I think the problem is that it only calculates the EV for that particular situation on that particular street, without considering other important factors like implied odds, game flow dynamics, future streets,and opponent player type adjustments.
It also has certain tendancies that are somewhat easy to exploit. For instance in HU, on dry flops like J62 or TT4, its betting range is very polarized, as it will check behind with mediocre hands like 2nd pair or worse. When it bets it either has top pair or air, and check raising your entire range is profitable since the vast majority of the time it has air. This is especially true on paired boards, I generally check raise 100% and it never realizes its being taken advantage of. Then when I review the analysis, it tells me that my check raise with 75s on an AA3 should have been a fold 100% of the time, even though a check raise is profitable 100% of the time against snowie.
Also some of the recommendations it makes are questionable. For example playing HU:
1) I raise Q
8
OTB, snowie calls, flop comes J
6
4
. I bet he c/r, I fold. It recommends calling. Not sure why it says call especially since there aren't very many semibluffs in his range, and my 8c blocks some them.
2)Snowie raises OTB I defend with Q
9
. Flop comes J
8
6
I c/r. Snowie recommends calling 100%. Although I don't always c/r here, I think its a fine spot considering my pot equity+fold equity combined, as his range is extremely wide and I block many of the hands he can continue with against a c/r.
3) Snowie raises to 3BB OTB I defend with A5s. It recommends 3-betting 100%. Pretty meh if you ask me considering I'm OOP and no worse aces will call my 3-bet. This hand would rather see a flop, keep the pot small, and hope to stack him when he makes a worse flush.
4) Snowie raises OTB to 3BB, I fold J2s. It recommends calling. Meh
5) I raise J9o to 3BB OTB, Snowie 3-bets to 9BB, I fold. It recommends calling. Pretty meh considering snowie's 3-bet % isn't all that high.
I think these examples show how snowie only looks at individual EV decisions without considering how the entire hand might play out.