15 January 2015

Appreciate What Miguel Gonzalez Is Doing... While It Lasts

Miguel Gonzalez's success with the Orioles has been baffling. He showed up out of nowhere and wasn't expected to provide anything, yet he's been pretty good at preventing runs. Since his arrival in 2012, he has thrown the third-most innings for the O's (420.2) and posted the second-lowest ERA (3.49) among starters, despite mediocre peripheral numbers and a FIP of 4.63 in 69 starts.

Other Camden Depot writers have established that Gonzalez has been an enigma, but it may be surprising that Gonzalez is the only player from 1901-2014 who has started at least 60 percent of his games and has thrown more than 400 innings with an ERA+ greater than or equal to 110 and a FIP greater than or equal to 4.50.

Rk Player ERA+ IP FIP From To Age G GS BB SO ERA
1 Miguel Gonzalez 116 435.2 4.59 2012 2014 28-30 75 69 139 308 3.45
Provided by Baseball-Reference.com: View Play Index Tool Used
Generated 1/14/2015.

Others have been close to Gonzalez's feat. The closest was Justin Thompson, who had an ERA+ of 118 and a FIP of 4.41 in 648.2 innings. Bill Walker, Jose Rosado, Wilson Alvarez, Monte Pearson, Wade Miller, and Mike Sirotka were also in the same neighborhood, but not as close.

For now, Gonzalez stands alone, which is both impressive and worrisome. How long can he keep it going? It's unlikely that he will continue to succeed at this rate. Regardless, he has been a fantastic find for the Orioles, is only 30 years old, and is under team control through 2017. He has been a tremendous value, and it's been fun trying to figure out how exactly he's been able to thrive. It's certainly a welcome change from watching talented prospects fail.

Photo via Keith Allison


Anonymous said...

Seems like the theme of the recent articles should be:

O's-All Smoke and Mirrors

O's-Only Dave Stewart likes Us

O's-Adam Jones Not as Good as You Think

O's-Brian Cashman says Not as Good as You Think Again


David Stone said...

I think the real problem is FIP as an accepted statistic. I get it, I get the theory: "let's remove all extraneous factors from a pitchers performance to see who they really are." Great, it seems like such a "clean" stat, how could you argue with it?
Well, here's a start:
To create this metirc, statisticians developed nearly the bluntest sledgehammer available. They include home runs, strike outs, walks, and hit by pitches. Only. This is not news to you, but i'm trying to form my point.
Recently, major league clubs have been looking at the effect of pitch-framing and how a catcher can effect the strike zone. This means that any walk or strike out that is attributed to a given pitcher has already been influenced by the defense around them, namely the catcher. Two of the four pieces of the supposedly defense-neutral stat are therefore at least somewhat definitely dependent upon the players playing with the pitcher being analyzed.
In the case of Oriole's pitchers, who pitch approximately half of their games in home run-friendly camden yards, a third piece of this very limited stat can also be called in to question.
This leaves us with hit by pitch as the only remaining "clean" portion of the stat. That's it. If you want to see what a pitcher can do independent of their teammates and ball park, the only bit of information that is utilized in FIP that you can actually rely upon to be based solely upon the pitchers ability is hit by pitch.
That is not very useful.
To create an actual working FIP, statisticians need to model what would happen to every ball hit by a pitcher given a league average defense, they need to model how every pitch would be called given a league average catcher, they need to model how every home run would end up given a league average fence.
In short, FIP is too broad and sloppy to be an effective metric. The thought is good, but it needs to be improved to actually become useful as more than an excuse for generalizations.
That being said, I believe that there is no way to create perfect "Fielding Independent Pitching," but we can certainly improve upon this crude tool.

Matt Kremnitzer said...

Anon: I can see why you'd think we have a negative vibe lately, but we're trying to be as objective as possible and have also written positive things as well regarding the O's lineup and their defense. How well the O's rotation has done the last couple years -- last year especially -- has stumped many baseball analysts. Obviously it's more welcome to overperform than underperform, but it's still worth investigating what is going on and doing our best to analyze the situation.

If you can figure out what Miguel Gonzalez has been able to do, that's great. He's currently doing something that no other pitcher has done. It's fun to watch, but I'm not sure I'd rely on it.

Anonymous said...

As stated by a previous commenter, I think a large part of the answer is that FIP is a garbage metric whose justification is largely based on hopelessly double-dipped statistics and extremely dubious heuristic notions of "things players have control over."

The mythology that there can be no sustainable and meaningful contribution from pitcher skill to outcomes when the ball is put in play (except for home runs...?) is silly and unfounded. I am reasonably certain that, were the hitFX data public, we would see that good pitchers' pitches are hit in substantially different fashion than bad pitchers' pitches. Gonzalez is a good pitcher.

Tim said...

FIP has utility. I am not sure why people claim there is no evidence when there are tons of studies like this one.http://www.baseballprospectus.com/article.php?articleid=12844

Anonymous said...

Tim: Because the data against which FIP is being tested are the same data used that were looked at while formulating it in the first place.

It's a major statistical no-no, but for some reason people who do sports stats (along with people who do observational research studies and people involved in "big data") seem to not know a damn thing about it.

Tim said...

No, anon. That is not the case with FIP measures. FIP that was calculated in say 2013 cannot include 2014 data. If you wish go back to previouschool calculations and compare to the following year without any future years' weights impacting those values.

Anonymous said...


You seem to not understand. I am talking about the metric itself being tuned for the data set against which it is being tested.

This is called "double-dipping." It is identical to shoddy "experiments" in nutritional science that look for correlations in large data sets without good external motivations, and then quote statistical significance figures for the correlations suggested by the data.

The *only* way to avoid this is to develop your metrics without looking at the data that you will use to validate them. This is clearly not done in most advanced SABRmetrics, largely because most of the people doing them don't seem to have solid backgrounds in mathematics or hard science.

Anonymous said...

As has been said, FIP fails to incorporate many factors. One of them that hasn't been mentioned is the infield fly ball rate. Gonzalez is one of the best in the business at inducing infield flyballs (5th overall in his time in the majors). At over 12% rate, it's an out nearly 100% of the time regardless of who's fielding. FIP hurts Gonzalez because he's not a strikeout pitcher and gives up too many walks. It's a bogus statistic.