fx-quant-lab

Systematic trading research — GBPUSD, GBPJPY, XAUUSD, NAS100, US30 (+ EURUSD reference). Every result is a deterministic, auditable backtest run: config in, pure-Python math, metrics out. No LLM sits in the signal path.

Matched-pair experiment — method class is per instrument

Identical data, dates, and cost model per instrument; only the method class differs (rolling z-score mean reversion vs EMA 20/50 momentum). Sharpe ratio by instrument, both methods:
Mean reversion Momentum
Table view — full metrics

Equity curves — growth %, both methods per instrument

Normalized to percentage growth (simulated capital bases differ per instrument). Hover for values. Each panel has its own y-scale.

Risk gate — one signal, sized per account regime

The execution gate (Exness tiers: Standard Cent → Standard → Raw Spread → Pro) converts a permitted signal into exact fractional lots. Identical setup — GBPUSD, 20-pip stop, 1% risk, 1:500 — across balances from $10 to $10M. The last row shows the drawdown circuit breaker vetoing everything. Rows are generated by the tested library code, not typed by hand.