MIA-PHI Match-Up: Pitch Type Linear Weights

I have been toying around with an idea for pitcher-hitter match-ups based not on prior head-to-head performance or platoon splits, but rather pitch type linear weights.

For those that are unfamiliar, pitch type linear weights basically takes a batter or pitcher’s performance on each type of pitch they throw or face during the year (e.g. four-seam fastball, slider, etc.) and converts that performance into runs created or runs saved relative to average. At FanGraphs, we show both the total runs created or saved for each pitch (e.g. wFB) and a normalized version for the value per 100 pitches thrown (e.g. wFB/C).

I thought it would be interesting to compare the starting pitcher’s pitch type linear weight performance against the lineup he is facing. To do this, I calculated the difference in run value between each pitch type for each starting pitcher and the hitters they might face. The difference is shown in the tables below. Green coding denotes an advantage to the pitcher, while red indicates an advantage for the hitter. I used the normalized version of each pitch type (i.e. run value per 100 pitches thrown/faced) to control for playing time, pitches seen, etc.

The tables below show the match-ups for tonight’s game between the Marlins and Phillies (7:05pm EST) for both Josh Johnson and Roy Halladay:

A few caveats. First, I only show data for hitters that accumulated >= 300 plate appearances over the past three years and currently appear on our active roster leader boards. Second, the pitch data is from BIS, so the classifications will vary. Often you will see a slider classified as a cutter, and vice versa. Ideally, I wanted to use our Pitch FX data here, but we do not list out the run values by pitch type for hitters.

The top of each table lists how often the pitcher has thrown each pitch type over the past three years. In the bottom table I will try to list where that hitter is slotted in the batting order, if I get my hands on the line up in time. For this post, I did not have a chance to incorporate batting order position, so I’ve just left it blank.

Looking at the match ups, Josh Johnson has a pretty significant advantage over the Phillies’ hitters. Jim Thome and Shane Victorino are pretty good fastball hitters, but outside of those two Johnson shows a clear advantage in terms of his two predominant pitches (fastball and slider).

Roy Halladay, on the other hand, is facing a more potent line up. Halladay’s cutter is his more dominant type of fastball, and you can see how much of a run expectancy advantage he has over most Marlins hitters; Jose Reyes, Hanley Ramirez, and Mike Stanton are the exceptions in this case. Halladay should generally dominate this lineup with his curveball. Halladay sports one of the best curves in the league, and outside of Chris Coghlan, no Marlins hitters has an advantage against the curveball when it comes to run values.

One big caution about these numbers; they are averages. More specifically, the values for each pitch type are stripped of just about all context. So when we see that some Marlins hitters have higher run values against the cutter than Roy Halladay, we have to remember that Halladay has one of the most dominant cutters in all of baseball. Hitters like Reyes and Stanton accumulated those averages against all sorts of different pitchers, not all elite hurlers like Halladay. It also doesn’t account for the unique combinations that different pitchers utilize in a game. So while a hitter might appear to have an advantage over a pitcher when it comes to curveballs, one can imagine that things may be different when facing someone like Justin Verlander, whose fastball-curveball combination makes both pitches more difficult to handle.

Another caution is that, generally speaking, pitch type linear weights are more retrospective than prospective. More simply, they tell you what has happened, not what is going to happen in the future. The weights are not adjusted for park or defense, so there are many factors that could affect how hitters and pitchers will fare in the future that the previous values do not accurately represent. Did the hitter or pitcher switch parks? Has the defense behind a pitcher changed? What catcher is the pitcher throwing to? These are things we must think about when interpreting the values.

Even if the predictive value is less than with other stats, I think the picture it paints is interesting and adds a bit more information to consider when analyzing a match up. As I mentioned at the outset, this is brand new and definitely in “beta” form, so feedback is welcome.




Print This Post



Bill works as a consultant by day. In his free time, he writes for The Hardball Times, speaks about baseball research and analytics, consults for a Major League Baseball team and appears on MLB Network's Clubhouse Confidential. Along with Jeff Zimmerman, he won the 2013 SABR Analytics Research Award for Contemporary Analysis. Follow him on Tumblr or Twitter @BillPetti.


20 Responses to “MIA-PHI Match-Up: Pitch Type Linear Weights”

You can follow any responses to this entry through the RSS 2.0 feed.
  1. I like it as an extra thing to look at when evaluating matchups but, like you said, you can’t take the hitter’s advantage against a cutter and expect it them to fair well against Halladay’s cutter. Also, each pitch value is based on 100 pitches and no hitter will ever see any pitch that many times in a game. It’s a fun and interesting to look at, to see which types of pitches hitters have faired well against, though.

    Vote -1 Vote +1

  2. pbjsandwich says:

    very interested to see how this turns out. Not expecting something revolutionary but I’m a big fan of pitch values and love this type of pregame analysis

    Vote -1 Vote +1

  3. Dan in Philly says:

    The interesting thing is how such data (more available to MLB than us) is affecting offense overall. Ever since Pitch FX and cheap technology, every pitching staff have been able to maximize their own pitchers’ abilities against the hitters they face. I’m quite sure they not only know this, but also how location, pitch speed, etc factors into the hitters’ ability to make good contact.

    In this way pitchers have more of an advantage over hitters than they ever have before. Hitters cannot utilize technology in a similar way, all the control of what kind of pitch is to be thrown is in the hands of the pitcher – the hitter can only try to hit what’s thrown. If he can’t hit fastballs up and in, pitchers know this to a percision and detail unthinkable just 10 years ago.

    Vote -1 Vote +1

  4. Whelk says:

    I really like the idea, but I’d much prefer it with linear weights based off batted ball type than off result. They’re real easy to implement, and I think they’d make the Fangraphs version more credible than they are now.

    Do you have the sway to change that yet? Soon?

    Also, I’d like to know more about matchups first. For instance, Pomeroy claims that in college basketball, the offense has way more effect on turnover rate than the defense does. I’d like to know more about whether the pitchers linear weight out the batters seems to Mayer more and by hire much.

    Vote -1 Vote +1

    • Whelk says:

      Thought about this a little bit more. First off, I agree with Slartibartfast that breaking it out by handedness would be nice.

      Secondly, thinking more on the matchups part, If you show the pitcher’s linear weights in addition to percentage used, you can break it into four quadrents (pitcher above average/hitter above average, pitcher above average/hitter below average, etc.)

      Vote -1 Vote +1

    • monkey business says:

      Handedness is a must. Obviously, a slider from a lefty is a completely different beast for a hitter to handle than a slider from a righty.

      Vote -1 Vote +1

  5. Frag says:

    I actually used something like this in ‘predicting’ the results of the postseason last year. I got the NLCS right, but missed badly on the ALCS and World Series (though, I was a strike away from getting it right in the WS haha).

    Vote -1 Vote +1

  6. But says:

    Nice work.

    Don’t worry about getting the batting order so much. The information could be done for each pitcher vs each team without scrambling to get it finished. Would delete it from the picture to avoid confusion. Thought Jim Thome was leading off for a second.

    Halladay vs. Marlins’ hitters, could also be confusing, since it’s Halladay vs. the league, and the Marlins’ hitters against the league. Would remove that as well.

    You did clarify after the picture, but said you wanted feedback.

    Be interesting to see if Halladay throws more than 19% curveballs, or if his pitches are so good, he doesn’t need to.

    Vote -1 Vote +1

  7. But says:

    Putting the yearly pitch values in(maybe career too, to account for aberrations), such as Halladay’s FB -0.8, CT 19.5, CB 14.1 would be very beneficial.

    Vote -1 Vote +1

  8. But says:

    “The top of each table lists how often the pitcher has thrown each pitch type over the past three years.”

    If it’s just % thrown by pitcher shouldn’t it be…

    Pitch Type FB, SL, etc.
    %Thrown last 3 years

    Bringing the /c is runs saved per 100 pitches if I read that correctly.

    Vote -1 Vote +1

  9. Slartibartfast says:

    As is, I’m not sure this is terribly useful. But if you split it up L/R I’d be more inclined to glean something from it. Not sure if the pitch lwts can be split.

    Vote -1 Vote +1

  10. jfree says:

    Well looks like Johnson got completely shellacked and Halladay has had a pretty good game so far. Isn’t it annoying when anecdotes refute a well-thought out idea?

    Vote -1 Vote +1

    • But says:

      “Halladay should generally dominate this lineup with his curveball. Halladay sports one of the best curves in the league, and outside of Chris Coghlan, no Marlins hitters has an advantage against the curveball when it comes to run values.”

      What is refuted?

      Vote -1 Vote +1

      • monkey business says:

        The idea of this post was to use a new tool to look at batter vs pitcher match ups. The fact that the tool apparently made a prediction that didn’t turn out to be anywhere near the truth is problematic when arguing for the usefulness of the tool.

        Yes, he does say that it is not a great example. This raises the question, why is it the example?

        Vote -1 Vote +1

      • jfree says:

        What is refuted?

        Looking at the match ups, Josh Johnson has a pretty significant advantage over the Phillies’ hitters. Jim Thome and Shane Victorino are pretty good fastball hitters, but outside of those two Johnson shows a clear advantage in terms of his two predominant pitches (fastball and slider).

        Box Score for Josh Johnson — 3.2IP, 11 H, 6R, 6ER, 1K, 1BB, 16.875 ERA, 3.27 WHIP

        Now I suppose you’re going to prattle on about what the word refuted really means — or perhaps we can argue about what the author really meant by “clear advantage”. Or perhaps we can argue that those box score numbers are all relatively good in some way.

        Vote -1 Vote +1

      • ccoop says:

        obvious troll is obvious

        Vote -1 Vote +1

      • But says:

        The prattling begins. I know well what refuted means, and don’t wish to argue.

        “Isn’t it annoying when anecdotes refute a well-thought out idea?”

        Conclusion of the author may have been refuted, no offense, but not the idea itself.

        Vote -1 Vote +1

  11. Ben Hall says:

    I love this idea, Bill. However you fine-tune it, I’ll look forward to them in the future.

    Vote -1 Vote +1

  12. jcxy says:

    forgetting the outcome of tonight’s game, i’m very intrigued by this as a way at looking at pitcher vs team matchups. however, without R/L data, i fear it’s pretty useless at this point.

    second, whoa! that marlins OF is quite an adventure defensively.

    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>