Pancake vs zipline (Quantopian)

At a glance

What's different

zipline was the research engine powering Quantopian, the now-defunct crowd-sourced algorithmic trading platform. The zipline-reloaded fork maintains the original API for US equity backtesting against OHLCV bar data. It is a mature, well-documented engine widely cited in academic and practitioner backtesting literature.

Pancake is hosting infrastructure for AI-built trading strategies targeting prediction-market binary outcomes. Its unit of analysis is a prediction-market event: a binary outcome (YES/NO) with a decision price at decision_time and a resolution at resolution_time. The EvidenceDataset schema enforces this structure. The batter runner computes P&L from the declared entry prices and resolution outcomes, without simulating an order book. Backtest is the on-ramp toward live execution (a v2-roadmap capability); the receipt is the artifact that travels with the strategy.

The practical consequence: zipline and Pancake do not overlap on the same strategy domain. A researcher studying Polymarket, Kalshi, or similar probabilistic forecasting markets would use Pancake; a researcher studying US equity momentum would use zipline or LEAN.

Methodology overlap

Both compute Sharpe ratio, win rate, and maximum drawdown from a trade-level P&L series. Both support fee and slippage modeling at the trade level. The Wilson CI95 and Bessel-corrected variance are Pancake-specific; zipline uses pandas-level statistics without formal small-sample doctrine.

See Pancake methodologyfor full math references (Sharpe 1994, Sortino & Price 1994, Bacon 2008, Wilson 1927).

When to use each

When to use Pancake

Use Pancake when your strategy trades prediction markets — binary outcomes on political events, sports, economic indicators, and similar probabilistic questions. Pancake gives you a structured receipt with a verification boundary that any downstream LLM can read and cite.

When to use zipline (Quantopian)

Use zipline (or zipline-reloaded) when you are backtesting US equity strategies against historical OHLCV bar data in a Jupyter notebook environment. zipline has a mature pipeline API and tight integration with the Zipline data bundle format.