So we can look at combining a few variables.

But first, there’s this

There’s some pretty interesting things there. But there’s also a big problem with it, though — he doesn’t tell us how or what he did to get his information. So, how much does that help us?

So instead of just the one variable, we can use the categories you used for the fifth starters: strike%, swing%, GB/FB ratio. The top half of pitchers in both strike and swing %, sorted into three equal groups for GB, median, and FB pitcher types:

GB RAA: 4.33
MG RAA: 4.59
FB RAA: 4.60

This is all pitchers over 100 innings over the last three seasons, from FanGraphs. It’s a pretty good size group, but not large (76 pitchers). No adjustment for park or defense, so hopefully that mostly evened out.

So there is a difference, about a quarter run a game. Is it significant? Well, maybe. If you chose a fifth starter based on a higher GB rate, with all other things being equal, and he could stay in the rotation all year, that would add up to about 1/4 to 1/2 of a win. That’s not nothing. But it’s also not a whole lot, either, and would depend on the 5th starter pitching a lot. And, could quite possibly be mitigated by pitching in Safeco.

But what you said before is right: everything is interconnected and we can’t expect that we can change one variable and not affect a whole bunch of others. There’s just a whole lot going on there and it’s too simplistic to say that giving up groundballs is better than giving up flyballs. There’s no pitcher that just gives up one or the other, and there are a bunch of tradeoffs in between that impact a pitcher’s ability and results. It’s almost impossible to expect we could seperate all those out from a pitcher’s style of pitching to be able to compare one to the other in such a way.

But what selection bias were you refering to?

The only “balls in the air” that are helpful are popups but its a pretty small percent and fluctuates a lot.

No need to be denigrating. Being informed also means being able to look at someone else’s point of view and consider what merit it might have without first dismissing it out of hand.

Also true: He has spent the majority of his career walking near or above the league average walk rate. Also true: He is a freak, and a great example of the kind of approach that you’d never teach but accept from him because he makes it work thanks to his freakishness.

One example isn’t proof. He’s an outlier.

Because like I just said, GB/FB is not the only thing that matters, it’s just an important part of the overall picture. Pitchers do not succeed because they give up a lot of flyballs, they succeed in spite of it. Like Johan Santana, for instance, strikes out so many people that fewer balls are put in play in general, reducing the number of homeruns he would likely give up with a batted ball profile like his. If you throw strikes or miss a lot of bats, you can still be a successful pitcher, but the fact of the matter is that a flyball put into play results in far more runs than a groundball put into play, therefore assuming that a flyball pitcher and a groundball pitcher are identical in all other aspects the groundball pitcher will be better.

Matthew listed ground balls along with throws strikes and misses bats. I was pointing out that if you have two pitchers with roughly the same ability to throw strikes and miss bats, you’d rather have the one who can keep the ball on the ground.

 As to how much a greater ability to throw strikes or miss bats can compensate for a lesser ground ball rate, I do not know. But say he had just listed “throws strikes” and “misses bats.” What I am suggesting is that this is not enough information to judge pitchers, because using just those two categories we would conclude that 2005 Felix was slightly better than 2006 Felix, rather than the reality that 2005 Felix was quite a bit better than 2006 Felix. Thus, the ability to generate ground balls should be listed as well in order to get a better picture of a pitcher’s ability to prevent runs.