Several years ago I had a brief debate with a friend about what kind of pitch profile leads to the most success for a major league starting pitcher. I think my friend suggested something about fastball-sinker-slider pitchers and I believe my position had something to do with fastball-breaking ball-changeup pitchers. But we were both blissfully unencumbered by facts or stats. So now I’m going to delve into the subject in a little more depth.
There are a number of different factors that lead to or success as a starting pitcher. My goal here is to dig into the stats to see what differences I can find between good starting pitchers and other pitchers in the pitches they throw, their velocity and their statistical profiles. After crunching the numbers, I found the differences between good starting pitchers, other starting pitchers and pitchers in general to be interesting. I also found the similarities between the groups in some areas equally intriguing.
Methodology
I used the cumulative pitching stats from all pitchers from 2003-2010 from Fangraphs.com. This included the pitch type data from Baseball Info Solutions (BIS), not Pitch f/x. I put the pitchers included in this data into three non-unique groups:
- All pitchers (N=1538)
- Starting pitchers – I wanted to include the entire population of starting pitchers, including 5th, 6th, etc. starters who don’t get many starts. But I chose to exclude the most fringy starters, pitchers who only got the occasional spot start. So I limited this group to pitchers who had at least 10 starts over this time period. (N=504)
- Good Starting Pitchers – I first had to pick a statistic (or group of statistics) which separated the good starting pitchers from the not so good. I couldn’t use any counting stat, including WAR because it would exclude good starting pitchers who had only been in the majors for a short period of time and would include some poor pitchers who had been around to rack up a bunch of WAR over several years. So I focused on rate stats. I didn’t want to just pick one of FIP, xFIP or tERA. Without going into a longer discussion of the relative merits of each stat, let's just say that they are all good stats which tell you something important, but they all leave something out as well. So I chose a somewhat inelegant solution. I created a weighted aggregate average of ERA, FIP, xFIP and tERA (weighted 1, 2, 2, 2) which I call "agERA". I chose the top 20% of the starting pitchers group above. But for this group I wanted to limit it to pitchers who were primarily starters. So I excluded pitchers in this group who started fewer than 50% of their games over the date range. (N=75)
Pitch Type Usage
The graph above shows how frequently the different pitcher groups use various pitches. The full data for all BIS pitch types is below.
The variation for each pitch from group to group are not huge. But there are some interesting differences. Starting pitchers throw fewer fastballs than other pitchers. This isn’t surprising, as one would expect starters to have more pitches in their arsenal and thus rely less on their fastball. There is a significant difference in the use of sliders. Starters use them much less than pitchers in general, and top starters even less than starters overall. Curveballs are thrown considerably more often by starting pitchers than other pitchers, although each group throws sliders more often than curveballs. Starters also throw more changeups than other pitchers.
This data doesn’t show a clear pattern in pitch arsenals among the three groups, which is unsurprising given the fairly low level of granularity of this data as I have presented it. There are some significant differences in fastball, breaking ball and offspeed pitch use, but I was surprised by the relatively modest degree of differentiation.
A cursory look at the number of pitches in a pitcher’s arsenal also shows a surprising lack of differentiation between good starting pitchers and all starting pitchers. Without meaningful data to show how many "plus pitches" various pitchers throw, I was left to look at how many pitches each pitcher throws at least somewhat frequently. For all starting pitchers, 69.4% have three or more pitches that they use 10% or more of the time. For the good starting pitcher group, it is 72.0%. While it makes sense that better starting pitchers would have a more diversified pitch selection, I was surprised that the difference is so small.
Pitch Velocity
Interestingly, all pitchers and starting pitchers have essentially the same fastball velocity. But the group of better starting pitchers throws fastballs which are almost 2 mph faster. For the most part, the rest of the pitch velocities follow a similar pattern with top starting pitchers throwing harder on average than the other two groups. I also calculated the difference between average fastball velocity and average changeup velocity. Top starting pitchers have a better fastball-changeup differential by about 0.5 mph.
Statistical Profile
The following table shows the general statistical profile of the three groups of pitchers.
It is no surprise that good starting pitchers have a better statistical profile. This is of course true given the very methodology I used to define the group. There is a lot here which is pretty much expected. Starting pitchers strike out fewer batters than relievers, but also give up fewer walks. Starting pitchers (and especially good starting pitchers) induce more groundballs and fewer flyballs than pitchers in general.
I find it particularly intriguing that two statistics which are supposed to be largely out of the pitcher’s control (either due to luck or some intervening cause like defense) show a distinct difference among the three groups. The HR/FB rate for all pitchers was 11.6%, but only 10.8% for starting pitchers and even lower at 9.2% for top starting pitchers. And BABIP was .303 for all pitchers, but a little lower for starting pitchers at .298 and even lower for top SP’s at .291. Is this difference significant? Perhaps these are in part just products of the smaller sample size for the top starting pitcher group. The lower BABIP is in part due to lower line drive percentages for the starting pitcher groups, and perhaps top starting pitchers tend to be on better teams with better defenses. It’s hard to say if these factors account for the entirety of the BABIP differences among the three groups of pitchers.
It is harder to explain the differences in home run rate. Do better starting pitchers disproportionately play in parks unfriendly to home runs? In part, the answer is yes in that very pitcher friendly parks will help starting pitchers with the various elements which give them good agERA and thus help them make it into this group at all. But I doubt this comes close to explaining the significant differences in home run rate. I think there’s something going on here which requires further study.
Conclusion
This little study of course merely scratches the surface on how good starting pitchers differ from other pitchers. There is clearly a great deal of work that can be done (and is being done) with Pitch f/x and other statistics to identify indices of starting pitcher success. Useful and reliable metrics on the effectiveness of each pitch thrown would go a long way towards building a comprehensive model of the successful starting pitcher.
Connect with Beyond the Box Score