prediction-market-backtesting

Thanks to PMXT for providing this data for free!

New in Version 2:

• Nautilus 1.225.0, via PyPI in lieu of a subtree
• Better backtest runner classes via EXPERIMENT objects
• IPython notebook support (.ipynb files)
• Joint portfolio multi replay runners
• Growing support for statistical optimizers
• New aggregate charts
• Massive improvements charting gen speed
• an attempt at a Tree-structured Parzen Estimator via Optuna

Looking for the old version? That was renamed to Version 1

Backtesting framework for prediction market strategies on Kalshi and Polymarket, built on top of NautilusTrader with custom exchange adapters. Plotting inspired by minitrade. This repo is still in active development.

Fantastic single & multi-market charting. Featuring: equity (total & individual markets), profit / loss ticks, P&L periodic bars, market allocation, YES price (with green buy and red sell fills), drawdown, sharpe (with above/below shading), cash / equity, monthly returns, and cumulative brier advantage.

If you find any bugs, unexpected behavior, or missing simulation features, PLEASE post an issue or discussion.

Detailed guides have been filed away in the docs index for better organization and long-term sustainability.

Table of Contents

• Docs Index
  • Acknowledgements
• Setup
  • Prerequisites
  • Install
  • First Run
  • Timing And Cache Defaults
  • Extension Architecture
• Backtests And Runners
  • Repo Layout
  • Runner Contract
  • HTML And Report Modes
  • Optimization Runners
  • Designing Good Runner Files
  • Multi-Market Strategy Configs
  • Running Backtests
  • Notebook Runners
  • Editing Runner Inputs
  • Data Vendor Notes
    • Native Vendors
    • PMXT
• Research
  • Overview
  • Warm PMXT Cache Before Notebook Runs
  • Scoring
  • Joint-Portfolio Mode
  • Samplers
    • Random Grid (sampler="random")
    • TPE (sampler="tpe")
  • Caveats
  • Notebook Output Persistence
• Execution Modeling
  • Fees
  • Slippage
  • Passive Orders And Queue Position
  • Latency
  • Limits
  • Vendor L2 Behavior
    • PMXT
• Data Vendors, Local Mirrors, And Raw PMXT
  • PMXT
    • Runner Source Modes
    • Lower-Level Loader Env Vars
    • What Works Today
    • Supported Local File Layout
    • Required Parquet Columns
    • Required JSON Payload Shape
  • Relay Mode
  • What Is Not Plug-And-Play Yet
• Mirror And Relay Ops
  • Active Mirror Service
  • PC-Side Alternative
• Vendor Fetch Sources And Timing
  • PMXT
  • Example Output
  • Timing Expectations By Source
  • How To See This Output
• Plotting
  • Scaling Model
  • Downsampling
  • Output Types
  • Output Paths
  • Example Summary Output
  • Multi-Market References
• Testing
  • Standard Repo Gate
  • Useful Smoke Checks
  • Docs Validation
• Project Status
  • Roadmap
  • Known Issues
  • Recently Fixed
• License Notes
  • Scope
  • NautilusTrader Attribution
  • Practical Meaning

Star History