A viral X post claimed BTC Up/Down 5m contracts on Polymarket were mispriced by 3–5...
The daily research thread from our internal kill ledger.
Posted first to X.
A viral X post claimed BTC Up/Down 5m contracts on Polymarket were mispriced by 3–5 points; we tested it against 41,317 resolved markets and refuted every edge claim.
Their sample: ~1,000 markets, 3.5 days. Every mispricing they named was 3–5pts. Per-bucket n was 100–250, putting binomial SE at 2.5–4pts. The claimed signal and the noise floor were the same size.
The structure replicated: up-share 50.4% (they had 50.3%), 91% of opens within 5c of 50c. A 2c lead at bar open wins only 52.5% — a coin flip. By 4:30 into the bar it’s 88.6%. Same shape, same regime.
The main claim: leaders at a tight band underpriced by +3–5pts at minute 2. We found the sign flipping across adjacent buckets and across time slices. Backtest at an 81c average entry: EV –0.5c gross, –2.5c with a cent of slippage.
The second claim: high-confidence leaders overpriced. In our sample the sign was opposite — 95c+ leaders at minute 2 won 98.6% against a 96.5c price. Every fade attempt was negative gross.
One pattern held across every time slice: near-50c leaders underperformed their price by about 4pts. It’s a methodology artifact. Near 50c a taker-buy print at the ask mislabels the leader; crossing the spread eats the apparent edge before the trade is on.
If every claimed edge is the same size as the noise floor of the sample used to find it, it’s noise.