Track Record
Our signals are backtested across 10 years of daily EUR/USD market data (2016–2026). The accuracy numbers below are out-of-sample: computed on held-out test data that the models never saw during development.
Accuracy by Horizon and Sensitivity
"Sensitivity" controls how often the model issues a signal vs staying neutral. Higher confidence = fewer signals, higher accuracy.
| Sensitivity | This week | Next few weeks | Next 3 months |
|---|---|---|---|
| High confidence only | — | 100% (3/3) | — |
| Balanced | — | 89.1% (41/46) | 91.7% (11/12) |
| Need to convert | 57.4% (136/237) | 71.3% (221/310) | 65.8% (223/339) |
How We Measure Accuracy
A signal is correct when:
- Convert signal → EUR/USD rate falls (your euros buy more dollars)
- Wait signal → EUR/USD rate rises (waiting saves you money)
We measure at the end of each horizon window: 5 trading days ("this week"), 20 trading days ("next few weeks"), or 60 trading days ("next 3 months").
All accuracy figures are out-of-sample: computed on held-out test data that the models never saw during development. This is the gold standard for validating quantitative models — it prevents overfitting.
When We're Wrong
No model is perfect. When our signal is wrong, the average rate move against us is small. Our errors are close calls, not catastrophic misses. The model gets the big moves right.