First off, ten thumbs up. This looks fan-****ing-tastic so far.
There are a few things I though of so far...
1) Obviously, +$EV pushes that don't get called won't be factored in. Also, +$EV pushes that get called by better hands will decrease a player's expectation even though the push was correct. The opposite is also true, but it's much less common to have a bad push called by an even worse hand. The result of all of this is I would expect even a very good player to have a long-term expectation below zero.
2) I wonder a bit if it will underestimate your "expectation" when you are running bad and overestimate it when you are running good. If, on the first hand of a tournament, you get AA vs. KK and lose, so the program correctly captures your -15% or so. The problem (I think) is then it's basically capped your equity for that tournament at a max of 18% and moved on. Does this make any sense?
I had one forced-quit and dwwin.exe didn't want to let go of about 50% of my CPU (about what SNGLA took while computing). It's probably a lolwindoments, but thought you'd like to know.
Lastly, I run g00t. The bottom plot is how g00t I run when making awesome calls against bad pushbots. The numbers aren't quite right because these are 6-max and I believe the 50/30/20 structure is hardwired in as of now, but you get the idea.