Claude Builds Polymarket BTC 5-Min Scalping Bot

The Breakthrough

Claude codes a Polymarket trading bot that scalps Bitcoin 5-minute up/down markets by entering 'yes' bets when prices exceed 0.95 in the final 35 seconds (skipping the last 8 seconds), holding positions to resolution without early exits to avoid fees.

What Actually Worked

• Creates a Polygon wallet via Claude, stores private key in .env, funds with 30 USDC.e for trading and 5-57 MATIC (initially ~$5) for gas on Polygon network using MetaMask transfers.
• Sets Polymarket client with Gamma API/SDK; approves USDC.e spending allowance via Polymarket account signup (no prior account needed initially, but enables trading/autoredeem).
• Researches Bone Reaper (profitable trader: $20k P&L in 6 days, 180k trades/week) via blockchain analysis: identifies late-window convergence scalping (bets >0.9 price on 'yes', high volume, hold to end); replicates with entry cap 0.95-0.97, live size scaling, one position max.
• Configures dry-run first, then live: tightens to 20s window (35s before end minus 8s), checks order book/spot prices, places orders only on convergence signals.
• Builds React dashboard (white-text Bloomberg-terminal style, 5s blink on trades): shows total equity, wallet address, recent trades, decision log, live window timer/prices (e.g., 'down at 94', ask 95), real-time updates.

Before / After

Bot starts with $30 USDC.e. After first trades: +$2 (7%). After 1hr: nearly $34. After 2hrs: above $34 (no losses, all profitable trades). Bone Reaper benchmark: $330 past month, $20k in 6 days, $30k total.

Context

Polymarket's highest-frequency market is Bitcoin price up/down in next 5 minutes. Author uses Claude (with sub-agents) in Cursor/Replit-like setup to read Polymarket docs, generate wallet/client, research Bone Reaper's public trades, implement backend (fetch markets, order placement), test dry/live, and frontend dashboard. Runs live for fun/scalability demo on small capital; warns against big bets without extensive testing. Latency less critical for this non-arbitrage strategy.

Notable Quotes

• "late window convergence scalping not any directional predictions... holds every position to resolution... puts in like a large bet... when they are above .9"
• "tighten the T remaining so we're only going to trade in the last 35 seconds but we're going to skip the 8 seconds at the end so we only have like a 20 second window"
• "is this like uh something that's going to make you rich I wouldn't put a lot of money on this like at least before you have done a lot of extensive testing over time"