The two methods mentioned here are quite flawed:

Kerbonut mentions the use of the signature of a gameserver to verify the loot. This obviously is not decentralized, and is essentially the same thing as having a centralized itemserver. This is part of what I mean when I said that "their development structure is not decentralized enough to warrant the use of cryptocurrency"

lmm mentions the use of a "proof-of-victory" given a random in-game challenge produced with a provably-fair RNG seed. Not only would this require a ton of bandwidth and CPU to store and verify these proofs (unless it's a turn-based game like chess or nethack I suppose), but it is not really sybil-proof. In other words, it puts people who play the game at a significant disadvantage to those who just bot. It incentivizes not actually playing the game.

Edit: note that this Decentralized-WoW thought experiment is fundamentally different from axie infinity. The case of axie infinity is a two-player game where they essentially gamble using the depreciation cost of their pokemon. It falls under the category I described as "gambling on the outcome of a game"