Results

No cherry-picking. No survivorship bias. Every chart generated from real backtest data β€” 42 million ticks of EURUSD, walk-forward validated, tick-verified where possible.

The Graveyard

31 strategies tested. 2 survived. Here's the full picture.

Strategy graveyard β€” all results with ALIVE/DEAD verdicts

Every strategy with its tick-verified profit factor. The yellow dashed line is breakeven (PF=1.0). Green = alive. Red = dead.

ECVT β€” Our Best Edge

Entropy Collapse Volatility Timing on EURUSD 1H. Based on Singha 2025.

ECVT walk-forward equity curve

+198 bps across 44 trades. PF 1.44, 41% WR, 2.08Γ— win/loss ratio. Walk-forward validated: 6M train / 2M OOS / 2M step.

ECVT per-window OOS results

10/20 OOS windows profitable. Cumulative (teal line) trends upward β€” the edge persists across time.

Entropy collapse visualization

The pattern: entropy drops during consolidation (yellow zones), then price explodes. The signal detects compression before breakout.

FVG Autopsy β€” Death by Tick Data

Fair Value Gap strategy looked great on OHLC daily (PF 4.28). Then we checked with tick data.

FVG multi-timeframe tick autopsy

Performance degrades monotonically: 5min PF 0.80, 15min PF 0.94, 1H PF 1.04. OHLC daily inflated results by 4Γ— due to Close-based SL/TP bias. Strategy formally dead.

Hurst β€” The Filter That Never Filters

R/S Hurst exponent with 100-bar lookback on crypto hourly. Supposed to detect trending vs mean-reverting regimes.

Hurst distribution on crypto hourly

100% of bars classified as "trending" (H > 0.6, mean = 0.758). Direction WR: 48-50% β€” a coin flip. The regime filter has zero information content.

Jump Signal β€” Tick Catastrophe

Jump detection on EURUSD tick data. Looked promising on Gold daily. Died on forex hourly.

Jump signal tick backtest results

All R:R ratios negative. PF 0.43–0.56. The fade component was poison; the trend component alone was profitable (55% WR at 1:1) β€” composite signals can hide working components.

R:R Sweep β€” The Variance Tradeoff

Risk:Reward ratio sweep across all signal types. Higher R:R = better math but worse drawdown.

Risk:Reward sweep across signal types

Win rate drops as R:R increases. For prop firms (WR > 50% + low DD), only Gold daily at 1:1 qualifies β€” but trades too rarely (17-23 per 2 years).

On-Chain Signals

What the LLM trading agent sees: funding rates, open interest, and liquidation cascade zones.

On-chain signals dashboard

7-day view. BTC price (top), funding rate bars (middle), open interest accumulation (bottom). Red zones mark liquidation cascades where OI unwinds violently.

Reproduce Everything

Every chart, every backtest, every signal β€” fully open source. Clone and verify.

View Code on GitHub