https://vitalik.ca/general/2019/04/03/collusion.html

In the first half he nicely elucidates concrete reasons behind an intuition I've always had: that using game theory to build decentralized financial systems always suffers because they have a dependency on identity (and how I'd put it: the problem with that fundamentally, comes from the fact that identity is a fiction, an arbitrary and unanchored construct).

This problem manifests as the impossibility of avoiding collusion in various forms.

Quote:

"But in the version of game theory that allows for the possibility of coalitions working together, called cooperative game theory, there are large classes of games that do not have any stable outcome that a coalition cannot profitably deviate from."

This is the problem - a lot of academic game theory posits isolated actors, however that is *never* the real world (see: sockpuppets or simply, communication!). That's why I've always told people, just like Nick Szabo did to Manfred Karrer back in the days of bitsquare, "don't rely on game theory, replace it with cryptographic verification").

He then correctly identifies, in the middle of the article, the best solution to the collusion problems described: proof of work, because it is identity-less.

The remaining part of the article feels like a reach, looking for increasingly Byzantine (pun intended) partial solutions to what he clearly understands to be an insoluble problem.