I think the posters in the other thread are right about the fact that you might not be able to use formulae that assume a normal distribution for this small of a number of hands. But I don't have other tools handy, so, we can look at the normal distribution and see.
So OK, let's say 80 hours, 5bb/hr win rate, and let's start with 50bb/hr standard deviation and see what that gets us.
http://rustybrooks.com/poker/bankrol...&Gogogo=Submit
So that gets you 750bb for a bankroll. I think that might be too low for stdev though, so maybe try 80...
Now we get 2000bb (20 buyins) which is probably more realistic.
However, this is a bankroll given the consideration that you're going to play, essentially, forever.
We can look at a different kind of calculator instead, such as
http://rustybrooks.com/poker/stats.h....x=&outside.x=
This one is basically saying, with a given win rate and stdev and # of hands, what are the chances of a particular statistical outcome. Note that I converted both WR and stdev to /100 instead of /hr by assuming 35 hands/hour.
This unhelpfully gives the probability of ending at worse than -1000bb as zero. I think this is because the chance is very small. HOWEVER, this calculator is not including the times you dipped below -1000 and came back. It's just saying that the chance of ending up worse than that is very low. If you look at the chance of ending up at -500 or worse, it's like 1.5%
I think this really shows the weakness of normal models for something like this, because I think the chance of losing 5 buyins is probably way more than 1.5%
I hope this gave you something to think about, but I'm sorry I can't give any actual answers.