Production Analytics
All metrics reported directly from the XTNL analytics pipeline. Dataset labels correspond to the internal subset naming convention. Live execution (N=29) reflects the current deployment phase; SESSION_FILTERED (N=106) and FULL_OPTIMAL (N=308) span the complete forward-test universe.
SESSION_FILTERED_OPTIMAL_SAMPLE
FULL_OPTIMAL_SAMPLE
LIVE_SESSION_FILTERED — Current Deploy
LIVE dataset context: N=29 live trades is statistically insufficient for edge confirmation. The [CAUTION] health status is expected and appropriate at this stage of deployment. The SESSION_FILTERED dataset (N=106) provides the validated statistical baseline from the forward-test period. WFO validation on the live dataset requires a minimum of 100 trades.
RECENT_100_SESSION_FILTERED
RECENT_200_FULL_OPTIMAL
95R_SESSION_FILTERED (Outliers Excl.)
REGIME: ALPHA_FLOW
REGIME: TOXIC_FATIGUE
HMM regime labels are assigned by the Gaussian Hidden Markov Model oracle using expectancy Z-score, execution density, and diurnal fatigue features. ALPHA_FLOW represents the high-performance regime; TOXIC_FATIGUE represents the degraded regime. Notably, both regimes show positive expectancy — regime detection is used to calibrate risk allocation, not to halt trading.
Walk-Forward Optimisation — SESSION_FILTERED (N=106)
Expanding-window validation. Each fold trains on historical data then tests on the immediately following unseen period. Aggregate result determines operational status.
| Stage | Train N | Test N | IS SQN | OOS SQN | IS E | OOS E | Degradation | Status |
|---|---|---|---|---|---|---|---|---|
| Fold 1 | 22 | 21 | 2.248 | 1.853 | 1.280 | 0.889 | 17.56% | — |
| Fold 2 | 43 | 21 | 2.938 | 3.059 | 1.089 | 1.583 | −4.11% | — |
| Fold 3 | 64 | 21 | 4.163 | 0.937 | 1.251 | 0.457 | 77.50% | — |
| Fold 4 | 85 | 21 | 4.094 | 1.300 | 1.055 | 0.686 | 68.26% | — |
| Aggregate | — | — | — | 1.787 | — | 0.904 | — | STABLE - OPERATIONAL_EDGE |
Fold 3 (77.5% degradation) and Fold 4 (68.26%) reflect the WFO pressure that prompted temporal quarantine of the 19:00 AEST cluster. The filtered dataset recovers to STABLE status.
Current-Week Inferred Metrics
Current-week inferred metrics reflect live performance relative to the theoretical optimal execution baseline. RISK REDUCE status triggers an automatic position size reduction via the performance-gating mechanism.