CLI Backtest

Run from src/:

cd src

# Default: BTC-USD, Bollinger + momentum, full grid search
python main.py

# Quick single backtest (skip grid search)
python main.py --no-grid

# Equity backtest
python main.py --symbol AAPL --asset equity --window 50 --signal 1.5

# Custom date range
python main.py --symbol ETH-USD --start 2020-01-01 --end 2026-01-01

# Different indicator + strategy
python main.py --indicator sma --strategy reversion

# Sweep multiple indicators and strategies in grid search
python main.py --indicator bollinger sma rsi --strategy momentum reversion

# Stochastic oscillator
python main.py --indicator stochastic --strategy momentum

# Optimize over both price and volume
python main.py --factor price volume

# Custom grid search bounds
python main.py --win-min 10 --win-max 60 --win-step 10 --sig-min 0.5 --sig-max 2.0

# Custom output directory
python main.py --outdir /tmp/results

All CLI Options

Run python main.py --help for the full list.

Flag	Default	Description
--symbol	BTC-USD	Yahoo Finance ticker
--start	2016-01-01	Backtest start date
--end	2026-04-01	Backtest end date
--asset	crypto	crypto (365 days/year) or equity (252 trading days/year)
--indicator	bollinger	bollinger, sma, ema, rsi, stochastic — accepts multiple
--strategy	momentum	momentum or reversion — accepts multiple
--factor	price	price, volume, or both
--window	20	Indicator window for single backtest
--signal	1.0	Signal threshold for single backtest
--no-grid	false	Skip parameter optimization
--win-min/max/step	5/100/5	Grid search window range
--sig-min/max/step	0.25/2.50/0.25	Grid search signal range
--outdir	../results	Output directory for CSVs and heatmap
--walk-forward	false	Run walk-forward overfitting test
--split	0.5	In-sample ratio for walk-forward split (0.0–1.0)

Output Files

Saved to results/ by default:

File	Contents
perf_<symbol>_<indicator>.csv	Daily PnL, cumulative returns, drawdown, positions
opt_<symbol>_<indicator>.csv	Grid search results (window, signal, sharpe)
heatmap_<symbol>_<indicator>.png	Sharpe ratio heatmap
wf_<symbol>_<indicator>.csv	Walk-forward in-sample vs out-of-sample comparison

Walk-Forward Overfitting Test

The walk-forward test splits data into in-sample (training) and out-of-sample (validation) periods:

|◄──────── 10 years of data ────────►|
|◄── in-sample (5y) ──►|◄── out-of-sample (5y) ──►|
   grid search here          test best params here

# Walk-forward with default 50/50 split
python main.py --walk-forward

# 70% in-sample / 30% out-of-sample
python main.py --walk-forward --split 0.7

Overfitting ratio: 1 − (OOS Sharpe / IS Sharpe). Values near 0 indicate robust parameters; values near 1 indicate the strategy performs much worse out-of-sample.