agent_pear pair trading bot posts 6,819 trades

Trader account @agent_pear said on May 15 that it had run a public pair-trading bot overnight and closed 6,819 trades across multiple exchanges and markets. The account said the bot posted a 64.8% live win rate overall and a 67.1% win rate on mean-reversion trades after applying cointegration, z-score and regime filters. In the same update, the account said it was holding 98 live positions across 94 pairs. Reuters could not independently verify the underlying trade log, fills or profit-and-loss from the post itself. ### What exactly did @agent_pear claim it did overnight? The May 15 post said the bot had executed 6,819 closed trades “overnight” in public and gave a breakdown of win rates and open positions, according to the account’s update on X cited in the card materials. The figures included a 64.8% live win rate overall, a 67.1% win rate for mean-reversion trades and 98 live positions across 94 pairs. Those numbers were presented as live trading results rather than a backtest, based on the wording in the post. The account also said the setup used cointegration and z-score filters, which are standard statistical-arbitrage terms for testing whether two assets move together over time and whether a spread has moved away from its historical average. ### How does Pear Protocol describe Agent Pear? Pear Protocol’s website describes Agent Pear as an AI-powered trading signal bot that watches market conditions and suggests entries based on mean reversion and correlation breakdowns between pairs. The site says Pear Protocol is a decentralized trading layer for pair trades that connects to venues including Hyperliquid and SYMM, and says users can go long one asset and short another in a single interface. Pear Protocol’s public homepage also shows rolling performance statistics for Agent Pear. A crawl captured on May 15 showed 5,369 total trades, 3,576 total wins and a 66.60% win rate on the site at that time, separate from the 6,819-trade figure cited in the X post. ### What do cointegration and z-scores mean in this context? Pear Protocol’s education materials say statistical arbitrage in pair trading focuses on the relationship between two assets rather than their outright direction. (pear.garden) The firm’s Sept. 1, 2025 explainer says traders look at correlation, hedge ratios, spreads and z-scores to identify when a pair has moved far enough from its historical relationship to justify a mean-reversion trade. The same explainer says a z-score measures how far a spread is from its historical mean and gives an example in which a trader would go long one asset and short another when the z-score falls below negative two standard deviations. Pear Protocol says the trade is then held until the spread normalizes. ### Can the reported performance be independently checked from public materials? (pear.garden) The public materials reviewed by Reuters did not include a full trade blotter, timestamps for all 6,819 closed trades, realized profit-and-loss, fees, slippage or exchange-by-exchange execution data. The X post in the card materials gave headline figures, while Pear Protocol’s website and documentation described the product and displayed separate aggregate metrics. (pear.garden) Pear Protocol’s documentation also says agent wallets are used to execute trades on Hyperliquid on behalf of users and are valid for 180 days with automatic 30-day rotations. That documentation describes wallet mechanics, not an audited public record of the May 15 results. ### Where are readers likely to see the next update? Pear Protocol’s app includes an Agent Pear page, and the company’s website says Agent Pear is being rolled out in stages with on-demand statistics for users evaluating specific pairs. (pear.garden) The homepage and docs were both publicly accessible on May 15, and the X account that posted the 6,819-trade claim is the most immediate place to watch for any follow-up screenshots, trade logs or updated open-position counts. (docs.pearprotocol.io) (app.pear.garden)