Verified Evidence & Returns
Explore cumulative equity curves, drawdowns, and trade attributions derived from frozen scientific ledgers under canonical 38.0 bps roundtrip friction.
Compare with Buy & Hold (HODL)
Buy & Hold (HODL Benchmark)During the 2026 validation window, raw crypto markets suffered severe drawdowns (-28.4%), while StrataRelay preserved capital and delivered +8.00% net alpha.
Testing Methodology & Verification Timeline
To ensure mathematical validity, StrataRelay strictly segregates historical model training from out-of-sample testing. The models never look ahead into future data, and all results deduct 38 basis points round-trip execution friction.
Understanding Model Specializations & Market Regimes
Trained across complete multi-year macroeconomic cycles (bull rallies, bear crashes, institutional post-ETF regimes). Employs adaptive holding horizons—trailing profits up to 6–8 days during strong structural bull expansions while dynamically reverting to strict capital protection during chop and bear phases.
Engineered specifically for the modern institutional, range-bound market structure of recent years. Operating with shorter tactical holding windows (~24–48 hours) to lock in gains quickly in high-frequency chop without giving back profits. On older vertical bull rallies (e.g. 2021) these tactical models prioritize capital preservation over buy-and-hold risk, whereas on recent data (2025–2026) they demonstrate superior consistency and win rates.
Cumulative Normalized Equity Curve ($10,000 Initial)
Untouched HoldoutEvaluation period: 2026-03-24 to 2026-09-20 (Holdout)
Historical Drawdown Profile (%)
Max DD: -2.88%Monthly Returns Breakdown (2026 Holdout)
Click a quarter or month card to isolate its equity curve
All models are trained with causal, stationary feature representations avoiding lookahead bias. Evaluations strictly enforce 38.0 bps canonical roundtrip friction (10 bps maker/taker fee plus 9 bps simulated slippage per side). We separate Backtest from Out-of-Sample (OOS) and Forward Paper validation.