Research project that streams Binance trades, builds per-second features and estimates the probability that a Polymarket 5-minute crypto Up/Down window closes up, comparing a Bayesian model with a LightGBM challenger against the market price. Ships a write-up of the ways the research produced false edges and how each was caught; the author's conclusion that there is no tradeable edge at this latency is self-reported.
Last updated · source link verified
Research project that streams Binance trades, builds per-second features and estimates the probability that a Polymarket 5-minute crypto Up/Down window closes up, comparing a Bayesian model with a LightGBM challenger against the market price. Ships a write-up of the ways the research produced false edges and how each was caught; the author's conclusion that there is no tradeable edge at this latency is self-reported. See the source for setup, usage and current status.
Tools: PythonLightGBMBinance and Polymarket data
Difficulty: advanced
Platforms: Polymarket
Research project that streams Binance trades, builds per-second features and estimates the probability that a Polymarket 5-minute crypto Up/Down window closes up, comparing a Bayesian model with a LightGBM challenger against the market price. Ships a write-up of the ways the research produced false edges and how each was caught; the author's conclusion that there is no tradeable edge at this latency is self-reported. It is listed under Data & analytics in this directory.
According to its source, it works with Polymarket.
The source lists: Python, LightGBM, Binance and Polymarket data.
It was published by madhur-guptaa. The public source is at github.com/madhur-guptaa/polymarket-5m-research.
This directory rates it advanced (beginner, intermediate or advanced).
It had 0 GitHub stars as of Oct 19, 2026.
The repository was last updated on Oct 19, 2026, as checked on Oct 23, 2026. Check the source for current status.
Compare all data & analytics tools → · How to build a Polymarket bot
More data: Prediction market datasets and free Kalshi JSON downloads
Framework for collecting and analyzing prediction-market data, including a large public Kalshi prediction market dataset.
Dataset stat Becker Kalshi dataset: 7.68M markets; 23.9% of settled markets resolved YES. Method on detail page.
Scripts to collect Polymarket market and trade data for analysis.
Student research notebook (Rennes School of Business graduating project) that compares the probability distribution implied by SPXW and NDX option prices, using a fitted SVI smile, with the distribution quoted by Kalshi's S&P 500 and Nasdaq-100 range contracts for the same expiry. Uses Yahoo Finance option data and the Kalshi public API, with a saved snapshot in the repo. Findings are self-reported and based on a single snapshot.
New bots weekly. One email, no spam.
Form not loading? Subscribe on beehiiv. Handled by beehiiv; see privacy. Unsubscribe any time.