Arbitrage Bots Surge; Weather Edge and NegRisk Scanners Emerge

Polymarket GitHub Intelligence — 2026-06-29

Volume and Composition

Today's crawl returned 19 repositories with Polymarket exposure, updated or created within the last 24–48 hours. Activity spans copy-trading automation, crypto price bots, arbitrage scanners, and at least one genuinely differentiated weather-market strategy. Repository sizes vary wildly — from 77KB micro-projects to a 3.3GB behemoth (King102681/polymarket-bot) that warrants scrutiny given its lack of description and implausible size for a trading bot.

Standout Repositories

**matthewnyc2/arbitrage** is the most technically credible find today. It describes a production-grade Python asyncio scanner specifically targeting NegRisk multi-outcome markets — the category where Polymarket's internal risk netting creates exploitable cross-leg pricing gaps. The repo claims 60 passing tests and a paper-trading mode, suggesting a developer doing serious work rather than publishing vaporware. Operators running NegRisk positions should note that systematic scanners of this type compress available edges quickly once deployed at scale.

**BallesJr/polymarket-weather-edge** is the most interesting strategic entry. It targets the NO side of overpriced Polymarket temperature markets, using live METAR observations and Open-Meteo forecasts with a Gaussian probability model. This is a narrow but legitimate edge: temperature markets on Polymarket are frequently mispriced by retail participants unfamiliar with meteorological base rates. The paper-trading framing suggests early-stage validation, but the architecture is sound and worth monitoring for live deployment signals.

**signorloops/polymarket-trading** claims a high-frequency arbitrage system using convex optimization and the Frank-Wolfe algorithm. This is an unusual choice for a prediction market context — Frank-Wolfe is typically applied to constrained optimization problems in ML and portfolio construction. Either this is a genuinely sophisticated operator exploring novel execution optimization, or the description is aspirational. Either way, it represents a more mathematically rigorous approach than the typical CLOB wrapper.

**nadnad250/polymarket-bot** deploys an ML ensemble (LightGBM, XGBoost, LSTM) on BTC Up/Down 5-minute markets with auto-retraining via GitHub Actions. At 599MB, it likely includes serialized model artifacts. The 5-minute BTC markets on Polymarket are high-volume and heavily contested; ensemble approaches here face significant adverse selection from faster participants.

Security Alerts

Two repositories warrant explicit caution. **nahuelvivas/Polymarket-Arbitrage-Trading-Bot** (412 stars, 1,438 forks) has a description consisting of pure keyword spam — the phrase "Polymarket Arbitrage Bot | Polymarket Trading Bot" repeated verbatim across the entire description field. This is a known pattern for repositories designed to rank highly in searches and funnel users toward malicious downloads or credential-harvesting setups. Do not clone or execute code from this repository.

**quipmnxailcrrgky/tradingbot** (98 stars, 60 forks) has an anonymous-style username and a generic description. The star and fork counts are elevated relative to its apparent age and documentation quality — consistent with artificial inflation. Treat with the same caution.

The 3.3GB **King102681/polymarket-bot** is anomalous. Legitimate trading bots do not approach this file size. Possible explanations include bundled datasets, embedded executables, or simply a developer who committed large binary files carelessly — but the absence of any description combined with the size makes this a repository to avoid entirely.

Market Intelligence Implications

The concentration of arbitrage-focused tooling — NegRisk scanners, cross-line sports arb, news arb dashboards — reflects a maturing operator ecosystem applying systematic pressure to inefficiencies that previously rewarded manual traders. Operators relying on NegRisk pricing gaps or slow news reactions should assume compression timelines have shortened. The weather-market angle is comparatively uncrowded and represents a category worth watching for edge durability.

The copy-trading bot cluster (at least three repos) continues to grow. As leaderboard-following automation proliferates, top-wallet alpha degrades faster and position sizing by followed wallets becomes a less reliable signal.

Bottom Line

Priority watches: matthewnyc2/arbitrage for NegRisk impact, BallesJr/polymarket-weather-edge for an emerging niche strategy. Avoid nahuelvivas and quipmnxailcrrgky repos entirely.