Coherence Arb Meets Fee Opacity: 6 Repos Worth Your Time
Today's sweep turned up six legitimate bots spanning weather arb, longshot bias, and coherence violations — plus three security alerts including two confirmed malware payload patterns. The coherence arbitrage repo (ArtBreguez) is the analytical standout, directly stress-testing whether probability-sum violations survive transaction costs.
Six Cloneable Repos, Two Malware Flags, and a Longshot Bias Paper That Actually Has Data
Saturday's GitHub sweep covered 8 new repositories. Six cleared our review threshold. None are confirmed malware, but three security alerts from this cycle demand attention before anything else.
---
🔴 Security First
**Two repos this cycle contain large binary blobs consistent with malware payload containers.** We are not linking them. The pattern — oversized binary assets committed alongside thin Python trading scaffolding — matches the delivery mechanism flagged in our July 31 advisory. A third alert involves a 460KB clone fingerprint that matches a known malware template circulating in the Polymarket bot community since June. If you pulled anything from GitHub this week that fits this description, treat your keys as compromised and rotate immediately.
Separately, at least one repo cluster is running **star-farming bots** to inflate credibility signals. Star count is not a safety proxy. Verify before you clone.
---
📦 New Bots Worth Cloning
**BallesJr/polymarket-weather-edge** — Weather arbitrage using live METAR observation data combined with Open-Meteo forecasts. The core model is Gaussian, estimating true temperature-outcome probabilities and systematically targeting the NO side when market odds overprice outcomes. Ships with paper-trading mode. Useful immediately for any operator already positioned in weather markets or looking for a low-correlation edge source.
**RonakToprani/plytrademoch** — Longshot bias exploitation in "unanchored" markets (markets without a numeric anchor, e.g. not "will price exceed $X" style). The bot buys underdogs based on documented research showing positive EV for longshots in this market class. Calibrated on 356,000 historical Polymarket markets, which is a serious dataset. Runs in DRY-RUN mode — validate before live deployment. The unanchored-market distinction is the operationally important detail here; don't apply this logic blindly to all markets.
**ArtBreguez/polymarket-coherence** — The most analytically interesting repo this cycle. Scans multi-outcome markets for coherence violations — cases where probabilities for mutually exclusive outcomes sum to something meaningfully different from 100%. Critically, it doesn't just detect violations; it tests whether they survive bid-ask spread. This pairs directly with two of today's alpha signals (see below). If you're building coherence arb infrastructure, start here.
**fisherben93/arbitrage-radar-odds** — Browser-based (HTML/JavaScript) cross-book scanner. Pulls proxy odds from external sources to flag cross-book arbitrage. Lower infrastructure barrier than Python-based equivalents. Useful for manual operators who want a live dashboard without standing up a full bot stack.
**midas93230-cell/donmarket** — Fee inference toolkit. Analyzes order flow and attributed executions to reverse-engineer the actual fee schedules applied by different Polymarket front-ends and market-making interfaces. This is foundational work if you're modeling net EV accurately. Most operators are underestimating effective fees on specific front-ends; this tool surfaces the delta.
---
📡 Alpha Signals
Three research papers surfaced this cycle, two of which are directly operationally relevant.
The first distinguishes **payoff-space no-arbitrage from protocol-executable no-arbitrage**. The key finding: Polymarket's negative-risk market structure allows probability bound violations to be *observed* but not *exploited* until settlement. This reframes coherence arb — you may be right and still unable to capture the edge intramarket.
The second analyzes **multi-condition market structure**, emphasizing that condition sets must be exhaustive and mutually exclusive for proper resolution. The probabilistic framework it presents is directly applicable to building detection logic on top of ArtBreguez/polymarket-coherence.
The third is a language-model merger arbitrage system — long-context reasoning over M&A documents to predict deal outcomes. Less immediately deployable on Polymarket but relevant for operators active in M&A resolution markets.
---
📰 Community Pulse
Three HN threads worth tracking: Spain has blocked Polymarket and Kalshi over gambling licensing — watch for regulatory contagion into other EU jurisdictions. The "Nothing Ever Happens" bot (always bets NO on non-sports markets) hit 471 points, which tells you something about the current meta. And the Iran missile story death-threat thread is a reminder that tail-risk markets attract tail-risk participants.
---
👁 Watch Next
Monitor whether Spain's regulatory action triggers copy-legislation discussions in France or Germany. Track the coherence arb paper thread for follow-up empirical work. Any new repos citing the 356k Polymarket historical dataset used by plytrademoch deserve a close read.