Christine Lagarde — the cross-bank generalization test.
Per-speaker drill-down. 52 ECB monetary-policy press conferences and Q&A sessions parsed since 2019-12 – 2026-04. ≈ 328K tokens. The Powell-built lexicon was applied to Lagarde without modification — the architecture transfers if and only if a Fed-watcher word list picks up signal at a different central bank operating in a different currency / equity bloc. 46 meetings scored walk-forward.
Cross-asset edges (walk-forward, n=46)
Same Ridge model architecture as Powell, applied to ECB-relevant assets. Both euro equity targets validate at scale. EUR/USD is honestly weak — the verbal signal does not robustly predict the cross-rate, likely because EUR/USD is jointly driven by Fed signals as much as ECB.
| Asset | Class | Edge over momentum | Sharpe (ann) | 95% CI | ρ | Status |
|---|---|---|---|---|---|---|
| EuroStoxx 50 | Equities | +14.3pp | 0.74 | [-0.03, 1.48] | +0.06 | validated |
| DAX | Equities | +12.2pp | 0.56 | [-0.27, 1.19] | +0.04 | validated |
| EUR/USD | FX | -10.1pp | 0.21 | [-0.60, 1.13] | +0.09 | weak |
Note: directional edges are positive on EuroStoxx and DAX (validated walk-forward Sharpes). Predicted-vs-actual Pearson ρ is small (~0.05) on equity targets — direction-acc edges can exist with low continuous-value correlation, particularly when the model is well-calibrated on the sign but conservative on magnitude. Disclosed honestly.
Last 8 walk-forward predictions — EuroStoxx 50
Each row is a real walk-forward prediction made using only data available at the source presser's timestamp. 4 of 8 directional hits. The recent run is below the n=46 lifetime hit rate of 58.7% — disclosed rather than cherry-picked. The full walk-forward sample is what validates; recent runs fluctuate around it.
| Meeting date | Predicted | Actual | Predicted dir | Actual dir | Hit? |
|---|---|---|---|---|---|
| 2025-06-05 | -1.97% | +0.10% | down | up | ✗ miss |
| 2025-07-24 | -1.32% | +0.20% | down | up | ✗ miss |
| 2025-09-11 | -1.50% | +0.47% | down | up | ✗ miss |
| 2025-10-30 | -2.18% | -0.12% | down | down | ✓ hit |
| 2025-12-18 | +0.68% | +1.06% | up | up | ✓ hit |
| 2026-02-05 | +1.12% | -0.75% | up | down | ✗ miss |
| 2026-03-19 | -1.31% | -2.14% | down | down | ✓ hit |
| 2026-04-30 | +0.31% | +1.12% | up | up | ✓ hit |
Word-level alpha — DAX
Pearson correlation between Lagarde's usage frequency of each lexicon term at the prior presser and the realized ECB-day DAX move. Top 5 in each direction across the n=52 corpus. "Expansion" being the most bearish word may surprise — the model picks up that Lagarde uses growth-language defensively when she's setting up a hawkish surprise.