Understanding Your Ranking Results
What the numbers mean, when to trust rankings, and how to interpret confidence scores.
Overview
After enough comparisons, Prevue produces a statistically ranked list of your images. This article explains what the numbers mean, when to trust the rankings, and how to interpret confidence scores.
Before You Start
- You have a session with completed evaluations
- Your session status is Active or Stable
Steps
Check the session status
Look at the status badge. If it says Stable, the rankings are statistically reliable and unlikely to change significantly. If it says Active, more comparisons may improve confidence.
Understand the ranking score (mu)
Every image gets a mu (μ) score — a number representing its estimated quality relative to other images. Higher mu = higher rank. All images start at 25.0 (equal). After comparisons, winners gain mu and losers lose mu.
Read the confidence score (sigma)
Every image also has a sigma (σ) score measuring uncertainty. All images start at 8.333 (very uncertain). Sigma decreases with more comparisons. Lower sigma = more confident the ranking is accurate.
Check evaluator and comparison counts
The results summary shows total comparisons, distinct evaluators, and catch pair accuracy. More evaluators and higher catch-pair accuracy mean more reliable rankings.
Export your results (Creator+)
Click Export to download CSV (rank, mu, sigma, comparisons), ZIP (images renamed by rank), or PDF reports (Team+). Free plan exports include a watermark.
Troubleshooting
Rankings seem wrong — my best image is ranked low
Symptom: An image you expected to rank highly is in the middle or bottom.
Fix: Rankings reflect evaluator consensus. Check the sigma value — if still high (>5), the image needs more comparisons. If sigma is low and the ranking persists, the data suggests others perceive it differently than you do.
All images have similar scores
Symptom: The mu scores are clustered together (e.g., all between 24 and 26).
Fix: If sigma values are high, continue evaluating — scores spread with more data. If sigma is low and scores are still clustered, your images are genuinely similar in perceived quality.
Session has not reached Stable status
Symptom: The session stays Active despite many comparisons.
Fix: The average sigma has not dropped below the convergence threshold, or the session needs more evaluators. Share publicly or opt into The Current for more diverse evaluators.
Related Articles
Still need help?
If these steps did not resolve your issue, contact support and include screenshots and error messages.
Contact Support