I've been playing around with pokersnowie checking it out.  In another post I want to give my first thoughts on pokersnowie.  However first I think I found one instance where it is actually playing sub-optimal even by its own limitations (.5,1,2 pot and all in bets only.  Can also only choose one bet size for its bets in a certain situation).  Wanting to confirm this is an error or find out why I'm wrong.

------------------------------------------------------------------------------------------------------------

if you need to download the program: www.pokersnowie.com

Instructions to view error (assumes you have the pokersnowie demo or full version open)

1. Click import on top bar

2.  Select sample and click 'view analysis'

3.  Click blunders

4.  Click the forward button above hand history till at blunder 2 (hand 4906512470)

5.  There should now be a $230 bet by Ad5d vs 9c9s into $400 pot on a Ks3s2d pot.

6.  Note that pokersnowie considers the 99 an error to call and an even bigger error to all-in.

7.  Click analysis tools

8.  Click view ranges.  Note the hero's range is QQ-55 and AQ, and that fold is the main choice.

9.  Close range advice

10.  Select the "Player-_-4 bets $230" spot in the hand history.

11. Click view ranges.

12. Select "Bet 1/2 POT 27.05%" and then Suggested action.

13.  Note that he is betting KK and a lot of high card hands (18.88% KK, 81.12% other).

----------------------------------------------------------------------------

Analysis:

First of all you should ignore the Ad5d in Player 4's hand.  It is not a hand pokersnowie recommends 4-betting pre-flop in this situation.  What we are concerned with is the how it plays the ranges it recommends to have gotten to this point in the hand.  

Its recommended betting range for player 4 is super polarized to being about 19% top set kings, and 81% being no pair bluffs.  This could easily be exploited by hero shoving his entire range (which none of player 4's bluff bets have equity to call) of 55-QQ,AQ over the 1/2 pot bet.  This is solely a protection raise as the only hand calling is KK which crushes every hand in his range, but because the bettor has so heavily weighted his hand towards bluffs simply shoving and taking down the pot 81% of the time becomes very profitable since the shove only needs to succeed  832.75/1462.75 = 56.9% of the time.  Whether or not calling with some of the range would be even more profitable cannot be confirmed (though tend to think its unlikely and would require player 4 to continue bluffing too much on future streets), but shoving all of hero's hands would be within the restrictions of pokersnowie's system and would be more profitable than the recommended fold.

So if this is correct Pokersnowie is not perfectly optimal even within the restrictions upon it.  I cannot say for sure whether it is a flaw in the neural network system that makes play adjustments or that even with the mass training it has done that dealer raise, sb 3 bet, bb 4 bet, sb calls, flop similar to Ks3s2d just doesn't come up often enough for this specific situation to be well trained.   I also do not know if this is an isolated case, or if there are mistakes hidden about in pokersnowie's play.