Conditional re-quoting on prediction-market crypto pairs is worse than static quoting...
The daily research thread from our internal kill ledger.
Posted first to X.
Conditional re-quoting on prediction-market crypto pairs is worse than static quoting at every tested price.
The idea: use the first fill as a signal. Whichever leg is hit first tells you direction. Cancel the other, repost it lower, collect on a corrected price. Smaller orphan exposure, we thought.
We ran two repricing rules — fixed discount and targeted pair sum — across two latencies and roughly 45,000 observations. Symmetric and asymmetric anchors. Both arms pre-registered before running.
Every cell negative. Losses on wide-market fills ranged from −6.39c to −7.68c per trade. Conditional was worse than the static baseline at every matched price — by up to −2.59c.
Why: leg one fills near 0.50. You are now holding a residual in the near-even region, where the expected value for a maker runs about −1.78c. Static orphans a 35c bet. Conditional orphans a 50c bet.
Three programs, ~80,000 bar-level observations. Programs one and two closed static quoting; this closes the conditional form. The overall maker fill average on these books remains positive — the pair is not where the edge is.
The first fill is not a signal. It is confirmation the market has moved against you. Repricing around it does not reduce the orphan. It increases it.