Skip to content

PyPI page (example)

Example only, not a release

This is a draft of what the racinglines page on PyPI could look like. The package is not on PyPI: nothing has been uploaded and the name is not registered. Publishing waits for the owner's go. Some commands below use the new CLI tree (task groups such as racinglines model forecast f1, --demo [PACK] on any command), which is still being built, and the licence wording is an open question.

Draft metadata

Field Draft value
Name racinglines
Version 0.2.0
Summary Fair prices and market making for race sports: F1, MotoGP, NASCAR and UCI downhill against prediction markets
Requires Python >=3.12; numpy>=2.5, pandas>=2.3,<3, rich>=15
Extras racinglines[remote], racinglines[server], racinglines[dev]
Licence LicenseRef-racinglines-NC (the MIT text plus a non-commercial restriction; see LICENSE)
Links racinglines.bet, source

Draft long description

Everything below is the draft page text as PyPI would show it (from PYPI.md on the PyPI draft branch).

Fair prices and market making for race sports. racinglines turns official timing into win, podium, head-to-head and championship odds, priced strictly as of any moment and tested on real market tapes, and gives makers and takers the tools to find and size an edge in race-sport prediction markets (Polymarket, Kalshi, OG.com) and against generic sportsbook lines.

  • Formula 1, the big end: a sector-aware car model, practice prior and grid/overtaking finishing model, backtested on 129 races (2021 to Azerbaijan 2026, backtest run 931). After qualifying, its top-10 Brier score is 0.149 against 0.163 for a grid-only guess.
  • UCI downhill mountain biking, the small end: modelled and forecast from split timing.
  • NASCAR Cup and MotoGP: results, market links on Kalshi, Polymarket and OG.com, and a simple result-only model in replay. No edge has held up in a second season yet.

Nothing is traded with real money: strategies run as backtests, replays on recorded exchange prices, and paper trading.

Install

pip install racinglines            # the CLI and the demo
pip install "racinglines[remote]"  # + the hosted server (sign in through the browser)
pip install "racinglines[server]"  # + everything to run your own database, web app and MCP server

Python 3.11 or newer, and PostgreSQL for the demo store.

Try it in a minute: the demo

--demo [PACK] on any command runs that same command on a demo store: a separate database and data folder built from a data pack by the real ingest. The default pack, kart, is a synthetic Formula 1 world with a Mario Kart grid; every pace, result and price is made up, and it is play money. It needs PostgreSQL (the demo database is created next to yours) but no network or account.

racinglines db stats --demo                                       # 1. build the demo store and look at it
racinglines model backtest kart --demo                            # 2. every past race priced and scored
racinglines model forecast kart --demo --save                     # 3. the rest of the season and the championships
racinglines live run kart/demo-2026-09 --demo --simulate --no-fetch # 4. a live private book on a simulated clock

These are the real commands: --demo only switches the data source (Quickstart).

Three ways to run it

What you need What runs
Demo pip install racinglines and PostgreSQL The five steps above, on a bundled synthetic data pack, with no network or account
Hosted A racinglines.bet account with MCP access racinglines book price FILE --remote prices your own book file on the hosted server; the first run signs in through the browser
Self-hosted PostgreSQL and racinglines[server] The full pipeline below on your own data, and your own MCP server (racinglines book config mcp_url URL makes it the CLI default)

Self-hosted quickstart

docker compose up -d                  # PostgreSQL (the repo's docker-compose.yml)
racinglines db init                   # tables + reference data
racinglines data fetch f1             # F1 sessions (FastF1)
racinglines data ingest f1            #   -> database
racinglines model forecast f1 --save  # live F1 prices: remaining races + championships
racinglines markets sync              # exchange market links
racinglines check                     # code, data endpoints, database

What it does

  • Collects official timing into one multi-sport PostgreSQL schema, and exchange history (prices, trades, order books) into a Parquet archive.
  • Models each sport and simulates race weekends and seasons thousands of times: win, podium, top 10, pole, head-to-head and team markets, plus championships.
  • Prices strictly as of a moment: only sessions that ended before the cutoff are used, and a leakage guard enforces it. Backtests, diagnostics, the live forecast and the signal engine call the same function.
  • Tests strategies on real market data: maker replays, taker strategies that re-trade after every session, and season-long championship strategies, on recorded exchange prices and trades.
  • Prices sportsbook lines and slips against the model and the exchanges, side by side, never blended.
  • Serves it all through an MCP server, so a chat client can read the data and run the simulations.

Licence

Source-available, not open source: the MIT licence text with an added restriction that allows personal, non-commercial inspection and study only, and prohibits commercial use (including as a hosted service) without written permission. See LICENSE.