May 2026 resolved 32 qualified calls: Naly 12/32 (38%), Brier 0.384. Calibration and Brier score are updated at /predictions/scorecard; this recap summarises only the answer-flip events with a component score of at least 20 that resolved between May 1 and May-end UTC.
- Qualified resolutions: 32 (Naly 12 correct, Polymarket 20 correct on the same set).
- Brier score: Naly 0.384, Polymarket 0.216. Lower is better; 0 is perfect, 0.25 is a coin flip.
- Best call: Will China announce a Boeing aircraft purchase by May 22? — component score 62, Naly top answer "Yes" resolved Yes.
- Worst call: US x Iran diplomatic meeting by May 15, 2026? — component score 81, Naly top answer "Yes" resolved No.
Brier and Calibration
This month we evaluated 32 qualified answer-flip predictions. Each Brier score below is the mean of (favored-answer probability − actual outcome)².
| Forecaster | Brier | Lower is better |
|---|---|---|
| Naly | 0.384 | — |
| Polymarket (same events) | 0.216 | — |
Calibration buckets
| Bucket | Predicted avg | Actual frequency | Count |
|---|---|---|---|
| 40-60% | 56% | 0% | 4 |
| 60-80% | 68% | 50% | 12 |
| 80-100% | 86% | 38% | 16 |
Accuracy by Category
| Category | Resolved | Correct | Accuracy |
|---|---|---|---|
| politics | 24 | 9 | 38% |
| stock | 3 | 1 | 33% |
| geopolitics | 3 | 1 | 33% |
| elections | 2 | 1 | 50% |
Top Qualified Calls This Month
| Event | Naly top answer | Market top answer | Component score | Resolved | Naly right? |
|---|---|---|---|---|---|
| US x Iran diplomatic meeting by May 15, 2026? | Yes (99%) | No (82%) | 81 | No | ✗ |
| Will Donald Trump visit China on May 14, 2026? | Yes (86%) | No (77%) | 63 | No | ✗ |
| Will China announce a Boeing aircraft purchase by May 22? | Yes (95%) | No (66%) | 62 | Yes | ✓ |
| Iran closes its airspace by May 27? | Yes (78%) | No (83%) | 61 | No | ✗ |
| US x Iran diplomatic meeting by May 15, 2026? | Yes (80%) | No (77%) | 57 | No | ✗ |
Methodology note
This recap only counts resolved predictions where Naly's top answer flipped away from Polymarket's top answer AND the answer-level component score was at least 20 points. Same-side recalibrations and weak flips are tracked separately and excluded from the core scorecard. The pipeline, selection rule, and Brier calculation are described in full at /methodology.
Disclaimer: not financial advice. Prediction markets are real-money instruments with real risk. Read the track record and methodology before trading anything.
