You’ve continually attacked the validity and applicability of tRA and FIP, yet refused to explain your reasoning in any way that is logically valid. You continually hold up straw men or debate subjects not being discussed without ever addressing the crux of the debate head on. As soon as you state why you feel the statistics in question are internally or externally invalid, your stature in secondary education does not lend more or less credence to your position.

It’s essentially the area under the curve of the WPA for the team. It a pitcher gives up 50 ground ball singles in a row he’s going to have a brutal WPA despite pitching well.

I didn’t bring up tRA* because with such a small sample size, it tells us basically nothing about a pitcher’s likely long run performance.

EVERY pitcher will regress to the mean. In statistics, there’s no such thing as a 100% chance that anyone will do anything, which I’m sure you know already. However, we do know that some things seem to occur time and again a large percentage of the time. And the fact of the matter is, regression of the mean is one of those things because statisticians have determined that’s what usually happens over a large enough sample size.

So essentially, statisticians have determined that it is far more likely that a pitcher WILL in fact regress to the mean, which is why we always have to account for it. As to this, I’ll cite an excellent article by Dave Studeman of THT which explains this far better than I ever could

Is Marcel saying that each of these players will regress to the mean? Absolutely not. Some of them won’t. But enough of them will regress to the mean to validate the entire approach.

That I figure, is the most important concept to be taken from that article. Your high school pitcher example is irrelevant because in MLB, the members of the population are at very similar talent levels. Certainly Dana Eveland, probably a #3 starter at best, is closer to that of an average MLB starter than a high school pitcher. Certainly it’s valid to believe that he has a far far greater chance of regressing to the population mean than not.

We cannot empirically differentiate between random fluctuations, which you expect in any system, and results that indicate a trend. What we can do is take into account the likelyhood that a number reflects a trend vs. randomness, and this is the function of regression.

If you look at Eveland and Braden’s regressed tRA, they are virtually identical because the samples right now are so small as to be nearly meaningless compared to the established range of likely values. This is why I’m not, and have not, made any predictions about future performance based on the results so far this season. This is about who has placed the team in the most likely position to win. While clearly Eveland’s result has been poor, the degree to which plain bad luck is responsible for this result is very open to debate.

I think we all enjoy the game in pretty much the same way. We want wins. Stats are just describing the underlying process.