Replay & Calibration
SYNTHETIC EPISODES Invented 2025 episodes — no historical track record is claimed
As-of date
2025-05-012025-01-012026-01-31
Evidence with available-at after this date is removed from the model input entirely; outcomes stay hidden until each resolution date.
Episodes at as-of
| Episode | Evidence | Model P | Market mid | Outcome |
|---|---|---|---|---|
| SYNTHETIC episode 01 · Macro | 4 / all | 53.2% | 69.0% | YES |
| SYNTHETIC episode 02 · Energy | 4 / all | 58.5% | 70.0% | YES |
| SYNTHETIC episode 03 · Digital assets | 4 / all | 42.3% | 29.0% | NO |
| SYNTHETIC episode 04 · Regulatory | 4 / all | 55.6% | 71.0% | hidden |
| SYNTHETIC episode 05 · Macro | 3 / 1 hidden | 50.0% | 56.0% | hidden |
| SYNTHETIC episode 06 · Energy | 2 / 2 hidden | 47.3% | 52.0% | hidden |
| SYNTHETIC episode 07 · Digital assets | 0 / 4 hidden | 50.0% | 44.0% | hidden |
| SYNTHETIC episode 08 · Regulatory(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 09 · Macro(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 10 · Energy(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 11 · Digital assets(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 12 · Regulatory(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 13 · Macro(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 14 · Energy(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 15 · Digital assets(not started) | 0 / 4 hidden | — | n/a | hidden |
| SYNTHETIC episode 16 · Regulatory(not started) | 0 / 4 hidden | — | n/a | hidden |
Selected episode
SYNTHETIC episode 01 · Macro
Model P
53.2%
Mid
69.0%
bid 0.67 / ask 0.71
Resolves
YES
- Synthetic signal 1 · confirmed · avail 2025-01-14 · g1−0.63
- Synthetic signal 2 · confirmed · avail 2025-01-28 · g1−0.31
- Synthetic signal 3 · confirmed · avail 2025-02-11+0.44
- Synthetic signal 4 · confirmed · avail 2025-02-25+0.63
Synthetic demonstration evaluation
NOT OUT-OF-SAMPLEResolved n
3
insufficient for inference
Brier (model)
0.1438
lower is better · 0.25 = coin flip
Log loss (clip 1%)
0.4714
0.693 = coin flip
Brier (benchmark)
0.1171
on 3 with benchmark · model 0.1438
Fewer than 10 resolved forecasts — the chart is shown for illustration, not as evidence of skill.
Perfect calibration: observed ≈ mean forecast in each bin. Bars = count per bin.
Your frozen forecasts
No resolved forecasts yet — scores appear only after resolution.
Trading P&L is not a substitute for calibration. Genuine validation requires forecasts frozen before outcomes were knowable, on real events.
No forecasts frozen yet. Freeze one above or record a paper trade.