September 24, 2020

Statistically Speaking: Is Stephen Strasburg’s Slider Hurting His Curveball?

As 2014 begins to slowly loom larger in the rear view mirror as the remaining games trickle in, fans of the Washington Nationals are yet again at a place where the promise of Stephen Strasburg lies somewhat orthogonal to his results. Despite some encouraging stats—in particular, a 3.00 fielding independent pitching and a 3.0 wins above replacement, all in the top-10 for National League starting pitchers—there is also plenty to point to a discouraging season for the righthander.

Hopes were high earlier in the season, as 2014 looked to be a year where Strasburg was finally completely healthy; add to it an addition to his already devastating four pitch repertoire in the form of a slider, and it appeared that the NL Cy Young Award and the NL East were all but wrapped up, courtesy of D.C.’s once and future ace.

Much was made of the inclusion (then the scrapping, only to be brought back again) of the slider, with most feeling this was a subtraction by addition. Recently, a great article by the venerable Pat Jordan added some body and tangibility to the underlying discouragement of the new pitch by Strasburg, with a particular passage being at the heart of the issue:

There’s a longtime axiom in baseball that a slider ruins a pitcher’s curveball. They are diametrically opposed pitches, a stiff-wrist slider and a loose-wrist curveball. When a curveball pitcher adds a slider to his repertoire, pretty soon he won’t have either. He’ll have a slurve. A slurve is a big, fat, right-to-left breaking pitch that loses the best qualities of both pitches. It’s slower than a slider and begins to reveal itself too soon, and it has a less definable break than a curveball. The greatest breaking pitches are the ones that break late, sharp, down and a lot.

In essence, the slider appeared to be a waste of time; Strasburg’s otherworldly fastballs, curve, and changeup were plenty to not only get hitters out, but dominate them. At best, it would be an infrequently used pitch, occasionally flipped up there to set up another pitch. At worst, it would make his once-in-a-lifetime curveball less effective—a slurve—and something that was in between the curve and slider, possibly losing the bite and movement in the process, as alluded to in Jordan’s piece on overhand curves. Adding insult to injury, it could make his other pitches also a little less effective in the process, thereby making Strasburg’s stuff a little more human.

Could the trials and tribulations of 2014 be caused by Strasburg’s flirtation with the slider, making his repertoire a little less effective and more hittable?

First, let’s figure out how often he has thrown the slider; for this and any following stats, this season will be compared to 2013, the most recent (and most healthiest) season where he threw his usual four pitch mix: a four-seam fastball (FA), two-seam fastball (FT), curve (CU), and changeup (CH). For the table below and in the sake of data robustness, the number on the left is taken from FanGraphs PITCHf/x data and the number on the right is from Brooks Baseball, which has an additional manual classification correction implemented.

Season FA% FT% SL% CU% CH%
2013 48.7/57.7 12.2/3.5  — 22.8/22.8 16.2/16.0
2014 40.2/57.5 18.3/1.3 0.5/1.9 17.9/17.6 22.8/21.3

For Strasburg, the slider is a rare pitch, being used at most two percent of the time. This being said, has its incorporation affected the success of his other pitches? To answer that, here is another table, including 2013-14 data on each pitch type, including their respective walk (BB%) and strikeout (K%) percentages as well as weighted on-base average against (wOBA) and pitch type linear weights per 100 pitches thrown (LW/100), with LW/100 in particular especially useful telling us how effective a particular pitch was, in terms of minimizing run expectancy.

Season Pitch BB% K% wOBA LW/100
2013 FA 8.70% 12.60% 0.325 0.62
2013 FT 10.10% 10.10% 0.351 -0.27
2013 CH 8.30% 43.80% 0.172 1.86
2013 CU 2.40% 50.60% 0.150 2.16
* * * * * *
2014 FA 5.50% 18.60% 0.353 -0.18
2014 FT 6.00% 13.50% 0.406 -1.72
2014 CH 5.50% 43.20% 0.210 1.65
2014 CU 1.60% 47.20% 0.252 0.30
2014 SL 0.00% 14.30% 0.310 -1.55

Between last season and current, all of Strasburg’s main four pitches have dropped in effectiveness, per LW/100, with the slider overall not being a terribly effective pitch, given its greatly negative linear weights value. wOBA shows the same trends, with all of the ‘big four’ getting hit around a little more this season. Strasburg’s strikeout percentages appear to be as strong as ever this season, however, with each enjoying a moderate hike in rate compared to last season.

These numbers are valuable, but don’t provide the entire story. Variables such as release points, pitch movements, and spin rates, to name a few, all play a role in a pitch’s effectiveness and aren’t necessarily reflected in the above numbers; in short, a pitch and its ultimate effectiveness is comprised of a number of dimensions.

Let’s look at those dimensions and see if we can visualize any changes from last year to today that might help answer whether the slider is hurting Strasburg’s other pitches. One way to do this is with the help of a statistical method that is used to reduce the dimensionality of a dataset: principal component analysis (PCA).

An approach that is commonly seen in psychology and sociology and used in things such as intelligence and personality tests, PCA is used to identify and analyze underlying linear structures and previously unsuspected relationships in order to reduce large datasets into more palatable results, using their variance to help collapse and parse out themes that help in explaining the underlying structure of the data.

For the curious, a great reference for PCA can be found here. For the purposes of this article, it isn’t so much reducing factors, but visualizing how much the factors ‘cluster’ for each pitch in 2013 and seeing if that clustering pattern is the same (or different) this year, so the math and nuts and bolts of the PCA will be glossed over, simply because we are just interested in a snapshot of the process versus the process itself.

The variables in question are primarily PITCHf/x based and include:

spin direction
break angle
pitch horizontal movement
pitch vertical movement
break length
spin rate
horizontal release point
vertical release point
vertical height of the pitch where it crosses hoke plate
horizontal height of the pitch where it crosses home plate

After applying a PCA to Strasburg’s 2013 and 2014 data, the following loading plots of the respective principal components were created (PC1 and PC2, which explain 99.9% of the variance of the factors discussed), which reveal the relationships between variables; here, it is with respect to pitch type.

First 2013:

Strasburg_13_PCAEssentially, Strasburg’s fastballs and changeup ‘run together’, with his curveball and the respective factors at play behind it being a separate and disparate entity. Thus, we have our template for what the various factors and dimensions of each pitch for comparison to this year, with the new pitch.

2014 looks like this:

Strasburg_14_PCAWith the new pitch, we do see some less-clustered aspects of the curveball, but more drastically, with the changeup; there is perhaps a smidgen of truth to the idea that the slider, with all other factors remaining the same, could perhaps lend itself to hurting Strasburg’s other pitches, though not necessarily the curveball in isolation. However, it isn’t such a cut-and-dried ‘his curve looks like his slider’ explanation, which was discussed in the Jordan article, or at least, not yet.

While the few specks of the slider are tough to discern, we do see them drifting between the two trend lines, with some bleed of the curve and change into an in-between area. However, we don’t have a grasp of what underlying factors are at play with this visualization. Interpreting some of the factor loadings (not presented), the biggest difference between this season and last is the exchange of importance between the horizontal release point (x0) and pitch height as it crosses the plate (pz), with x0 having a slightly larger role as a factor in the loading of the variables used this season.

remain stable across the two seasons of interest. Other factors could be at play with respect to Strasburg’s somewhat shaky 2014; injury and even simply being a year older and the ramifications that has on a pitcher’s ultimate success could be variables that are responsible for the hiccups that the PCA does not capture.

Overall, through the interpretation and visualization of many of the factors and their respective dimensions at play with respect to Strasburg’s pitching, we do find that while his curve still might ‘look’ like his curve after the incorporation of a slider, there is some potential for some trending towards a slurvier offering if this small amount of data are to be trusted. With that said, Strasburg might do well to scrap the slider, as not only is the curve at risk of looking less like a curve, but his changeup as well is threatened by the unfavorable, neither here nor there characteristics of his slider.


Data courtesy of Baseball Savant, Brooks Baseball, and FanGraphs.

Stuart Wallace is a Contributor to District Sports Page. A neuroscientist by day, the Nevada native also moonlights as an Associate Managing Editor for Beyond the Box Score and a contributor at Camden Depot and Gammons Daily. A former pitcher, his brief career is sadly highlighted by giving up a lot of home runs to former National Johnny Estrada. You can follow him on Twitter @TClippardsSpecs.

About Stuart Wallace

Stuart Wallace is a Contributor to District Sports Page. A neuroscientist by day, the Nevada native also moonlights as an Associate Managing Editor for Beyond the Box Score, stats intern at Baseball Prospectus, and a contributor at Camden Depot. A former pitcher, his brief career is sadly highlighted by giving up a lot of home runs to former National Johnny Estrada. You can follow Stu on Twitter @TClippardsSpecs.

%d bloggers like this: