Descriptions:
Creator Sharbel A. shares the unvarnished 14-day results of running an OpenClaw-powered trading bot on Polymarket’s 15-minute cryptocurrency prediction markets. Starting with $100 on a Mac Mini running 24/7, the bot placed 550 trades and achieved a 92% win rate — but netted only $36 profit on $7,400 total wagered. The video’s central lesson is that win rate is a vanity metric: the bot’s small wins (a few cents per share) were frequently offset by outsized losses of $10–$30 on a single bad bet.
The per-asset breakdown is illuminating: Solana markets drove +$36 at a 94% win rate, Ethereum came in near flat at +$3, and Bitcoin was a consistent drag at -$30. Three expansion attempts all failed — 5-minute markets were too volatile with worse spreads, whale copy trading consistently entered positions after smart money had already moved prices, and weather prediction markets simply didn’t fit the strategy’s edge. Each failure is analyzed with specific reasoning rather than dismissed.
The video also details a practical monitoring architecture: a secondary AI agent named “Max” checks on the trading bot every two hours and reports status to Telegram, illustrating how production autonomous systems require supervisory layers. Sharbel closes by describing a reset to V5 of the 15-minute Solana strategy with improved position sizing logic, which was already showing $22 in a single day versus $36 across the prior two weeks combined. Essential viewing for anyone building or evaluating autonomous AI trading systems using OpenClaw.
📺 Source: Sharbel A. · Published February 18, 2026
🏷️ Format: Workflow Case Study
