Resting a quote on both sides of a Limitless market for an hour

The profitable number in the plan was computed at a two-second reaction time. The design the code actually implemented quotes once per market per hour and never reprices. Those are two different strategies and only the first one makes money. The bind underneath is structural: quote fast enough to avoid being picked off and you almost never fill, quote slow enough to fill and you are picked off essentially always.

45,963 two-sided observations across 130 markets on the hourly book, against a half-spread of 0.025. Breach rates by how long a quote rests: 12.4% at two seconds, 60.4% at one minute, 74.8% at five minutes, 85.3% at fifteen, 93.8% at thirty, and 100.0% at fifty-nine minutes. Median fair-value movement over thirty minutes is 0.235, about ten times the half-spread being captured. Simulated against the real trade tape over 11.7 hours, the paired fill rate is 0.0% on both the daily and hourly books, so the deterministic exit almost never completes. Capacity is the second wall: 0.99 fills per market-hour, a median trade of 9.4 shares against a 100-share venue minimum, so 79% of the flow cannot absorb our smallest possible order, and the market in question turns over $18 a day against a single clip worth roughly $50. Edge and volume run in opposite directions: the largest market has seven times the volume and one-fifth the edge and turns negative within thirty seconds, so you cannot scale into the liquidity. Disclosed limitation: the recorder never captured a trade signal, which is why the pre-registered fill-rate criterion never accrued, and the realised-edge-given-fill section rests on 5 to 31 observations and is indicative only.