How Good Is Matt Garza?

The Cubs need to prioritize. As a struggling franchise, they need to significantly overhaul their roster and farm system. A possible trade chip, as Buster Olney wrote last Monday, is Matt Garza. This year was excellent for Garza; to highlight a few of his achievements, he posted career bests in WAR, FIP-, and xFIP-.

Much of his success can be traced to an improved strikeout rate. He struck out opposing batters 23.5% of the time, placing him right in between Roy Halladay and C.C. Sabathia for 12th among all qualified starters. In addition to an increased strikeout rate, we also know that he has been depending much more heavily on his off-speed and breaking stuff. Therefore we can attribute his great 2011 performance to his decreased usage of his fastball.

But is it that simple?

Just two years ago, Matt Garza posted a strikeout rate nearly as good as in 2011 — 22% of all PAs. Factor in the difference in run environments and the switch from the AL East to the NL Central, and his rate from this year is no longer unprecedented. He wasn’t limiting his fastball usage either back then, throwing about 70% fastballs according to my classifications, so it’s not as simple as saying that a drastic change in his fastball usage definitively led to the drastic jump in strikeout rate. He’s seen this kind of spike before, even without changing the amount of fastballs he threw, so we have to look deeper to see if there’s anything else going on.

We like to think of strikeouts as immune from luck. They are paramount in sabermetric analysis of pitchers, and have a very strong relationship with all major run estimators. And while the importance placed on strikeout rate is the correct decision for these metrics and analyses, it’s important to be mindful of the fact that lucks plays a role in strikeout rates, too. No, strikeout rate is not nearly as inconsistent as BABIP or HR/FB, but there are times when a pitcher’s observed strikeout rate significantly diverges from his true-talent strikeout rate. This is because the ability to record strikeouts is not really one ability, but a composite of skills. The pitcher’s prowess in getting swings and misses, called strikes, and even expanding the zone all affect overall strikeout rate. Of course not all of these skills are equally important.

A skill that tells us a lot about a pitcher’s strikeout ability is whiff rate — the amount of whiffs divided by the total number of pitches thrown. Here are Garza’s whiff rates and contact rates (whiff/swings), split up by year:

year whiff contact
2008  0.08    0.83
2009 0.08    0.81
2010  0.08    0.83
2011 0.12    0.76

With a huge jump in strikeout rate we would expect to see a corresponding increase in whiff rate — and that’s exactly what we see with 2011. There is a clear improvement in his ability to get swings and misses in some parts of the zone:


The extra whiffs that he’s getting in 2011 are right on the outside edge of the strikezone to a right-handed batter — the location where he’s throwing many of those additional breaking balls. Dotted lines indicate the strikezone and the bands indicate confidence.

But we still haven’t explained why his strikeout rate was so high in 2009. Turning our focus to called strike rates, we find the following proportions of plate appearances that ended in a called strike three (denoted by cs rate):

year cs rate
2008 0.05
2009 0.06
2010 0.04
2011 0.04

But the opportunities in which Garza can record a called strikeout are affected by other variables — like whiff rate — that we want to ignore. If we only look at pitches that can end up as a called strikeout — pitches thrown in two-strike counts that are not swung at — we find the following proportions:

year cs rate oppurtunities
2008 0.13    296
2009 0.12    403
2010 0.09    385
2011 0.10    305

A visual examination yields a bit of a mess:


Again, only looking at pitches here that could potentially result in a called third strike.

The table helps to explain some of the variation in Garza’s strikeout rate. While his called third strike rates were similar in 2008 and 2009, he had many more opportunities for such a result in 2009. This can be explained by the fact that more of his pitches came in two-strike counts in 2009 (27.6%) than 2008 (26.1%). In 2010, he simply did not have as many of his potential called third strikes turn into actual called third strikes. This had a significant effect on his strikeout rate; if we apply his 2009 called strike three rate to his 2010 number of opportunities, we would expect him to have 46 called strikeouts — 12 more than he actually had. These 12 additional strikeouts would have bumped his strikeout rate from 17.5% to 19%. Given the small samples at play here with called strike three opportunities, luck can have a larger role than what we expect. A significant test between the two proportions (2009 and 2010) yields a result that is not significant at a 95% level.

Often times explanations are simply drummed up to explain so-called “breakout” years. Many of these are simply the result of our own cognitive dissonance — it just feels so wrong to a attribute a breakout performance to luck, so we craft our own narratives to reduce the mental tension. Most of the time these explanations are just that — rationalizations with a sprinkle of truth and a dollop of imagination. But not every time. Jose Bautista, for example, completely revamped his swing. Matt Garza looks like he could be another legitimate improvement. Is he the 5 WAR stud he was in 2011? No, we can’t just throw out the rest of his career. But there is reason to believe that he is better than he was before. His peripherals entirely support his performance and can be explained by a shift in pitch selection. Of course it’s possible that the league adjusts in 2012, but there is no reason to dismiss 2011 as a fluke.

References and Resources

*PITCHf/x data from MLBAM via Darrel Zimmerman’s pbp2 database

*Plate discipline statistics calculated by author

Print This Post

21 Responses to “How Good Is Matt Garza?”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. chel says:

    Very interesting stuff, have you read about swing area? (
    I would like to see this exact same study about Bartolo Colon

    Vote -1 Vote +1

  2. Paul says:

    No. No. No.

    The answer is that Matt Garza is who we (some of us) thought he was. When reality doesn’t match what the model says it should, you might just have a problem with your metric. I am so tired of people on this site claiming to be scientific, justifying their own magical thinking by claiming that those who simply acknowledge events that actually happened in reality are the weak-minded who are simply fooled and create mythologies.

    On a technical note, the big story here that nobody on FG has bothered to analyze is that “true” talent levels need to be regressed significantly and the factor for SO% needs to be adjusted based on a dramatic three year trend upward for starting pitchers. There’s also the small detail that a catcher might just have a lot to do with pitch location on two strike counts. I know, that’s heresy, but I’ll stick to my narrative – doesn’t make much sense to have bruised and bloody knuckles without it.

    Vote -1 Vote +1

    • Keystone Heavy says:

      U mad? If you are so tired of the writers, don’t read whats on the site!

      Vote -1 Vote +1

      • Paul says:

        Right. It’s better to just engage people who you agree with on every single thing. The Matt Garza is a bum narrative and counter discussion has been going on a for a while here. Nice try though.

        Vote -1 Vote +1

      • Dash says:

        No one said Garza is a bum. He had a good year, but people shouldn’t proclaim him as the next Cliff Lee without more data.

        Vote -1 Vote +1

      • t ball says:

        Paul, who is saying Garza is a bum? You’re exaggerating. The argument is more like he’s a stud – no, he’s almost but not quite a stud. That’s a bit different than saying he’s a bum.

        Vote -1 Vote +1

      • Paul says:

        t ball: Thanks for catching the exaggeration. You clearly didn’t read the entire article. This summation thing is pretty important. And it clearly continued the theme on this site of twisting themselves into knots, not to discredit Matt Garza, but to continue to prop up a set of supposedly predictive metrics that are fatally flawed. In other words, what mister_rob says below.

        Dash: That’s such a great point, because it’s really the heart of this ongoing argument. I didn’t see one person at any point in this entire discussion claim that Matt Garza is the next anything, much less Cliff Lee. We were merely saying that he is what he is, and that he’s much better than normalized metrics say, and that multiple writers on this site have used to claim that he’s actually more like a back end starter, or easily replaceable. And then the cheering in support of a trade that turned out to be a salary dump for a bunch of guys. The entire point all along has been that if you are willing to at least question the theory underlying FIP, Matt Garza is clearly not a replacement level starter.

        Vote -1 Vote +1

      • t ball says:

        No, Paul, I clearly read the entire article, but without the incredibly jaundiced eye you have. Man, what an axe you’re grinding. I prefer to take every article for what it actually says rather than fit it into my personal agenda.

        I see nothing in this article that looks like some sort of metric-propoganda. I do see in your post an angry rant against some of the metrics used but your rant is not supported by any evidence. The writer here references metrics supported by much research. It’s fine to dispute it, but I don’t have to just take your word for it.

        Vote -1 Vote +1

      • Paul says:

        tball: You’re right, I’m a member of Garza’s posse, so I have a personal axe to grind about this.

        That the author wound up in the right place is interesting vis-a-vis Garza. But I frankly couldn’t care less about him. FIP was used to dog him prior to the season, and the prior SO rate was dismissed as an outlier. Now that it fits the narrative, he suddenly has the ability to strike people out. It’s a cover for a metric that may be predictive over a large population, but it did a terrible job of informing us of Matt Garza’s true talent level. And worse for me is that a lot of people use these metrics to tell you that your eyes are fooling you (see, e.g., the entire no-hitter’s are flukes narrative).

        Vote -1 Vote +1

      • DavidCEisen says:

        “FIP was used to dog him prior to the season, and the prior SO rate was dismissed as an outlier.”

        His ERA has never been greater either. Most writers here would agree consistently out preforming your FIP raises the possibility something a pitcher is doing results in a lower ERA than FIP. However even assuming that to be the case, nothing Garza showed in the past made what he did in 2011 predictable. His career ERA is still 3.83.

        “It’s better to just engage people who you agree with on every single thing.”

        You may have a different definition of engagement, but generally belittling and insulting are not strategies for informed discourse.

        Vote -1 Vote +1

    • Angelsjunky says:

      Interesting points, Paul. I’m still wondering if anyone is ever going to come up with a metric that (somehow) combines the best of stathead metrics with What Actually Happened. I too am tired of advanced sabermetrics becoming so divorced from what happens on the field that they end up clashing with a more intuitive sense of player value; for instance, I have a hard time believing that Dan Haren was a better pitcher (6.4 WAR) to Jered Weaver (5.6 WAR) when the difference in their ERAs is rather significant (Haren 3.17, Weaver 2.41). I know, I know–FIP and BABIP and all that jazz…but what these sophisticated metrics don’t seem to account for is a pitcher’s ability to find ways to prevent runners from scoring that don’t show up in rate stats and percentages, but do show up in ERA (to some extent). In other words, ERA is directly related to results while while WAR tends to be a bit separate from that.

      To illustrate this point, look at two very different pitchers: Tom Glavine and Javier Vazquez. Sabermetricians want to give props to pitchers like Vazquez who put up gaudy rate stats, but seem to ignore the question as to why Vazquez’s ERAs have never come in line with his secondary stats (except for in a couple seasons). In other words, my point is that there is something that Vazquez does that keeps him from being a great pitcher and WAR shouldn’t try to bypass that. Tom Glavine didn’t have the sexy rate stats, and thus never had very high WARs (his highest was 5.7 and he was only above 5 twice) but he knew how to win ball games. That has to account for something and, in the end, made him a better pitcher than Javier Vazquez, even though Vazquez’s best three WAR seasons were higher than anything Glavine ever did.

      In a similar fashion, I find it difficult to swallow the degree to which Fld and BsR can impact WAR. I know that WAR is meant to be a catchall stat, but to be honest I don’t really care all that much about how bad of a baserunner Adrian Gonzalez is — his -8.2 BsR lowers his WAR by almost 1 to 6.6, which doesn’t accurately describe his overall value. A .338/.410/.548 hitter with very good defense is more than a 6.6 WAR player, and certainly shouldn’t be 1.1 WAR lower than Ian Kinsler’s .255/.355/.477 line, even though Kinsler’s defense was excellent and his baserunning very good.

      I could go on, but in short I think we have to remind ourselves that abstract metrics like WAR are just tools and shouldn’t be taken as absolute. Even Grandmaster Cameron has said as much. I would just take it a step further and posit that when we start abstracting too far away from What Actually Happened, I think we lose sight of actual value and start playing a more intellectual version of fantasy baseball. I am wondering if the sweet spot with statistics is going back to stats like Adjusted OPS and Adjusted ERA that actually refer to something actual. I would love to see someone take something like Total Average and contextualize that, like “Adjusted Total Average” and then replace OPS with that and add in seperate rates for defense and baserunning. In other words, make sure the advanced stats refer to actual performance, then contextualize that somewhat for era and park factors and call it a day (although maybe learning to optimize your park’s dimensions is a skill in and of itself?)

      Vote -1 Vote +1

      • kick me in the GO NATS says:

        Would you care if Adrian ran like Reyes? You sure would, so you should care that Adrian is an Anti-Reyes. RThe fact that adrian is one of the esieat guys to get into the front end of a DP matters a ton!

        Vote -1 Vote +1

  3. e says:

    Very nice article.

    Vote -1 Vote +1

  4. mister_rob says:

    The common opinion around here at the time of the trade was that the cubs greatly overpaid for a pitcher who owed much of his success to the defense behind him and the stadium around him

    common sense would tell you the difference between the ALE and the NLC would more than make up for those things.

    Maybe, just maybe the stats this site uses dont give the pitcher enough crefit for his own destiny, and rely way too much on normalizing and rely way too much on defensive metrics

    Vote -1 Vote +1

    • Joe says:

      Variability in stats, as well as confidence intervals, aren’t given nearly enough due on faux-pas statistical sites. But anyone who says, “maybe the stats this site uses dont give the pitcher enough crefit for his own destiny”…is clearly barking up the wrong tree.

      Vote -1 Vote +1

      • mister_rob says:

        barking up the wrong tree = looking at actual rates of Garza’s between 2010 and 2011, and coming to the conclusion that he was ONE AND A HALF runs better per 9 innings in 2011

        just way too much reliance on defense in those numbers

        Vote -1 Vote +1

  5. Dan Conley says:

    Good piece — we need a similar reality check story about Brandon McCarthy.

    Vote -1 Vote +1

  6. Daniel says:

    “Of course it’s possible that the league adjusts in 2012, but there is no reason to dismiss 2011 as a fluke.”

    This was a reality check? Did I misread or was the article not arguing that there are legitimate reasons to believe that Garza’s 2011 success is sustainable? Which people have been arguing here since like May. Sure, there’s a wait-and-see approach but that’s only sensible. How many people really rated Cliff Lee right after he won the Cy Young, in and out of the stat community?

    We don’t all have to agree, everyone. We don’t all have to agree.

    Vote -1 Vote +1

  7. Jack says:

    I called the crap out of his season…right down to the ERA @ or under 3.5.

    Vote -1 Vote +1

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>