Tier 4

pairwise_comparison

Instead of rating each option absolutely (hard), compare pairs

Usage in Claude Code: /pairwise_comparison your question here

Pairwise Comparison Matrix

Overview

Instead of rating each option absolutely (hard), compare pairs directly (easier). “Is A better than B?” is easier than “How good is A?” Aggregate comparisons to produce ranking.

Goal

Rank options by comparing all pairs and counting wins. Each individual comparison is simple; aggregation is mechanical. Avoids the difficulty of absolute rating.

Steps

Step 1: List All Options

Enumerate all options to be compared. Label them for easy reference (A, B, C… or 1, 2, 3…).

Output: Labeled option list

Step 2: Define Comparison Criterion

Clearly state what you’re comparing on. Can be single criterion or composite.

Single: “Which is more important?” Composite: “Considering cost, quality, and time, which is better?”

Output: Criterion statement

Step 3: Create Comparison Matrix

Create N×N matrix with options on both axes. Diagonal is empty (don’t compare to self). Only need to fill upper triangle (lower is symmetric).

Output: Empty matrix

Step 4: Compare Each Pair

For each pair (i, j) where i < j: Ask: “Is [Option i] better than [Option j] on [criterion]?”

Record:

  • “1” in cell (i,j) if i wins
  • “0” in cell (i,j) if j wins
  • “0.5” each if tie

Be consistent. If A > B and B > C, then A > C (transitivity).

Output: Filled comparison matrix

Step 5: Check for Inconsistencies

Look for intransitive cycles: A > B > C > A If found, reconsider those comparisons. Some intransitivity may be acceptable (different criteria).

Output: Consistency check result

Step 6: Calculate Scores

For each option, sum its wins. Win = 1 point, Tie = 0.5 points, Loss = 0 points Total possible points = N - 1

Output: Score for each option

Step 7: Produce Ranking

Sort options by score (descending). Handle ties by re-comparing or accepting tie.

Output: Final ranking

When to Use

  • Ranking options when absolute scoring is difficult
  • Prioritizing features/tasks
  • Multi-criteria decision making
  • Reducing bias from anchoring
  • When options are hard to rate but easy to compare

Verification

  • All pairs were compared
  • Criterion was applied consistently
  • No significant intransitivity
  • Scores sum correctly
  • Ranking reflects scores

Input: $ARGUMENTS

Apply this procedure to the input provided.