Pancake vs Tardis.dev
Tardis.dev is a crypto market microstructure data archive, not a backtester. Pancake is a prediction-market backtester. They solve different problems.
At a glance
| Capability | Pancake | Tardis.dev |
|---|---|---|
| Open-source engine | ✓ Apache-2.0 (batter, Python) | ✗ proprietary data API |
| Prediction-market native | ✓ Polymarket, Kalshi, binary outcomes | ✗ crypto perpetuals and spot only |
| Backtesting engine | ✓ batter runner with P&L + risk metrics | ✗ data archive only — no backtesting engine |
| Verification boundary doctrine | ✓ explicit 3-tuple in every receipt | ✗ data quality docs only, no strategy verification |
| Agent-callable MCP surface | ✓ 6-tool surface (v1.3) | ✗ REST data API only |
| Receipt URLs with byte-stable hashes | ✓ /r/<short_id> | ✗ not applicable (data archive) |
| Crypto tick data (L2 order book) | ✗ prediction markets only | ✓ tick-level L2 data for major crypto venues |
| Replay infrastructure | ✗ | ✓ exact replay of historical L2 order book state |
What's different
Tardis.dev and Pancake solve fundamentally different problems. Tardis.dev is a data company: it archives tick-level L2 order book data for major crypto derivatives venues (Binance, FTX, BitMEX, etc.) and provides APIs and replay infrastructure for researchers building microstructure strategies. It is not a backtester.
Pancake is hosting infrastructure for AI-built trading strategies targeting prediction-market binary outcomes. It has no tick data, no order book replay, and no crypto perpetuals support. The unit of analysis is a prediction-market event with a binary resolution — the antipode of Tardis's L2-tick regime. Backtest is the on-ramp; live execution (a v2-roadmap capability) is where the strategy is headed.
The two tools are complementary for researchers who study the overlap between crypto market microstructure and prediction markets — but that overlap is narrow. A researcher studying Polymarket USDC pools or Kalshi crypto event contracts would use Pancake for the strategy validation receipt; they might use Tardis for data context on correlated crypto spot prices.
Methodology overlap
Both platforms deal with timestamped financial data and care about data provenance. Tardis documents its data quality (missing ticks, venue outages) in the same spirit as Pancake's unmodeled_risks list. Neither platform claims to model market impact in its primary use case.
When to use each
When to use Pancake
Use Pancake when your research question is "did this binary prediction strategy have edge?" on a corpus of Polymarket or Kalshi markets. Pancake gives you a structured receipt with a verification boundary and reproducible statistics.
When to use Tardis.dev
Use Tardis.dev when you need historical tick-level L2 order book data for crypto derivatives venues, exact replay of historical microstructure state, or research into market-making and execution on perpetual futures.
Citation
Tardis.dev provides historical crypto market microstructure data and replay infrastructure. tardis.dev. Pancake comparison: usepancake.com/compare/pancake-vs-tardis-dev