I'm happy to hear from the poker community, but you must appreciate this is research. I do have a great understanding of the game of poker and the dynamics of the game. Therefore, it was a conscious decision to use a simplified version of poker (i.e., no position - heads-up, no re-raising, no stack-size considerations, no flop, turn, river, no feedback about outcomes to develop opponent models, etc.).

Although this greatly simplifies poker, it gains experimental control: we can more cleanly interpret our findings. What if I would have allowed for the addition information and found the same results? What could I say? At the end of the day, this is research! I must control for variables that I am not manipulating, so that I can interpret my findings.

Your comment that people should fold 99% of the time doesn't support your comment that people should have believed their opponent was random: I designed the hand distributions to have equal numbers of *optimal* call/fold decisions (against a random opponent). Therefore, folding 99% of the time, would clearly be a suboptimal strategy! In fact, that's why I used such a high raise amount: as bet size -> infinity, the optimal decision boundary approaches p(win) = .5.

What you may find interesting is that some people (perhaps like yourself) report not using face information at all. However, upon inspecting their data, it was clear they still folded at higher rates against a trustworthy appearing opponent, suggesting this effect may have an implicit factor.

The message I'd like you to walk-away with is that my decisions to simplify the poker game wasn't due to my lack of knowledge about the game, but rather my appreciation for its complexities. Good luck at the table!