Not to be argumentative, but there are always optimal play strategies. While they may not be unique, they tend to be relatively few, unless the player wins more or less no matter what they do.
It sounds like you're saying that the player has to be so responsive to chance item acquisitions that there can be no one size fits all strategy. It's an interesting goal, but I wonder how you can be sure the player can win with what he gets in that case without essentially requiring certain "kits" to appear in each game, if indeed the items are really different/not interchangeable. (Again, I'm sort of responding to the notion from the blog post...) In other words, can you really generate items with nontrivial interactions and non-interchangeable functions randomly and independently (in other words, without biases toward certain combinations, "kits," appearing together in a given play) in such a way that each play is almost surely winnable, not frequently too easy (e.g. you can often find broken combinations of items), and there does not exist a small set of strategies that are consistently optimal or at least game winning regardless of items (in other words, a number of more or less interchangeable combinations of items exist that collectively occur with high probability and lend themselves to a small number of winning strategies)?
My feeling is that this is too much to ask. If you have so many items that you don't even see all of them in a given play (even if short), all of which are useful and have some kind of synergy with other items (this usefulness and synergy being critical to a winning play), it sounds like it's a hell of a balancing act making these things work together properly. Has anyone really done something like this?
