Bot Arms Race Accelerates: Arb and Copy-Trade Dominate

Polymarket Dev Activity — 2026-06-16

Overview

Today's GitHub intelligence sweep returned 20 repositories touching Polymarket infrastructure, up meaningfully from last week's baseline. Three dominant themes emerged: **cross-platform arbitrage** (Polymarket ↔ Kalshi), **copy-trading automation**, and **crypto price prediction bots** targeting short-duration BTC/ETH/SOL markets. The volume of activity suggests the retail automation layer is maturing — and that serious operators need to assume more bot-driven order flow in liquid markets than ever before.

---

Key Repos to Watch

**`bit-nexusxtitmtdsuy/Polymarket_Bot`** is the standout by raw traction: 32 stars, 17 forks, and a 35MB repo size that suggests bundled assets or models rather than a lean script. No detailed description, which is a pattern worth flagging. High fork counts on anonymous-style accounts can indicate coordinated promotion rather than organic developer interest. Treat with caution.

**`BallesJr/polymarket-edge-model`** is the most technically credible entry today. The description references a Random Forest calibration model, backtesting infrastructure, and automated paper trading — a legitimate, structured pipeline. At 777KB with zero stars, this looks like genuine solo developer work rather than a promotional repo. Worth monitoring for future updates; the calibration angle is exactly where edge lives in mature prediction markets.

**`Panca2341/polymarket-trading-bot`** (4 stars, 3 forks) and **`samuel483/poly-kalshi-arb`** (5 stars, 2 forks) represent the mid-tier retail automation stack — Python-based, beginner-friendly framing, gasless transaction promises. These are the tools your counterparties are running.

**`tduic/prediction-market`** deserves a closer read. The description calls out Kelly sizing and risk-managed execution explicitly — rare for this class of repo. An async Python system with an embedded analytics dashboard targeting Poly/Kalshi cross-platform arb is a more sophisticated build than the typical copy-paste bot. Zero stars suggests early or private-primary development.

**`kmjjjj/polymarket-arbitrage-bot-btc-sol-15m`** targets 15-minute BTC and SOL price markets specifically in Rust — a performance-conscious choice that hints at latency awareness. If this bot is live, it is competing in the most crowded short-duration markets on the platform.

---

Security Alerts

Elevated caution is warranted across this batch. Several patterns recur that experienced operators should recognize:

• **Anonymous or low-history accounts** pushing large binary repos (Sahil56899 at 42MB, mariomerinom at 170MB, IanSmith2005 at 28MB) with no descriptions. Oversized repos with no readme are a classic vector for credential-harvesting tooling dressed as trading bots.
• **Credential management language** in bot descriptions (Menna-Awad11) is a specific red flag — any tool that touches your API keys or wallet signing deserves full source audit before execution.
• **Copy-trading bots** by definition require access to your account to mirror trades. giovanna24623 and mahmoooud194 both fit this profile. Do not run these against funded accounts without complete source review.

General posture: assume any repo with under 10 stars and an oversized file footprint is unaudited. Paper-trade environments only.

---

Strategy and Alpha Signals

**Cross-platform arb is crowded but real.** Five repos today explicitly target Kalshi/Polymarket spreads. This confirms persistent pricing gaps exist — but also that the arb is being automated by an expanding pool of participants. If you run this strategy manually, your execution window is compressing. If you run it programmatically, latency and fee optimization are now the differentiator.

**Short-duration crypto markets are a bot battlefield.** BTC/ETH/SOL 5-minute and 15-minute markets are explicitly named across multiple repos. Human traders operating in these windows are price-taking against automated flow. Consider avoiding or fading sharp moves in these markets that lack news catalysts — they may reflect bot positioning rather than genuine probability updates.

**Calibration models are the next moat.** BallesJr's Random Forest calibration approach points toward where sophisticated edge is being built: not raw prediction, but probability calibration against market prices. Operators who invest in calibration infrastructure will increasingly outperform those relying on raw model outputs or news reaction alone.

---

*Next sweep: 2026-06-17. Flag unusual star velocity on any repo — coordinated promotion remains the primary noise source in this feed.*