Yeah i remembered... isn't there a very simple way to calculate how much variance should you expect? You can calculate standard deviation (statistics). It will not be me because i don't remember how to do it and i don't want to kill my brain...

Anyway, made a little program, i think it's rigged lol. It should generate about 3.35 BB/100 but it gets around 1.8 and can't explain why so i won't post it yet. If you'll take my word for it (and if i wrote it well, which i doubt) things stay this way (10 simulations for each sample... humanly speaking it follows the difference between the luckiest and unluckiest from 10 people doing the same thing)

Over a 1K sample: best 5.5bb, worst -2.2bb

Over a 10k sample: best 2,39bb worst 0.41bb

Over a 100k sample: best 2,369bb worst 1,357bb

Over a 1M sample: best 1,969bb worst 1,7516bb. So i guess what they do is worth about 1.8bb/100 hands. It's so strange, even in a million hands (!!!) you get a + or - of 0.1bb/100hands. At least over the recommended 20k sample you can't expect to have a deviation from your real win rate bigger than 1.5bb/100hands. You have like 0.5?! Oupsyy...

And in case you wonder how i got the numbers, it's a process like this: it's a 30% chance of winning 10bb, a 20% of winning 0bb, 5% chance of losing -15bb etc... (God i am really that bored?)