Illustrative figures based on logged runs
Passive quoting on the 5-minute markets loses on every exchange, at every size: the fills you get are the ones you did not want. The same seat on the 4-hour markets clears the bar, and the reason is arithmetic in the fee schedule rather than anything clever.
The mechanism
Liquidity rewards are paid as a rate per market per day. A 4-hour market therefore collects 48 times a 5-minute market's pool per window while trading only about 7.5 times as much, so the reward per share of forced flow is about 9.5 times higher. Adverse selection per share is unchanged; it is the subsidy that grows.
The seat
Quote both sides at the touch within the reward band, minimum 50 shares (smaller orders earn nothing), inventory capped at 100 shares, no gates. Every gate tried was matched by a placebo that knew nothing, so the ungated seat at the pessimistic queue position is the honest number.
Why it does not scale
So it is a $75-to-$160-a-day line on about $300 of capital, with a worst day near minus $200 and a worst single window near minus $360. It is real, it is small, and it depends on a share of the reward pool that no public data can confirm; the verdict survives that share being three times smaller than modelled, not four.
What kills it
- A second seat of the same size in the same market halves the pool share and fails the bar.
- The payout landing at the bottom of its published range.
- Fill losses 35 percent worse than modelled, which is inside the range real makers on those markets showed.
- An order path slower than a 250-millisecond cancel-and-replace.
Watch two API fields: the maker rebate rate and the reward band width. Either one moving changes the verdict.