采用 Joshua Della Vedova 的机器人定义（任意活跃日交易次数超过 50 次，或在 2025 年 1 月 1 日至 2026 年 4 月 8 日期间累计交易超过 1,000 笔），机器人式钱包的交易活跃度远高于散户钱包，平均每个活跃交易日 89 笔交易，对比 2.2 笔。这个小群体产生了大部分交易量，并且合计净赚 131,000,000 美元，利润集中在 823 名每人盈利超过 100,000 美元的用户。更少交易的用户以及许多其他普通钱包整体上也亏损了大致同样的金额。顶端 1% 的交易者获得了约四分之三的总利润；Martineau 较早的研究（自 2022 年以来）已显示 69% 的交易者在亏损。
数据还显示原因：机器人式钱包并非主要靠更好的预测获胜，而是通过更早进入市场并获得更好的价格。散户有时预测更正确，却因为迟到、执行价格差不理想而支付高昂代价，累积亏损达数千万美元。Polymarket 在实践上也类似传统交易所：market makers 和高频自动化者可以占主导，而不存在明显庄家优势；而预测市场中错误押注的下行风险与任何赌博相似，一笔仓位可能损失 100%。再加上 Polymarket 的数据显示，仍缺乏其用户明显比 Kalshi 或其他平台更差的证据。因此结合 Pat Akey 与 Martineau 的研究，预测市场可被视为一种工具，但不是可靠的“补房租”方式。
Social media often promotes prediction markets as a side income, yet Bloomberg’s blockchain-level analysis of Polymarket activity from Jan. 1, 2025 to Apr. 8, 2026 reached the opposite conclusion. More than 100,000 wallets lost at least US$1,000, almost double the number that gained at least US$1,000. Nearly half of about 2 million active wallets lost or made less than US$10, showing broad experimentation; excluding a very small winning core, the rest together lost US$131 million in total.
Using Joshua Della Vedova’s bot definition (more than 50 trades on any active day, or more than 1,000 total trades during Jan. 1, 2025–Apr. 8, 2026), bot-like wallets were far more active than retail wallets, averaging 89 trades per active day versus 2.2. This small cohort generated most volume and collectively netted US$131 million, with gains concentrated in 823 users earning over US$100,000 each. Less active users and many other regular wallets lost roughly the same amount overall. The top 1% of traders captured about three-quarters of total profits; Martineau’s earlier work (from 2022 onward) had already shown 69% of traders losing money.
The data also explain why: bot-like wallets did not mainly win from better predictions, but from entering markets earlier at better prices. Retail traders sometimes predicted correctly, yet paid a high cost from late entries and weak execution, losing tens of millions in total. In practice, Polymarket resembles traditional exchanges where market makers and high-frequency automation can dominate despite no explicit house edge, and prediction markets carry a full-loss risk equivalent to bad wagers, with a position able to lose 100% of value. Combined with findings from Pat Akey and Martineau, prediction markets can serve a function but are not a reliable side-income method.