Benchmarks
Predicted vs. observed. Every revealed cohort.
We publish the calibration of the Living Outcome Oracle every quarter. The predictions are the LOO posterior at cohort start - locked at OEA-V1 commit time. The observed values are the revealed cohort means. This is the page that exists only because Peptide MD operates the human-use endpoint layer.
Predicted vs. observed · |Δ| absolute effecty = ŷ · x = y
revealed cohort ideal y = x
Brier-style MSE per bin
| |Δ| bin | n | MSE | RMSE |
|---|---|---|---|
| [0, 5) | 1 | 0.410 | 0.640 |
| [5, 20) | 1 | 0.029 | 0.170 |
Lower MSE means a better-calibrated posterior in that effect-size band. We publish the underlying CSV at /api/export/cohorts.
Peptide leaderboard · revealed |Δ| sum
| # | Peptide | Cohorts contributing | Σ |Δ| revealed | Posterior mean | Grade |
|---|---|---|---|---|---|
| 1 | Semaglutide | 1 | 14.9 | 29.7% | Grade B |
| 2 | Selank | 1 | 3.4 | 54.5% | Grade B |