Loading live weather ensemble models and market contracts...
| Contract / Target | Date | Side | Shares | Entry Price | Current Price | Total Cost | Current Value | Unrealized PnL | Live Forecast & Win Prob | Action |
|---|---|---|---|---|---|---|---|---|---|---|
| No active positions open. Trigger a market scan to execute +EV trades. | ||||||||||
No active positions open. Trigger a market scan to execute +EV trades.
| City | Target Date | Bracket | Side | First Seen | Peak Price | # Seen | Best Edge | Model % | Why not bought | Bought? | Status | Link |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Loading sighting history... | ||||||||||||
| Status | Market Title | City | Settles | Side | Contracts | Entry | Exit Mode | Observed | Hold to Close | Total Cost | Payout | Realized PnL |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No closed trades yet. Expired contracts will settle here. | ||||||||||||
No closed trades yet. Expired contracts will settle here.
Model Calibration Trend
Every point is the average over the 10 most recently settled contracts, so the lines move as new results come in.
- Brier (blue, left axis) — how well the model's stated probabilities matched what actually happened. Lower is better: 0 = perfectly confident and always right, 0.25 = no better than a coin flip, 0.5 = confidently wrong.
- Forecast MAE (orange, right axis) — how many °F the ensemble's predicted daily high missed the observed high by, on average. Lower is better.
- + climatology (shadow) (green dashed) — the same Brier if we blended the model with 30-year station climatology. Not traded on yet; if it stays below the blue line over a real sample, we switch the live model to that blend.
A Brier pinned near 0 isn't automatically good — it usually means the model only took bets far from the bracket edge (easy calls), or was briefly overconfident on a bad forecast. The number that matters is Brier on the close calls.