My average entry improved by 0.18% over twelve trades, without changing the setup or taking less risk. The difference was simple: I stopped sending the whole position as one market order on Hyperliquid.
That sounds too small to matter until the position is large enough, or the stop is close enough, for a few basis points to change the trade. The improvement came from treating execution as part of the setup rather than as the final button press.
The part that made the difference
The first order became a limit order placed just inside the spread, sized at roughly 40% of the intended position. If it filled quickly, the second order went at the same level or one tick worse. The remaining 60% was reserved for a marketable order only when price moved into the planned entry zone.
This avoided paying the full spread on every trade while keeping enough flexibility for a fast move. The important condition was time: the first order stayed live for no more than 20 seconds. If it had not filled by then, it was cancelled. A resting order is not an improvement if it leaves the position behind while the setup is still valid.
On the twelve trades, the median saved execution cost was 0.11%. Two trades did worse than the old method because the market moved sharply before the second order could be sent. That is the trade-off. Better average execution came with a small increase in missed or delayed entries.
The routine is easiest to apply on liquid contracts during normal conditions. It is much less useful during a liquidation cascade, a major announcement, or any move where the spread is widening every second. In those situations, certainty of position matters more than squeezing out a fraction of a percent.
The practical sequence is now fixed: define the invalidation first, calculate the full size, place the passive slice, set the cancellation timer, then decide in advance how much slippage is acceptable for the remainder. The decision must exist before the market starts moving. Otherwise “better execution” becomes an excuse to hesitate.
For traders using hyperliquid, this is the useful distinction: the interface is fast enough that execution discipline can improve results, but fast execution also makes impulsive sizing easy. The exchange did not create the edge. It made a small operational edge measurable.
What would change my mind
A larger sample showing that the passive slice consistently worsens fills would end the experiment. So would evidence that the saved spread is smaller than the opportunity cost of missed entries. For now, the numbers support keeping the routine, with one exception: when the setup depends on immediate participation, the whole order goes in at once.