fixaiprompt
All templates
⚡ Advanced Techniques·Works on: claude, chatgpt, gemini

Tournament Pairwise Ranking

Technique: Pairwise Tournament

Don't ask 'which is best?'. Ask all N×(N−1)/2 pairwise comparisons. Far more reliable.

Advanced#ranking#tournament#comparison
0/3
Fill the template
3 placeholders left.
Live preview
Use Tournament Pairwise Ranking. Do NOT rank everything at once — pairwise is more reliable than absolute rating.

What we're ranking:


Candidates:


Quality criteria (what "better" means):


Protocol:

**Round 1 — Pairwise duels.**
For each unique pair (A vs B), state:
- Which wins under the quality criteria.
- The single sharpest reason.

**Round 2 — Aggregate.**
Count wins per candidate. Produce the ranking by win count.
For any tied candidates, run a tiebreaker duel using only the strongest criterion.

**Round 3 — Sanity check.**
- Does the #1 candidate beat the #N candidate by a large margin or a small one?
- What's the closest call (smallest win margin)?
- Is there a candidate that ALMOST won — would shifting one criterion flip the result?

Final output: ranked list with score, plus the one observation a smart reader of this ranking should know.
Inputs0 of 3
›See the lazy version this template replaces
Before — the lazy prompt
Which of these is best: <list>

Why it works

  • Pairwise comparison is robust to ordering bias and easier for the model to reason about than absolute scoring.
  • Counting wins surfaces upsets — sometimes the obvious favorite loses head-to-head to a quieter candidate.
  • The 'closest call' analysis catches near-ties that an absolute rating would have hidden.

Make this one yours

Replace the bracketed placeholders, then paste into the Prompt Fixer to lint your customisation before hitting send.

Open Prompt FixerBrowse more templates