Calibration
Every other screen here makes a claim about your future. This one checks whether those claims have been holding up — by standing at points in your past, forecasting forward from only what was knowable then, and comparing against what actually happened.
Verdict
The bands are overconfident
Reality escapes the outer band more often than it should. The forecast is understating how uncertain it really is, which is the failure mode that actually hurts people. Treat the numbers as directional until this closes.
- Outer band
- 55%
- claims 90%
- Inner band
- 20%
- claims 50%
Accuracy decays with distance, as it should
| Horizon | Outer band held | Inner band held | Band width | Median off by | Leaned |
|---|---|---|---|---|---|
| 14 days | 20% | 20% | 17% of balance | $1.5k | optimistic |
| 30 days | 20% | 20% | 20% of balance | $943 | optimistic |
| 60 days | 80% | 20% | 24% of balance | $926 | optimistic |
| 90 days | 100% | 20% | 28% of balance | $1.1k | optimistic |
Band width is the p05–p95 spread as a share of the starting balance. Narrower is better — but only while coverage holds. Anyone can be perfectly calibrated by predicting somewhere between bankrupt and rich.
Each test, individually
- Forecasting from Mar 25started at $10,395+14d$11kmissed+30d$11kmissed+60d$12kouter+90d$15kouter
- Forecasting from Apr 1started at $11,383+14d$14kmissed+30d$12kmissed+60d$15kmissed+90d$15kouter
- Forecasting from Apr 8started at $11,302+14d$11kmissed+30d$12kmissed+60d$13kouter+90d$16kouter
- Forecasting from Apr 15started at $13,585+14d$14kmissed+30d$14kmissed+60d$15kouter+90d$16kouter
- Forecasting from Apr 22started at $10,757+14d$12kinner+30d$12kinner+60d$13kinner+90d$16kinner
243 days of history · 5 origins tested · pinball loss $391 · scored Jul 26