pmintel.report
← all digests
Jul 2, 2026daily

Arb Bots Flood GitHub as Cross-Platform Edges Narrow

Today's GitHub sweep surfaced 19 Polymarket-adjacent repositories updated on 2026-07-02, with a notable cluster targeting cross-platform arbitrage between Polymarket and Kalshi and a separate thread of weather-market exploitation. Operator tooling is maturing fast: whale-watching, mobile P&L dashboards, and fractional Kelly sizing are all showing up in public repos.


GitHub Sweep: 2026-07-02 — 19 Repos Flagged

Lead Story: Cross-Platform Arb Is an Active Research Area

Two repos today point directly at Polymarket/Kalshi spread trading. `iamshuuu/Polymarket-Kalshi-Arbitrage-Bot` (26KB, created February) is a lean, focused tool hunting price discrepancies between the two platforms on correlated markets. More polished is `nayrbryanGaming/arbiter`, which targets *intra-Polymarket* no-arbitrage violations across related election markets — specifically, cases where a set of mutually exclusive outcomes fails to sum correctly to 1.00. The arbiter repo is notable for its explicit survival-first philosophy: fractional Kelly sizing, mandatory backtested out-of-sample edge before any live execution, and paper-trading as the default mode. For serious operators, this is worth cloning and auditing. The discipline described in its README is ahead of most public tooling.

Weather Markets: An Underattended Edge

`BallesJr/polymarket-weather-edge` (2 stars) is a paper-trading bot targeting the NO side of overpriced temperature markets, feeding live METAR observations and Open-Meteo forecasts into a Gaussian probability model. This is a narrow but structurally sound edge: Polymarket's weather markets are lightly traded and likely priced by unsophisticated participants. The repo is from the same developer as `BallesJr/polymarket-edge-model`, a Random Forest calibration pipeline with backtesting, suggesting a single operator building a multi-strategy stack in public. Watch this account.

Whale Monitoring Enters the Public Toolkit

`kydlikebtc/polymarket-whalewatch` (1 star, 1 fork, created June 25) is a real-time whale-tracking tool monitoring large positions and smart-money flows across prediction markets. The fork count is low but the creation date is recent — this is early signal of operator interest in on-chain flow analysis as an alpha source. If whale-watching tooling starts accumulating stars, it indicates the market is becoming information-efficient faster than price action suggests.

Size Anomaly Worth Flagging

`King102681/polymarket-bot` has no description and no stars, but its repository size is **5.5GB** — orders of magnitude larger than anything else in today's sweep. The next largest repo is `IanSmith2005/polymarket-dashboard` at 35MB. A 5.5GB bot repo is unusual and warrants scrutiny: possible causes include committed model weights, large dataset artifacts, or binary blobs. Operators should not run unknown code from this repo. Flag it as unvetted.

Strategy Taxonomy: What's Being Built

Across today's 19 repos, the strategy distribution breaks down as follows:

The momentum and copy-trading clusters are worth watching. `Yury617/polymarket-impulse-monitoring-trading-bot` adds trailing stops and hedging to standard momentum signals — a sign that developers are accounting for Polymarket's thin liquidity and adverse fill risk.

Security Posture

Several repos reference `.env` files for secrets management — standard practice — but `UnitedStars111/polymarket-copy-trading-bot` explicitly names secrets storage in its description. Any operator forking these repos should audit for hardcoded keys before execution. The `bit-nexusxtitmtdsuy/Polymarket_Bot` repo (32 stars, 17 forks) is the most widely distributed tool in today's sweep. High fork count on an anonymous account with no organizational affiliation is a standing risk: review all dependencies before running.

Operator Takeaway

The tooling layer is professionalizing. Fractional Kelly, out-of-sample validation, and live data feeds from authoritative sources (METAR, Open-Meteo) are appearing in public repos that six months ago would have been simple scripts. The Kalshi arb and election cross-market arb threads deserve direct monitoring. If these edges are real, they will compress quickly once more operators find them.