Event workspace
Explain my rank
Published revision 1. Your team's result and the evidence behind it.
This is the frozen published revision. Review labels are local to each project and do not identify judges. Older publications or duel-only rankings may have no rubric review breakdown.
Glass Signal
Raw rank: 24 · Calibrated rank: 18 · Rank movement: 6
Raw mean: 3.4444 · Published score: 3.5508 · Panel grand mean: 3.5661
Panel grand mean in cyan · Contribution above panel baseline in green · Contribution below panel baseline in amber · Published score in indigo with the reported analytic rubric interval. Colors describe calibrated review contributions, not reviewer severity.
| Review | Raw total | Baseline | Scale | Information weight | Shrinkage trust | Calibrated contribution |
|---|---|---|---|---|---|---|
| Review 1 | 2.6667 | 2.6979 | 1.2507 | 1.5642 | 0.6000 | 3.5412 |
| Review 2 | 3.3333 | 3.3336 | 1.0000 | 1.0000 | Scale held at 1 | 3.5659 |
| Review 3 | 4.3333 | 4.3487 | 1.0000 | 1.0000 | Scale held at 1 | 3.5508 |
How the calculation works
Each raw total is the rubric's weighted criterion score. Baseline is the panel grand mean plus the fitted reviewer leniency and the final fit's pending leniency correction. Each calibrated contribution is grand mean + (raw total − baseline) / scale.
The published calibrated score is the information-weighted average of these contributions, with information weight = scale². Values are rounded for display. A positive rank movement means the project rose relative to raw ranking.
Shrinkage trust is n / (n + κ): it blends the fitted scale toward 1 before scale bounds apply. When there are too few observations to estimate a slope, the scale stays at 1. This is uncertainty about reviewer scale, not a penalty to your project. Review labels restart within your project; they do not match judge identities or review labels elsewhere.