result before relying on a mean.
The question
A fictional team needs three repeats with circuit build B and the required scoring method. Each result must lie within the inclusive band 0.42 to 0.48.
| Build label | Estimate | |
|---|---|---|
| Run 1 | B | 0.45 |
| Run 2 | B | 0.47 |
| Run 3 | B | 0.46 |
The team reports a simple mean for context. The reuse rule checks every eligible estimate.
“Does the evidence meet the stated reuse rule?”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: What records are required?
What would you do here? Open the move you’d make.
Require only the mean to be within the band
Tempting, but it misses something. The rule checks each of the three eligible estimates.
Keep any three values regardless of build
Tempting, but it misses something. Only runs for build B with the required scoring are eligible.
Check the build and scoring record for each estimate
The strong move. The labels and the per-run limits both matter.
See how a strong candidate says it
“I need three records with build B and the same scoring method. Every eligible estimate must be in the band.”
Why it works: The mean is a summary. The rule applies to each run.
Step 2: What do the summaries show?
What would you do here? Open the move you’d make.
The mean of 0.46 alone establishes reuse
Tempting, but it misses something. The rule also checks individual estimates and build identity.
All three are in band, pending record checks
The strong move. The values pass the limit, and the run labels still need confirmation.
The value 0.47 exceeds the upper bound
Tempting, but it misses something. The upper bound is 0.48.
See how a strong candidate says it
“The three listed values are in band, and their mean is 0.46. I’d verify the run records before accepting them.”
Why it works: The summaries satisfy the numeric rule if their labels and scoring are correct.
Step 3: Does the record qualify now?
The raw run record maps Run 2 to earlier build A. A replacement run with build B and the required scoring method gives 0.51. The other two runs match.
What would you do here? Open the move you’d make.
Hold reuse. Only two of three eligible estimates pass.
The strong move. The replacement is eligible and exceeds the upper bound.
Accept because the new mean is about 0.4733
Tempting, but it misses something. The mean cannot substitute for the three required in-band estimates.
Keep the original 0.47 and discard the replacement
Tempting, but it misses something. The 0.47 estimate came from an ineligible build.
See how a strong candidate says it
“Keep the two valid original runs and the replacement. Their simple mean is about 0.4733, but only two of three estimates are in band. The replacement at 0.51 is 0.03 above the upper bound, so the reuse rule fails.”
Why it works: The older build is ineligible. The new estimate must count even though the mean still fits.
Step 4: What would you hand off?
What would you do here? Open the move you’d make.
Drop the replacement as an outlier
Tempting, but it misses something. The rule requires it to count unless there is an independent reason it is ineligible.
Widen the band to 0.51 after seeing the result
Tempting, but it misses something. Changing the limit after seeing the result cannot show that the original rule passed.
Report the failed rule and preserve the run records
The strong move. This shows what changed and keeps the outcome checkable.
See how a strong candidate says it
“I’d keep Run 2 in the audit trail and leave it out of the build B comparison. I’d report the failed rule with the replacement estimate and ask the test owner to investigate.”
Why it works: A later reviewer can see which records counted and which one missed the limit.
The answer, in short
- Decision
- Hold reuse. Two of three eligible runs meet the per-run rule.
- Because
- The replacement uses the right build and exceeds the upper bound, even though the simple mean is in band.
- What would change it
- Keep the declared rule and ask for an investigation of the out-of-band result.
Common traps
- Treating a summary label as verified provenance
- Replacing a per-run rule with a favorable mean
Check your answer
- Which records match the build?
- Does each eligible value meet the limit?
- Why is the mean insufficient?
- What would the test owner need to investigate?
Last reviewed Nov 11, 2026