Pancake vs Alpaca Paper Trading

At a glance

What's different

Alpaca Paper Trading is a sandbox layer on top of a live broker API. It gives developers a real-time simulation environment where orders are matched against actual NBBO quotes (IEX feed) without routing to a live exchange. A new paper account starts with $100,000 in simulated capital. The API surface — REST and WebSocket — is identical to the live account; you switch environments by pointing your client to paper-api.alpaca.markets and using a separate set of API keys. PDT rules, margin, and position limits are all simulated.

Pancake is hosting infrastructure for AI-built trading strategies — not a broker simulator. It does not simulate order routing, real-time fills, or account margin. Its unit of analysis is a prediction-market event: a historical binary position with a known entry price at decision_time and a known resolution at resolution_time. The batter engine computes P&L deterministically from those inputs. Backtest is the on-ramp; the receipt travels with the strategy as it advances toward live execution (a v2-roadmap capability) via your own connected accounts. There are no random partial fills, no NBBO matching, and no real-time data feed — only a validated EvidenceDataset and a reproducible calculation.

The epistemic gap is meaningful. Alpaca's paper trading documentation lists what it does not model (market impact, information leakage, price slippage, order queue position, borrow fees) but that list lives in the docs, not in the result payload. A developer reading paper trading results has to recall those caveats manually. Pancake's unmodeled_risks block — listing market_impact, resolution_lag, resolver_risk, and small_sample — is embedded in the receipt JSON so any model or reader consuming the result encounters the caveats at the same time as the statistics.

Methodology overlap

Both platforms apply fee and slippage concepts to simulated trades. Alpaca Paper Trading simulates realistic NBBO fills with stochastic partial fills; Pancake applies a deterministic fee_bps to each prediction-market entry. Both disclose known modeling omissions — Alpaca in documentation, Pancake in the receipt JSON unmodeled_risks field. Neither platform claims to model market impact in its standard configuration.

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-market binary outcomes and you need a reproducible, agent-readable receipt with a structured verification boundary. Pancake is the right tool when the evidence is historical (assembled from resolved markets) and the receipt must survive downstream LLM scrutiny without additional context lookups.

When to use Alpaca Paper Trading

Use Alpaca Paper Trading when you are developing an equity, options, or crypto trading algorithm and want to test it against real-time market data with simulated NBBO execution before committing real capital. Alpaca Paper is the right sandbox for API-driven live trading development on US markets.