The thing that has always made pitching so fascinating to me is that while everyone is doing basically the same thing, there are so many different ways to do it. How a pitcher throws the ball and the types of pitches he’s able to throw have always seemed to be as much an expression of the pitcher’s personality as their talent or training.

Reading Patrick Jeter’s excellent mini-series on the statistic RE24, and the general concept of looking at the change in expected runs after each play to determine the play’s value, got me thinking about breaking it down even further to look at the modern pitcher’s arsenal of pitches. While a little complicated to explain, the idea is pretty simple at its core. It goes like this:

1. Calculate the value of every offensive play, on average, in terms of expected runs. (Here’s a helpful table for that. We are currently in an era where we’re seeing teams score 4 runs per game, and there is a column for that in the table. The number of runs per game changes the value of each play slightly because, for example, a single is worth more if there are more runners on base, on average.)

2. Calculate the value of every pitch that does not result in a play, on average, in terms of expected runs. These values are obviously a lot smaller than those of the actual plays, but there are a lot more of them and they add up. If a pitcher throws a strike and gets to an 0-1 count, that slightly reduces the average expected runs because he got one strike closer to a strikeout and got ahead in the count. If he throws a ball, and brings the count to 1-1, that slightly increases the expected runs, because he’s one ball closer to a walk and the count is even.

3. With today’s technology, look at every pitch thrown, identify the type of pitch it is, and calculate the change in expected runs after each. In some cases, this is done with computers only, and in others it’s done with a combination of computers and people. This could be a tiny change, say from a 0-0 count to a 1-0 count, which changes the expected runs by about .04, or it could be a big change, say from a HR, which has an average value of about 1.4 runs.

From there we can actually calculate the value of different type of pitch thrown by every pitcher in a way that simply wasn’t possible not all that long ago. Just to explore a little of what is out there, I looked at every starting pitcher from this year and last (combined) that has pitched at least 100 innings.

Pitching is Hard

It turns out that being a starting pitcher in the big leagues is a tough job. Relievers get to come in and throw as hard as they can for one inning, while starters have to save their strength for later innings and see the same professional baseball hitters three or more times in the same game. To keep those hitters off balance in later at-bats, starters have to have additional pitches, which probably aren’t as good as their best offerings. (The average number of pitches thrown by starters in the sample was 4.05.)

The table below shows the percent of starters using each type of pitch, and the average value of each type (per 100 pitches thrown). Value is given in runs above average, so positive is bad for a pitcher. (I didn’t include split finger fastballs or knuckleballs because they were thrown so infrequently that I didn’t trust the data.)

Pitch Vlues

Those numbers are all in comparison to the expected value of an “average pitch,” a sort of Platonic ideal of a pitch. We know that hitting is very hard in baseball, and this isn’t telling us that it’s easy. What it shows is that hitters really do have the advantage over the average starting pitcher, compared to relievers and top of the rotation types. The only pitch that starters throw that does better than average is the slider. And boy, if you’re a starter, throw a cutter at your own peril.

But Some Guys Are Really Good At It

Despite the challenges described above, most starting pitchers have at least one good pitch, and some have multiple great pitches. Only 9 percent of starters in the sample had no pitches that were above average, and man would I not want to face a big league lineup with their arsenals. These are the Alfredo Simon’s of the world. It would feel much better going to the mound knowing that I had more good pitches than bad, and 44 percent of starters had that warm and fuzzy feeling. The best in the biz have no below average pitches, and as of right now, that list consists of (in no particular order): Jaime Garcia, Jake Arrieta, Zack Greinke, Dallas Keuchel, Jacob deGrom, Adam Warren, Chris Sale, Matt Harvey, Madison Bumgarner, Andrew Heaney, and Jon Lester.

Here’s the cream of the crop for each pitch type:

•    Best Fastball: Jake Arrieta; -1.5 runs per 100 pitches
•    Best Change up: Danny Salazar; -3.2 runs per 100 pitches
•    Best Curveball: Carlos Carrasco; -3.7 runs per 100 pitches
•    Best Slider: Jorge de la Rosa; -4.4 runs per 100 pitches
•    Best Cutter: Jake Arrieta; -2.5 runs per 100 pitches

And The Reds…

Well the Reds aren’t very good at pitching right now. But hey, there are still some really nice things to watch for. For example, Raisel Iglesias’s slider (-1.8), Robert Stephenson’s curveball (-3.7) and Brandon Finnegan’s change up (-2.9) are all excellent offerings.

Can the Reds front office use any of this information? Maybe, I don’t really know. But I think it’s great that this type of information is available, so that we can put some real information behind all the great pitches and pitchers that we get to see on the mound.

3 Responses

  1. Jeremy Conley

    Last year Lamb’s curve and change were both slightly below average pitches, but with so few innings, I would feel find just calling them average. I did note last year that Lamb had the second largest difference between his fastball and change up in the majors, so that could be a pitch to look out for as well.

  2. IndyRedMan

    If you can calculate all that then how about sequencing and the end results there? For example…Cingrani to David Wright the other night….it was 0-2 on fastballs and maybe 1 foul before another fb out over the plate and there goes the game. Compare that to when Cingrani throws atleast 1/3 breaking balls just to show the hitter something offspeed before going back to your bread-n-butter. I think that’s why I got so excited about Lamb after I saw him even though his #s weren’t that good last year. That 68 mph curve or whatever it is would’ve made his fb seem like 98 if he would’ve thrown it more? Of course the Price and co. want him to throw it more but I wonder if its a injury concern? Sean Marshall curveballed himself right out of the league!

    • Jeremy Conley

      You’re definitely tapping into some of the interesting issues in this area Indyredman. With the data available to the public through fangraphs and a few other sites, you can only see the values of individual pitches.

      So say a pitcher gets ahead with 0-2, then throws a letter-high fastball to change the batters eye level (which is called a ball), and then gets a strikeout with a down and away slider. With the data we have, the fastball would get downgraded slightly, because it was a ball, and the slider would get upgraded significantly because it resulted in the strikeout. In reality, we know that the fastball was used to set up the slider, and the slider may not have worked without it.

      What a team can do is pay a company called Baseball Info Solutions for more detailed data, or they can set up an analytics department to gather the data themselves. Then you could do an analysis of a pitcher’s fastball value when it is thrown in an at-bat with a different pitch, compared to when it is thrown as the only pitch in the at-bat, to get to your Cingrani example.

      When people talk about teams “relying more on modern analytics,” or setting up an “analytics department” doing what I just described might be one of the many purposes. It could give you and edge in how you coach your pitchers. Or, conversely, all of these measures can be used (in reverse) for hitters, so you could see which hitters struggle more with different types of pitches, in different sequences, etc.