Copy-trading the most profitable Polymarket sports wallets doesn't work. We tested it...
The daily research thread from our internal kill ledger.
Posted first to X.
Copy-trading the most profitable Polymarket sports wallets doesn’t work. We tested it four independent ways, out-of-sample, look-ahead-free — and lost money every time.
The premise was reasonable. Some wallets consistently show profit on slow sports markets — US sports, soccer, tennis — where copy latency isn’t fatal. Mirror their fills, capture the same edge.
We built a look-ahead-free event-driven copier: match each buy and sell at fill price plus a slip allowance, pay the 3% taker fee, settle open positions at resolution. No mint channel. No survivorship shortcuts.
The structural problem: the profitable wallets’ edge isn’t directional. The copyable long-side cohort of top sports wallets loses $1.1M in aggregate. The short/mint side makes $760K — by selling minted inventory that never appears in order flow.
In-sample selection looked good — one formulation hit +9.7%. We caught a bug before deploy: the copier’s “profit” exceeded the wallets’ own realized returns, which is impossible. The honest out-of-sample result: −8.6%.
Track records don’t persist. Wallets selected on historical quality go dormant or flip negative within months. Past copy-profit predicts future copy-profit roughly 28% of the time.
If the edge is uncopyable, mirroring is just paying fees to lose.