Well you are maybe right that I should have stated it differently : there is no actual "flaw" in it, it has a rigorous mathematical definition. The flaw, as usual with statistics, is in its interpretation. I think I already explained why.

As for the rest, yes adjusted ROI will, on average, be closer to your true ROI of course. Just don't go too far with the interpretations, I also gave examples in one of my replies.

First, I love the post. I have a statistical question. I understand truncating all losers for estimate ROI. However, if we want to look at skill, we need to have confidence intervals around places 1, 2 and 3. I would argue that these intervals are conditional on the probability of reaching the top 6.

In a non-statistical analysis, I look at my position finishes. I expect 9th to be the lowest, followed by 8th and 7th. In fact, in my case, these add up to less than 25% for the stakes I currently play.

When I look at 1-6 I get a relatively uniform distribution (every place within 1 or 2%). Perhaps I should have more success once I get to 6 players left, however, that is a different point.

Do others have similar position stats? If we want to set confidence intervals (and thus standard deviations) shouldn't we look at position finish rather than money won?

Hoping someone more sophisticated in statistics than me can address these questions.

Thanks