PERFORMANCE & VALIDATION

Out-of-sample metrics validated across 13+ years of data. Six independent layers of anti-overfitting evidence answer the central question of quantitative finance: how do you know it generalizes?

KEY METRICS

SHARPE RATIO (OOS)
0
Ensemble performance
Net of all transaction costs
VALIDATED TRACK RECORD
0
2012 to 2025
Zero losing calendar years
DATA ANALYZED
0
Price bars across 5 pairs
7 timeframes

PER-PAIR PERFORMANCE

PAIR TIMEFRAME PROFIT FACTOR SHARPE POST-GATE SHARPE WIN RATE
PAIR_AH1█.███.████ to ████.█%
PAIR_BH1█.███.████ to ████.█%
PAIR_CH1█.███.████ to ████.█%
PAIR_DH1█.███.████ to ████.█%
PAIR_EM15█.███.████ to ████.█%
CLASSIFIED
DETAILED PERFORMANCE BREAKDOWN AVAILABLE UNDER NDA

YEARLY PERFORMANCE

Every calendar year profitable since inception. Net of all transaction costs. Zero losing calendar years across the full track record.

YEAR TRADES PNL (PIPS) AVG PNL/TRADE
Y-01███+█,███+██.█
Y-02███+█,███+██.█
Y-03███+█,███+██.█
Y-04███+█,███+██.█
Y-05███+█,███+██.█
............
CLASSIFIED
YEAR-BY-YEAR PNL AVAILABLE UNDER NDA
Every calendar year profitable. Zero losing months in the last 24 months.

GATE IMPACT

Each gate layer independently validated via 5-fold walk-forward cross-validation. All improvements are additive and statistically significant. The gate pipeline transforms raw signal performance into institutional-grade risk-adjusted returns.

GATE IMPACT NOTES
Triple Gate (MAE+MFE+TP)Significant improvement5-fold CV, 100% positive
+ Hybrid TP (Analytical + MFE)Additional uplift5-fold CV, 100% positive
+ Quad Gate (ML PnL)Incremental gain5-fold CV, 5/5 positive
Full PipelineSubstantial combined effectAll layers combined
Each gate layer validated independently before stacking. No circular logic. Post-gate performance available under NDA.

VALIDATION FRAMEWORK

Six independent layers of anti-overfitting evidence. Each layer uses a fundamentally different methodology. Together they make the strongest case possible that the observed performance is structural, not an artifact of data mining.

PBO = 0.0000
PROBABILITY OF BACKTEST OVERFITTING: ZERO ACROSS ALL COMBINATORIAL FOLDS
LAYER 01
Theory-Driven Indicators
All core indicators derived from established mathematical fields: stochastic processes, information theory, state estimation, extreme value theory, spectral analysis, fractal geometry. None invented to trade FX.
LAYER 02
Constrained Search Space
Compact rule set with a constrained parameter space. Each parameter has clear mathematical interpretation. Not millions of opaque parameters.
LAYER 03
Temporal Train/Test Split
Strict 80/20 temporal split. OOS Sharpe actually exceeds in-sample, a strong anti-overfitting signal indicating the model generalizes to unseen data. This confidence interval inversion is rare and meaningful: it suggests the strategy captures a structural edge that strengthens over time rather than decaying.
LAYER 04
Combinatorial Purged Cross-Validation
The gold standard of backtest validation. PBO = 0.0000 across all combinatorial folds. Near-perfect rank correlation between in-sample and out-of-sample performance.
LAYER 05
Deflated Sharpe Ratio
Passes at 1% significance after correcting for thousands of independent trials, non-normality, short track records, and multiple testing bias.
LAYER 06
Walk-Forward Cross-Validation
5-fold anchored expanding windows. 100% of all fold-pair tests positive across triple gate, hybrid TP, and quad gate. Walk-Forward Efficiency exceeds 1.0.
DETAILED VALIDATION RESULTS
METRICPAIR_APAIR_BPAIR_CPAIR_D
CPCV PBO0.00000.00000.00000.0000
IS-OOS rho█.████.████.████.███
DSR p-value<0.05<0.05<0.05<0.05
WFE Ratio█.███.███.███.██
SR0 (null)█.████.████.████.███
SR observed█.███.███.███.██
CLASSIFIED
FULL VALIDATION METHODOLOGY AVAILABLE UNDER NDA

HONEST REPORTING

Transparency is not a compliance checkbox. It is a competitive advantage.

The backtest-live gap is documented openly. A gap was identified between backtest and live execution. Root cause identified across 15+ research sessions. A targeted fix was deployed, and the fix has been validated. Honest reporting builds more trust than hidden disclaimers.

Every edge decays. Every system breaks. What matters is how fast you find the gap and how rigorously you close it.