Weather Bots and Cross-Market Arb Dominate June 18 Commits

Polymarket Developer Activity — 2026-06-18

Weather Markets Are Now a Bot Category

Three separate repositories targeting Polymarket weather markets appeared in today's sweep: `kshitij406/polymarket-weather-bot`, `kk20054/polymarket-weather-bot`, and `BallesJr/polymarket-weather-edge`. The last of these is the most technically detailed — it describes an automated paper-trading system that fades the YES side of overpriced temperature markets using live METAR observations and Open-Meteo forecasts, with a Gaussian probability model for resolution estimation.

The clustering here is significant. When multiple independent developers converge on the same market category within weeks of each other, it typically signals either a perceived inefficiency or a shared information source pointing to one. Operators active in Polymarket weather markets should assume pricing is increasingly contested by systematic players with access to real-time meteorological data feeds. Manual traders in these markets are likely at a structural disadvantage.

Cross-Market Arbitrage Infrastructure Maturing

Several repos today reinforce a trend that has been building for months: the cross-market arbitrage stack between Kalshi and Polymarket is becoming more accessible. `pjmerica/pred-arbitrage` explicitly targets Kalshi, Polymarket, and PredictIt in a single scanner. `Max-Brunelle/market-arb-portfolio` is notable for a different reason — it is explicitly framed as architecture documentation for a private live system, a transparency choice that suggests the author is comfortable signaling the system's existence while keeping the actual edge private.

`matthewnyc2/arbitrage` describes a production-grade Python asyncio scanner focused specifically on NegRisk multi-outcome arbitrage within Polymarket itself, with 60 passing tests and a live browser demo. NegRisk arb — exploiting mispricings across the legs of a multi-outcome market — is a known opportunity that tightens quickly as tooling proliferates. The existence of a polished, test-covered implementation suggests this particular edge is being systematically harvested.

PMXT Ecosystem Expanding

`pmxt-dev/pmxt` continues to accumulate traction with 1,896 stars and 225 forks as of today's update. More interesting is `pmxt-dev/pmxt-mcp`, an MCP server that allows AI agents and LLMs to query and interact with prediction markets via PMXT's unified API. With 7 stars and 3 forks since its April launch, adoption is early but the architectural direction is clear: LLM-driven agents are being wired directly into execution infrastructure. Operators should expect AI-assisted market participation to become a routine part of the liquidity landscape within the next two quarters.

Political Market Quant Bot Notes Discipline

`nayrbryanGaming/arbiter` is a small repo worth noting for its stated philosophy: fractional Kelly sizing, a hard requirement for backtested out-of-sample edge before live trading, and paper-trading as the default. It targets cross-market no-arbitrage violations in political election markets. The discipline described is more rigorous than most public repos in this space, though the project is new and unvetted.

Security Advisory: Spam-Fork Signal

`MstKail/polymarket-trading-bot-services-polyedge365` should be treated with caution. The repository was created and last updated on the same day, has 104 stars and 251 forks against a file size of only 43KB, and its description is a keyword-stuffed repetition of service names with no technical substance. This pattern — high fork count, tiny codebase, keyword spam, same-day creation — is consistent with repositories used to distribute malware or credential-harvesting scripts under the guise of trading tools. Do not clone or execute code from this repository.

Operational Takeaway

The density of arbitrage tooling across today's sweep — NegRisk arb, cross-platform arb, complete-set arb, weather market fading — points to a market environment where structural mispricings are being systematically identified and closed. Operators relying on slow price discovery in any of these categories should reassess their edge assumptions.