I Ran 3 Claude Bots on Polymarket | Here's What They Are

The speaker discusses their recent success in five-minute markets, particularly in women's tennis, and introduces three new bots they've built, emphasizing that while they share ideas, building profitable bots remains challenging. They highlight the accessibility of coding with AI and aim to provide conceptual frameworks rather than plug-and-play solutions. The first bot presented is a "copy bot," designed to replicate the trades of highly successful traders on Polymarket's five-minute up-down markets. The speaker explains how they identify profitable profiles by analyzing the "activity tab" on Polymarket, looking for traders with consistent gains over $100. They demonstrate how to input a trader's address, like "Hitman" or "Latka," into the bot. This copy bot, named "Lopka bot," employs a "stink bid" approach, meaning it observes the size of a successful trader's position and then takes a slightly discounted position on the same side, aiming for a better entry price rather than blindly mirroring the exact trade. A bonus idea, "stink bid on momentum," is also briefly mentioned, where the bot identifies market momentum and places a bid in the desired direction. The second bot, called "Hype 5-Minute" or "Hyperliquid Spread Tightener," addresses the wide spreads found in the newly introduced five-minute markets on Hyperliquid Crypto. The bot's purpose is to provide liquidity by simultaneously bidding and asking, thereby tightening the spread between the bid and ask prices. The speaker shows the bot in action, noting that it had been previously rejected due to insufficient cash, a problem resolved by claiming recent winnings. During the demonstration, the speaker identifies a potential flaw where the bot was bidding in the last 30 seconds of a market, leading to a loss. They promptly instruct Claude (an AI) to implement a rule preventing bids in the final 30 seconds, demonstrating the iterative nature of bot development. This bot aims to profit from the difference between the bid and ask prices by acting as a market maker. The third and final bot is the "five-minute close bot." This idea emerged from observing a trader who consistently bought at specific price points (e.g., $0.98, $0.95, $0.94, $0.88, $0.99, $0.98, $0.96) towards the end of the five-minute markets. The "five-minute close bot" is designed to provide liquidity by placing bids at the very end of each five-minute period, mirroring this observed trading pattern. The speaker implies this strategy is "very interesting and very opposite" of the earlier discussed gambling-like trades observed. In summary, the speaker showcases three distinct bot ideas: a copy bot for replicating successful traders with a discounted entry, a spread-tightening bot for Hyperliquid's five-minute markets, and a liquidity-providing bot that trades at the close of five-minute intervals based on observed profitable patterns. They emphasize that these are ideas for others to build upon, stressing the difficulty and continuous learning involved in bot development, even with AI assistance.