In the snowie faq they have a write up explaining the inconsistencies that occur between the move advice given and the range displayed for villain:

First of all it has to be noted, that the evaluation (the move advice) is NOT based on the hand range of the opponent. The evaluation is the output of a neural network and has evolved during training. The neural network has been trained over trillions of hands, against various counter-strategies. Therefore these evaluations are very robust and can be trusted most.

Surely this is an admission that snowie is some distance from nash equilibrium? And what about this idea:

two slightly different neural nets that play almost identically may have significantly different hand ranges

I can only see this as true if 1) there are literally different snowies with slightly different EV evaluations causing strategies to diverge by the river or 2) some sort of rounding error causes the villain range advice and hero strategy advice to diverge as the hand progresses.

Either way we have to put a limit on how much attention we pay to range advice in close spots because we can never have the confidence that we get from a true equilibrium solver and therefore miss out on some of the nicer ways to visualise a hand range, such as the nice charts and graphs we get in Cardrunners EV.

Thoughts welcome.