I just put together a quantitative backtesting tool and open-sourced it for anyone who needs it. If it helps you, feel free to follow and show some support — thanks!!!

Is your trading idea actually any good? Don’t rely on gut feelings — throw it into historical K-line data and see what happens!

As a Hermes newbie, I used to only validate trading strategies through paper trading. After joining the BG hackathon, I learned that data backtesting is also an option.

So I open-sourced a local crypto backtesting tool: Stone Quantitative Backtesting Lab (Hermes Backtest Lab)

Open-source repo: https://github.com/stong123456/stone-backtest-lab…

It doesn’t need an API Key, it’s not an exchange bot. Its one core job: take your trading idea, run it through real historical data, and let the numbers speak.

——————————————————

Core capabilities:

1. OKX spot + USDT perpetual swap backtesting
2. Multi-currency portfolio backtesting (simulates a real multi-asset account)
3. Long/short both sides (long-only / short-only / both)
4. On-chain token backtesting (GeckoTerminal data source)
5. Full cost modeling (fees + slippage)
6. Dynamic position sizing, leverage, margin, drawdown circuit breaker
7. Built-in common factors (trend, mean reversion, ATR, ADX, RSI, EMA, Bollinger Bands, volume z-score, etc.)
8. Report export — each backtest generates three files:
   • report.md: human-readable backtest report
   • metrics.json: machine-readable metrics
   • trades.csv: detailed log of every simulated trade

——————————————————

You can run it with PowerShell or directly feed it to your AI.

Quick start for beginners (recommended):

python -m pip install -r requirements.txt
python scripts/hermes_backtest_lab.py --preset demo

Run with common presets directly:

--preset balanced / conservative / aggressive / spot / onchain-demo

Custom OKX perpetual backtest:

python scripts/hermes_backtest_lab.py --symbols BTC,ETH,SOL,SUI --inst-type SWAP --days 180 --bar 1H

Spot backtest (short disabled):

--symbols BTC,ETH,SOL --inst-type SPOT --allow-short 0 --days 180

On-chain token backtest:

--data-source geckoterminal --symbols base:token:0x... --days 30 --bar 1H

See the repo README for more usage.

——————————————————

Why do I think backtesting is the most valuable thing?

It’s not meant to prove you can make money. It’s a cold, impartial quality inspector:

• Does your strategy survive fees and slippage?
• Is it propped up by a few big wins?
• Are win rate, profit/loss ratio, max drawdown, trade frequency reasonable?
• Can it still survive on a different set of coins or a different time period?

If a strategy can’t even pass the historical-data-plus-costs test, putting it on a live account is mostly just donating money.

So this tool is for three kinds of people:

• People who want to validate their own trading ideas
• People who want to systematically study different coins, timeframes, and parameters
• People who want to automate trading but don’t want to trial-and-error with real money right away

——————————————————

Final reminder: Historical backtesting does not guarantee future results. But without backtesting, most of the time you won’t even know where you lost.

Feel free to star, clone, and run it. After running, issues and feedback are welcome!

stong123456/stone-backtest-lab

Source: https://github.com/stong123456/stone-backtest-lab

Stone Quantitative Backtesting Lab

English name: Hermes Backtest Lab.

A local cryptocurrency backtesting tool that requires no API Key.

It uses OKX public K-line data and GeckoTerminal public DEX data. Supports spot, USDT perpetual swaps, multi-chain on-chain tokens, multi-currency portfolio backtesting, fee/slippage modeling, dynamic leverage, dynamic margin, take-profit/stop-loss, time-based exit, factor attribution, and report export.

This is not investment advice. Historical backtesting does not represent future returns. Please validate with small samples and paper trading first.

Features

• Backtest any OKX spot or USDT perpetual swap symbol
• Backtest any GeckoTerminal-supported on-chain token or DEX pool
• Single-currency backtesting or multi-currency portfolio backtesting
• Supports long/short both sides, long-only, short-only
• Built-in trend-following and mean-reversion logic
• Built-in indicators: ATR, ADX, RSI, EMA, Bollinger Bands, volume z-score, etc.
• Supports fees, slippage, max position, account drawdown circuit breaker
• Outputs report.md, metrics.json, trades.csv
• Does NOT read .env, exchange accounts, private keys, API Keys, or Telegram Tokens

Installation

Requires Python 3.10+.

cd hermes-backtest-lab
python -m pip install -r requirements.txt

Windows users can also run directly:

.\run_setup.ps1

Quick Start

python scripts/hermes_backtest_lab.py --preset demo

If you’re not sure which command to use, first view the Chinese short menu:

python scripts/hermes_backtest_lab.py

After running, files will be generated in outputs/hermes_backtest_lab/<timestamp>/:

• report.md: human-readable backtest report
• metrics.json: machine-readable metrics
• trades.csv: detailed log of every simulated trade

Common Presets

# View all presets with Chinese explanations
python scripts/hermes_backtest_lab.py --list-presets

# Dry-run: only show config, no network or backtest
python scripts/hermes_backtest_lab.py --dry-run --preset balanced

# 30-day quick check, good for confirming environment is working
python scripts/hermes_backtest_lab.py --preset demo

# 180-day balanced portfolio backtest
python scripts/hermes_backtest_lab.py --preset balanced

# More conservative: only mainstream coins, reduce portfolio exposure
python scripts/hermes_backtest_lab.py --preset conservative

# More aggressive: include more high-volatility assets
python scripts/hermes_backtest_lab.py --preset aggressive

# Spot, long-only
python scripts/hermes_backtest_lab.py --preset spot

# On-chain token backtest, automatically picks the highest-liquidity pool
python scripts/hermes_backtest_lab.py --preset onchain-demo

# Base chain WETH token
python scripts/hermes_backtest_lab.py --data-source geckoterminal --symbols base:token:0x4200000000000000000000000000000000000006 --days 30 --bar 1H

# Specify a particular DEX pool directly
python scripts/hermes_backtest_lab.py --data-source geckoterminal --symbols base:pool:0xPOOL_ADDRESS --days 30 --bar 1H

You can override any parameter in a preset:

python scripts/hermes_backtest_lab.py --preset balanced --symbols BTC,ETH,SOL --days 90 --max-leverage 8

You can also put symbols in a txt file, one per line, for easier maintenance:

python scripts/hermes_backtest_lab.py --preset balanced --symbols-file examples/symbols_okx_bluechip.txt

Windows one-click scripts:

.\run_demo.ps1
.\run_balanced.ps1
.\run_onchain_demo.ps1

Custom Backtesting

# Any OKX USDT perpetual
python scripts/hermes_backtest_lab.py --symbols BTC,ETH,SUI,LINK --inst-type SWAP --days 180 --bar 1H

# Specify OKX instId directly
python scripts/hermes_backtest_lab.py --symbols BTC-USDT-SWAP,ETH-USDT-SWAP --days 90 --bar 4H

# Spot backtesting
python scripts/hermes_backtest_lab.py --symbols BTC,ETH,SOL --inst-type SPOT --allow-short 0 --days 180

# On-chain token. Format: chain:token:contract. Script automatically selects highest-liquidity pool.
python scripts/hermes_backtest_lab.py --data-source geckoterminal --symbols eth:token:0xA0b86991c6218b36c1d19D4a2e9Eb0cE3606eB48,base:token:0x4200000000000000000000000000000000000006 --days 30 --bar 1H

# On-chain pool. Format: chain:pool:pool_address.
python scripts/hermes_backtest_lab.py --data-source geckoterminal --symbols solana:pool:POOL_ADDRESS --days 30 --bar 1H

# Multi-currency sharing a single portfolio account (capital not split by symbol)
python scripts/hermes_backtest_lab.py --symbols BTC,ETH,SOL,SUI,LINK --inst-type SWAP --portfolio-mode 1 --starting-balance 100000 --min-margin 1000 --max-margin 3000

Important Parameters

See More Commands

python scripts/hermes_backtest_lab.py
python scripts/hermes_backtest_lab.py --list-presets
python scripts/hermes_backtest_lab.py --examples
python scripts/hermes_backtest_lab.py --advanced-help

How to Read Results

Don’t only look at win rate. Prioritize:

• Whether total return and max drawdown are aligned
• Profit Factor > 1
• Average profit significantly larger than average loss
• Sufficient number of trades (don’t be fooled by small samples)
• Whether profits are concentrated on a few days or a few coins
• Whether the strategy still works after adding fees and slippage

On-Chain Data Notes

• Token mode automatically selects the highest-liquidity pool returned by GeckoTerminal; it may not be the specific pool you intended to trade.
• Pool mode is more precise; use it when you already know the target pool address.
• New coins, small pools, or low-liquidity pools may have incomplete K-lines; data warnings will appear in the report.
• On-chain backtesting is only a price-series study by default. It does not simulate real on-chain execution risks such as MEV, gas, pool impact costs, buy/sell taxes, blacklists, trade pauses, routing failures, etc.
• For meme and small-cap tokens, manually increase slippage, e.g., --slippage-bps 30 or higher.

Open-Source Sharing Suggestions

If you publish to GitHub, only commit these:

• README.md
• SKILL.md
• requirements.txt
• scripts/hermes_backtest_lab.py
• examples/commands.md
• .gitignore

Do NOT commit:

• outputs/
• runtime-logs/
• cache/
• .env
• API Keys, Telegram Tokens, exchange account files

Stone (@Stone141319): A lot of people talk about AI Web3, focusing on models, applications, airdrops, etc. But when it comes to AI agents executing on-chain tasks, calling data, coordinating trades, managing assets, the problems become very, very real: Can its reasoning be verified? Is the data it calls reliable? Where is its memory stored? When it collaborates with other agents, who audits if something goes wrong?