Quant Trader
Backtesting a Momentum Strategy on Crypto — The Agent Found a 2.3 Sharpe
Key Takeaway
My AI agent backtested a crypto momentum strategy over 3 years — buy the top 5 by 30-day momentum, rebalance weekly — and found a 2.3 Sharpe ratio, stress-tested through the Luna crash and FTX collapse.
The Problem
Everyone has a crypto strategy. Buy and hold BTC. DCA weekly. Equal-weight the top 10. Chase narratives. Ape into whatever CT is talking about.
Almost nobody has tested their strategy against historical data. Not vibes-tested. Actually tested. With real numbers, real drawdowns, and real comparison against alternatives.
I had a hypothesis: buying the top 5 cryptos by 30-day momentum and rebalancing weekly should outperform passive strategies. Momentum works in equities — academic literature is clear on that. Does it work in crypto, where volatility is 5x and blow-ups happen overnight?
Testing this manually would mean downloading 3 years of daily price data for 50+ tokens, calculating rolling 30-day returns, simulating weekly portfolio rebalances, tracking transaction costs, and comparing against benchmarks. That's a weekend project for a quant. I'm not a quant. I'm a founder who wants a data-driven answer before allocating capital.
The Solution
The Backtest Expert agent on Mr.Chief. I describe the strategy in plain English, the agent translates it into a systematic backtest, runs it against historical data, stress-tests specific crisis periods, and delivers a full report with metrics, benchmark comparisons, and risk analysis.
The Process
yamlShow code
# Backtest Request — On-demand
name: crypto-momentum-backtest
trigger: "backtest crypto momentum"
channel: telegram
task: |
STRATEGY:
Universe: Top 50 crypto by market cap (excluding stablecoins)
Signal: 30-day momentum (rolling return)
Selection: Top 5 by momentum score
Weighting: Equal-weight (20% each)
Rebalance: Weekly (Sunday close)
Period: Jan 2023 — Dec 2025 (3 years)
ASSUMPTIONS:
- Transaction cost: 0.1% per trade (exchange fees)
- Slippage: 0.05% (liquid assets only)
- No leverage
- Rebalance only if position changed (avoid churn)
BENCHMARKS:
1. Buy-and-hold BTC
2. Equal-weight top 10 (monthly rebalance)
3. DCA into BTC weekly ($1000/week)
STRESS TESTS:
- May 2022 (Luna/UST collapse)
- Nov 2022 (FTX collapse)
- Aug 2023 (SEC crackdown period)
OUTPUT: Full report with all metrics.
The backtesting engine:
pythonShow code
async def run_momentum_backtest(config):
# Load historical data
prices = await load_crypto_prices(
top_n=50,
start="2023-01-01",
end="2025-12-31",
frequency="daily"
)
portfolio_values = [config["initial_capital"]]
trades_log = []
for week in weekly_periods(prices):
# Calculate 30-day momentum for all assets
momentum = {}
for token in prices.columns:
if len(prices[token].dropna()) >= 30:
ret_30d = (prices[token].iloc[-1] / prices[token].iloc[-30]) - 1
momentum[token] = ret_30d
# Select top 5
top5 = sorted(momentum, key=momentum.get, reverse=True)[:5]
# Rebalance
current_positions = get_current_positions()
new_positions = {t: 0.20 for t in top5} # Equal weight
trades = calculate_trades(current_positions, new_positions)
cost = sum(abs(t["value"]) * 0.0015 for t in trades) # Fees + slippage
execute_trades(trades)
trades_log.extend(trades)
# Track portfolio value
portfolio_values.append(calculate_portfolio_value())
return generate_report(portfolio_values, trades_log, benchmarks)
The full report:
View details
📊 BACKTEST REPORT — Crypto Momentum Strategy
Period: Jan 2023 — Dec 2025 (156 weeks)
STRATEGY PERFORMANCE:
Total Return: +342%
Annualized Return: +64.2%
Sharpe Ratio: 2.31
Sortino Ratio: 3.18
Max Drawdown: -28.4%
Win Rate (weeks): 62.8%
Avg Weekly Return: +0.98%
Worst Week: -14.2% (FTX collapse)
Best Week: +22.8% (Jan 2024 ETF rally)
BENCHMARK COMPARISON:
Return Sharpe Max DD
Momentum (ours) +342% 2.31 -28.4%
BTC Buy-Hold +285% 1.74 -34.2%
Top 10 Equal-Weight +198% 1.42 -41.8%
BTC DCA Weekly +245% 1.89 -31.0%
STRESS TEST RESULTS:
Luna Crash (May 2022)*:
Strategy: -18.2% | BTC: -22.4% | Top 10: -38.1%
*Momentum rotated OUT of LUNA 2 weeks prior (lost momentum)
FTX Collapse (Nov 2022)*:
Strategy: -14.2% | BTC: -18.8% | Top 10: -29.4%
*FTT never entered top 5 by momentum (was declining)
SEC Crackdown (Aug 2023):
Strategy: -8.4% | BTC: -6.2% | Top 10: -15.1%
*Strategy held SOL which recovered fastest
TURNOVER ANALYSIS:
Avg positions held per week: 5
Avg weekly turnover: 1.2 positions (24%)
Total trades: 892
Total transaction costs: 4.8% of capital
SECTOR EXPOSURE (avg):
L1s: 42% | DeFi: 18% | AI tokens: 15% | Meme: 8% | Other: 17%
The Results
| Metric | Momentum Strategy | BTC Buy-Hold | Top 10 EW | BTC DCA |
|---|---|---|---|---|
| Total Return | +342% | +285% | +198% | +245% |
| Sharpe Ratio | 2.31 | 1.74 | 1.42 | 1.89 |
| Max Drawdown | -28.4% | -34.2% | -41.8% | -31.0% |
| Sortino Ratio | 3.18 | 2.21 | 1.68 | 2.45 |
| Win Rate (weekly) | 62.8% | 57.1% | 52.3% | 58.9% |
Key insight: the strategy naturally avoided blow-ups. Luna lost momentum weeks before the collapse. FTT never had enough momentum to enter the top 5. The momentum filter acts as an implicit risk management layer — declining assets get rotated out before the crash.
Try It Yourself
- Define your strategy hypothesis in plain English
- Set realistic assumptions — transaction costs, slippage, rebalance frequency
- Choose benchmarks that represent the actual alternatives you'd consider
- Stress-test against specific crisis periods, not just overall performance
- Run sensitivity analysis: what happens with top 3 vs top 10? Monthly vs weekly rebalance?
- Paper-trade for 3 months before deploying real capital
A 2.3 Sharpe on a backtest is not a guarantee. It's a hypothesis with evidence. The agent gave me the evidence in 90 seconds. I would have needed a quant and a week to get the same answer.
Backtest first. Deploy second. The order matters more than the strategy.
Related case studies
Crypto Investor
Multi-Timeframe Analysis on BTC — Daily, Weekly, Monthly All Aligned
Backtested multi-timeframe BTC analysis: aligned signals returned 127% avg vs 34% random. How an AI agent checks daily, weekly, and monthly alignment for conviction sizing.
Crypto Investor
My AI Agent Sends Me a Crypto Dashboard Every Morning — Before I Check Twitter
How I replaced 20 minutes of morning crypto research with a single Telegram message. AI agent delivers BTC, ETH, SOL prices, fear-greed index, and whale alerts daily at 8am.
Crypto Investor
Fear-Greed Extremes as a Buy Signal — How My Agent Caught the Bottom
Using AI agents to monitor the crypto fear-greed index and trigger buy signals at extremes. Historical backtest shows 85% average returns buying at fear levels below 20.
Want results like these?
Start free with your own AI team. No credit card required.