Thanks guys. I'm more interested in getting a handle on the big picture before tweaking my game. I'd like to know what a good standard would be to set as my ITM/ROI goal. Since your ITM has a more direct relationship to ROI in HU than 6 and 10 man SnGs, the numbers play out differently for quality play(e.g. the 'good' ITM of 45% most SnGers play for gives you a -15% ROI in HU). Another aspect that's different is the rake which cuts into your ROI. Taking the rakes, entry fees and pize money at each pokerstars level, I've come up with the following:

ITM <-> ROI
52.5% = 0% (you need to play winning poker just to break even )
57.75% = 10%
60.38% = 15%
63% = 20%
65.63% = 25%
70% = 33.33%

My next question is what would be a good sample size? I've heard as low as 20 games, but I smell bs. I'd suppose the variance of ITM/OTM streaks could be the same as 6/10 man SnGs, but HU is more dependant on beating your opponent's style than beating the game, so perhaps an alternative sampling is needed...

What if we were to use something like sharkescope to break down each game/opponent into sub-categories along the lines of the following:

donk: <3% ROI
break even/average: 3-13%
solid: 14-19%
shark: >20%

These ROIs would assume mostly SnG play since no one here can point me towards purely HU players(If the numbers are off, they're only my guesses). It would be nice to know what percentage of each category a player's opponents tend to be at each level. If anyone would be interested in some research, shoot me a message . I can only do so much with 5 searches a day, hehe. I'd think that the ITM% you have against each category would give you a more well rounded indication to your level of play, and playing an equal(and by equal, I mean equal to the average of a given level) number of each category would give you a more truthful measure of ROI. What do you think of this sampling technique over, say, 200 games? Is this notion off the wall and a straight 200 games not taking into account your opponent equally acceptable?

In conclusion, what's a good ITM/ROI, what's a good-sized sample, and how would you go about getting that sample?