Look. Competitive games with any element of variance in them have one underlying principle: If you are winning, reduce variance at all costs. If you are losing, INCREASE variance at all costs. The reason for this is that losing badly is the same result as losing just barely, so if you know your current position isn't winnable there's no reason not to gamble everything to try and improve it.
Yes, this card averages a negative return in terms of raw numbers. But you have to look at the return UTILITY. Let's say the corp is 6 points up with a probable agenda advanced once in a server you need 20 credits to crack. You have 13. What's the utility of those 13 credits? Nothing. They're useless to you. You lose whether you have them or not. So Push Your Luck in this situation is gambling an item of no utility for a chance to win an item with actual utility.
However, that doesn't make this a good card. The simple reality is that in almost every scenario where you're almost guaranteed to lose with the credits you have but might stand a chance with just under double that, Stimhack performs just fine and without the horrible variance. No, while I can see the point of this card (take the gamble only when you have nothing to lose, turning a guaranteed loss into a 50/50), the payout isn't high enough to justify including this.
Which is sad, because the concept and usage would be neat, but understandable because randomness at a high competitive level annoys the hell out of competitive players. Psi gets a pass because it's like a yomi game, but this is really just a coin flip.