This indeed hits the key point. The initial yes or no inventory status is essentially uncertain, and I have encountered this pitfall when doing arbitrage. The solution I finally devised is to perform buy up and buy down operations simultaneously, combined with periodic merge settlements. This way, we can bypass the dependency on inventory status.

From a code logic perspective, there's no need to judge or maintain the yes or no inventory, because both directions are active, and the merge step in the middle will automatically balance out the redundancies. The advantage of this approach is a simpler process, avoiding the hassle of state issues during initialization, and it also makes the smart contract easier. If you try to determine the inventory status from the start, it can instead introduce bugs related to edge cases.