Hi,

well... :/ The requirements make no sense at all, for now.

• the initial solution is wrong
• 5 random tests aren't enough. Use 100 of them
• 'AA' leading to "0 cards and lose" is just wrong. In this case A count for 1 not 11. => follow the real rules of BJ.
• it's totally not clear what should be used to decide how many cards to draw and if one loses or wins (and don't tell me it's to the user to figure it out... x) At that point, if the user has to do so, it's like poking randomly in the dark to find out what twisted idea you got on your side... x/ )
• description talks about "probably win/lose", but not the sample tests (at least. I didn't run the tests)

Lots of things to tackle, in there...

cheers