[Lori]

What I mean is that your sample could be:

270 small pots: break even
70 large pots: lose lots

in that example, you haven't really had 340 hands of bad luck, just 70.
Averaging it doesn't work because it inflates the sample size and makes you look like you've had a sustained bad run (assuming the guess that you've run bad in big pots is corret), rather than just a 'standard' bad run.

Edit: ICM would still carry the same problem that some hands will be more important than others.