[mosdef]

as usual, the "debate" itself usually boils down to the wide variety of different used of various words on the forum, and as a result both sides of the debate are right because both sides are talking about two different things.

it also has a lot to do with the questions people choose to ask about their data. For example, you can play, let's say, 1000 tournaments and observe an specific ROI over those tournaments. This is really just one outcome for a random variable that had a large range of possible outcomes. there are two questions you can ask.

1. what does this mean with regard to my expected ROI (actual ROI, if you will) over my next set of tournaments. this is subject to the non-constants mentioned by the OP (i.e. developing skill and non-static opponent pool)

2. what does this say about my expected actual ROI (expected ROI) over the 1000 tournaments that i just played? the ROI random variable over those tournaments just played is what it is, and is not subject to the variations in skill and opponent pool that worry the OP.

anyway, for the umpteenth time, failure to precisely define terms and scenarios so that there is no doubt about what you're talking about is what causes the debates; after both sides have interpretted a statement their own way, they can both be right.