quant-engine

Vectorised backtesting engine for systematic trading strategies. Built to research and validate multi-instrument strategies across equities and FX — not a wrapper around backtrader or zipline.

Capabilities

• Vectorised position simulation with full trade accounting (slippage, commissions, fractional sizing)
• Risk-per-trade position sizing (fixed % of equity)
• Daily loss limit halting
• Walk-forward validation (rolling windows)
• Tested across 10 instruments, 5+ years of data

Quick start

import yfinance as yf
from engine import run_backtest, build_indicators, DEFAULT_PARAMS
from metrics import full_report

data = yf.download("QQQ", start="2019-01-01", end="2024-12-31", interval="1h")
htf  = data.resample("4h").last().dropna()
df   = build_indicators(data, htf)

params = {
    **DEFAULT_PARAMS,
    "risk_pct": 0.005,
    "reward_ratio": 3.0,
}

equity, trades = run_backtest(df, params)
report = full_report(equity, trades)
print(f"Sharpe: {report['sharpe']:.2f}  MaxDD: {report['max_drawdown']:.1%}")

Engine parameters

Parameter	Default	Description
initial_capital	10000	Starting equity
risk_pct	0.005	Fraction of equity risked per trade
reward_ratio	4.0	Take-profit in R-multiples
daily_loss_limit	0.06	Halt new entries after -6% intraday
commission_pct	0.0001	Round-trip commission as fraction of notional
slippage_pts	2.0	Slippage in price points per side

Walk-forward validation

from walk_forward import rolling_holdout, wfo_summary
from engine import run_backtest, build_indicators

results = rolling_holdout(df, run_backtest, train_bars=504, test_bars=126, params=params)
summary = wfo_summary(results)
print(f"Pass rate: {summary['pass_rate']:.0%}  Mean OOS Sharpe: {summary['mean_sharpe']:.2f}")

Validated across 10 instruments, 5+ years of data: 13–15 of 16 out-of-sample windows profitable per strategy.

Installation

pip install -r requirements.txt

Examples

• examples/simple_ma.py — MA crossover (minimal working example)
• examples/run_backtest.py — full API walkthrough with report printing