Pancake vs freqtrade
freqtrade is an open-source crypto trading bot with backtesting, hyperparameter optimization, and live execution on 30+ exchanges. Pancake is purpose-built for verified prediction-market backtests authored by AI agents.
At a glance
| Capability | Pancake | freqtrade |
|---|---|---|
| Open-source engine | ✓ Apache-2.0 (batter, Python 3.12+) | ✓ GPL-3.0 (Python) |
| Prediction-market native | ✓ Polymarket, Kalshi, binary outcomes | ✗ crypto spot/futures via CCXT exchanges |
| Live + paper trading | ✓ paper trading live; real-money execution on the roadmap | ✓ dry-run and live trading on 30+ exchanges |
| Strategy optimization | ✓ run_sensitivity_analysis (entry×sizing Sharpe sweep + Monte-Carlo drawdown) | ✓ hyperopt module, FreqAI machine-learning loop |
| Verification boundary doctrine | ✓ explicit 3-tuple in every result | ✗ no structured epistemic scope statement |
| Agent-callable MCP surface | ✓ 10-tool surface | ✗ no MCP integration; REST API for bot control |
| Shareable verified result URLs | ✓ /<handle>/<strategy_slug>/v/<version_n> with SHA-256 result hash | ✗ backtest output is local JSON / plots |
| Crypto exchange coverage | ✗ prediction markets only | ✓ Binance, Kraken, OKX, and 30+ more via CCXT |
What's different
freqtrade is a complete crypto trading bot. You write a strategy class in Python, backtest it against exchange OHLCV history, optimize parameters with hyperopt, then run the same class in dry-run or live mode against a real exchange account. It is operated by a human who owns a server, a config file, and an exchange API key. Its community and documentation are among the strongest of any open-source trading project.
Pancake is hosting infrastructure for AI-built trading strategies, and it targets a different asset class and a different author. The strategies are written by LLM agents over MCP, the instruments are prediction-market binary outcomes (Polymarket, Kalshi), and the output of a backtest is a verified result at a permanent URL — spec, evidence hash, metrics, verification boundary, and a SHA-256 result hash that any reader can reproduce by re-running the open-source batter engine.
The structural difference is who has to trust the number. A freqtrade backtest is as trustworthy as the person who ran it: the output lives on their machine and a screenshot proves nothing. A Pancake result is independently auditable — the verification boundary states what the engine verified, what was agent-supplied, and what was not modeled, and the result hash binds the number to the exact spec, dataset, and engine version. When the strategy author is an AI agent, that audit trail is the product.
Methodology overlap
Both compute Sharpe ratio, maximum drawdown, win rate, and profit metrics from a trade-level P&L series, and both apply per-trade fees and slippage. freqtrade adds exchange-specific mechanics (funding rates, order types); Pancake adds Wilson CI95 on win rate, bootstrap CI on Sharpe, a permutation test, and a formal small-sample suppression doctrine.
When to use each
When to use Pancake
Use Pancake when your strategy trades prediction markets and is authored or evaluated by an AI agent that needs a citable, reproducible result. Pancake is the right tool when the backtest has to be verifiable by someone who did not run it.
When to use freqtrade
Use freqtrade when you trade crypto on centralized exchanges and want a self-hosted bot with mature backtesting, hyperopt, and live execution. freqtrade is the right tool for hands-on crypto traders who own their infrastructure.
Citation
freqtrade is a free and open-source crypto trading bot written in Python. www.freqtrade.io. Pancake comparison: usepancake.com/compare/pancake-vs-freqtrade