Which method gives more usable outputs?
Compare correct outputs from two methods on a known-answer circuit.
The question
A fictional lab wants the larger fraction of submitted runs that return a correct, usable bit. Each scored run gives one bit.
| Scored | Correct | |
|---|---|---|
| Method A | 1000 | 910 |
| Method B | 1000 | 970 |
The table shows the runs that reached scoring.
“Which method would you choose for the lab’s stated goal?”
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 does the table count?
What would you do here? Open the move you’d make.
Treat the scoring counts as submission counts
Tempting, but it misses something. The table does not establish that every submission reached scoring.
Ask how many runs each method received
The strong move. This gives the denominator for the lab’s decision.
Compare the number of bits per run
Tempting, but it misses something. Each scored run gives one bit, so this does not distinguish the methods.
See how a strong candidate says it
“The table gives correct results among scored runs. I need the submission counts to compare usable outputs per submission.”
Why it works: The question counts all submissions. The table labels a narrower group.
Step 2: What can you compare now?
What would you do here? Open the move you’d make.
B already meets the submission goal better
Tempting, but it misses something. The table gives no full submission counts yet.
A leads because it has more incorrect outputs
Tempting, but it misses something. The goal is to get correct usable outputs.
B leads among scored runs, 97% to 91%
The strong move. Both rates use scored runs as the denominator.
See how a strong candidate says it
“A scored 910/1,000 = 91%. B scored 970/1,000 = 97%. B leads by six percentage points among scored runs.”
Why it works: The opening counts support a comparison within the scored runs.
Step 3: Which method meets the goal now?
A scored all 1,000 submissions. B scored 1,000 of 1,250 submissions and discarded the other 250 before scoring.
What would you do here? Open the move you’d make.
Choose A. It yields 91% usable outputs to B’s 77.6%.
The strong move. Both rates now include all submitted runs.
Choose B. It scored 97% to A’s 91%.
Tempting, but it misses something. The 97% rate omits the 250 unscored B submissions.
Choose B. It returned 970 correct bits to A’s 910.
Tempting, but it misses something. B received more submissions. The objective is the fraction that yield a correct output.
See how a strong candidate says it
“A gives 910/1,000 = 91% correct usable outputs per submission. B gives 970/1,250 = 77.6%. I’d choose A for this goal.”
Why it works: B scored 80% of its submissions. Its 97% score among retained runs gives 77.6% usable outputs across all submissions.
Step 4: Which numbers belong in the report?
What would you do here? Open the move you’d make.
Classify all discarded B runs as measured wrong bits
Tempting, but it misses something. Those runs did not produce a scored output.
Report both rates and the number of unscored runs
The strong move. This preserves the evidence about output quality and usable yield.
Claim A gives more usable outputs per minute
Tempting, but it misses something. The packet gives no timing evidence.
See how a strong candidate says it
“I’d give correctness among scored runs and correct usable outputs per submission. The discarded runs had no score. I’d ask for time and cost before comparing throughput.”
Why it works: Both rates matter and they use different denominators.
The answer, in short
- Decision
- Use A for the stated goal of correct outputs per submission.
- Because
- A gives 91% usable outputs and B gives 77.6%, a gap of 13.4 percentage points.
- What would change it
- If the objective changes to cost or throughput, ask for the resources and time per method.
Common traps
- Using scored runs as the full denominator
- Treating discarded runs as measured wrong bits
Check your answer
- Which runs does the opening table count?
- How many runs were submitted?
- What rate answers the question?
- What would establish usable outputs per minute?
Last reviewed Nov 11, 2026