Needs your decision 0
Performance β realized P&L curve
Open & closed positions
| # | Market | Side | Shares | Entry | Mark | Win% | P&L | Return | Opened | Held | Status |
|---|
Proposal history 0
| # | Market | Side | Entry | Risk | Integrity | Analyst | Status | Outcome |
|---|
Recent trades
| # | Time | Mode | Side | Price | USDC | Shares | Status | Detail |
|---|
Last scan β weather markets considered 0
| Market | Our P(hit) | Why it didn't trade |
|---|
Analysis Β· out-of-sample calibration, breakdown, what-if replay, untracked cities
Out-of-sample calibration (is the model honest forward?)
'Recent' = markets resolved in the last 7 days (the model didn't see these when its bias was fit).
If recent predicted β actual with a low Brier, the corrections work forward β safe to scale.
'Older' is the in-sample baseline.
Calibration breakdown
predicted vs actual per slice, with the offset that would close the gap for THAT slice
| Slice | N | Predicted | Actual | Brier | Suggested offset |
|---|
What-if replay
re-pick over every logged scan with alternate params; reports how the trade selection would shift
Untracked cities (no station mapping)
cities the scanner sees but we don't have a settlement-airport ICAO for; these run on city-center grid only