prediction-market-backtesting

New in Version 4.1-alpha:

• Live sandbox plumbing for Polymarket BTC 5min markets
• Example runner showing how to use live BTC 5min hooks
• (strategy & model now included)
• open sourced some /private files (felt like it)

New in Version 4:

• Nautilus 1.226.0
• Rust-native data conversion
• Faster staged data loading
• Improved materialized caches
• Unified cache/local/archive/API message bus

New in Version 3:

• Telonex vendor support
• Local Telonex download script
• Many bug fixes & accuracy improvements
• Book replay order book deltas with trade ticks

New in Version 2:

• Nautilus 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 Polymarket, built on top of NautilusTrader with custom exchange adapters. Limitless.exchange and Opinion.trade are planned next; Kalshi support depends on access to L2 historical book data. Current Kalshi components are research and fee-modeling plumbing, not a public runnable backtest path. 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
  • Start Here
  • Core Framework
  • Advanced / Experiments
  • Project
  • Acknowledgements
• Setup
  • Prerequisites
  • Install
  • First Run
  • Timing And Cache Defaults
  • Extension Architecture
• Backtests And Runners
  • Repo Layout
  • Archived Private Research
  • 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
    • Telonex
• Sandbox And Live Runners
  • Current Scope
  • Directory Contract
  • Sandbox Runner Contract
  • Shared Live Helpers
  • BTC 5m Sandbox Plumbing
  • Example BTC Snapshot Runner
  • Archived Strategy Boundary
  • Model And Parameter Placement
  • Running Sandbox
  • Public Polymarket Data
  • Path To Live Polymarket Trading
• Data Loading
  • Mental Model
  • Staged Loading
  • PMXT Flow
  • Telonex Flow
  • Caching
  • Downloading Local Data
  • Progress And Timing
  • Failure Semantics
• Polymarket Account Ledger Replay
  • Runner And Notebook
  • What The Strategy Does
  • Why Exact Reproduction Fails
  • Copy-Trading Interpretation
  • External Source Check
  • Observed Result
  • Latest Terminal Output
  • How To Use This Experiment
• 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
    • Maker Rebates
  • Slippage
  • Passive Orders And Queue Position
  • Latency
  • Limits
  • Vendor L2 Behavior
    • PMXT
    • Telonex
• Data Vendors And Local Mirrors
  • PMXT
    • Runner Source Modes
    • Lower-Level Loader Env Vars
    • What Works Today
    • Supported Local File Layout
    • Required Parquet Columns
    • Legacy JSON Payload Shape
  • Telonex
    • Download Local Telonex Files
  • What Is Not Plug-And-Play Yet
• Vendor Fetch Sources And Timing
  • PMXT
  • Example Output
  • Telonex
  • 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