Slider Effectiveness & Spin: Unexpected Results
I heard in passing from a credible source:
“The effectiveness of a pitcher’s slider relies on it having the same spin as his fastball.”
I figured it would be an easy test and could help to immediately identify top-rated sliders. After looking at the data every conceivable way and came up with the following conclusion: publicly available spin information has near ZERO correlation to a slider’s effectiveness. But while rooting around, I did find two factors which do matter, fastball velocity and the difference in slider and fastball velocity.
The theory behind the quote is that a hitter has a tougher time differentiating a fastball and slider if they are spinning at the same rate. So, the closer the difference, a higher chance for a swing-and-miss.
I compared 2018 pitchers with at least 200 sliders and 200 four-seamers thrown. Then, I compared just the difference, the absolute value of the difference, square of difference. Nothing tangible. Nothing matched.
It just might be something I’m missing with an input. For example, a factor I don’t have available is the spin axis, so I don’t know how much the spin is working vertically and horizontally. This information is not publicly available but could be a major input.
Then I began to add in other factors to see if it was a combination of slider spin and pitch velocity. Zip.
Finally, I added in fastball velocity. First, slider velocity and swinging-strike-rate (SwStr%) had an r-squared of .014 and that was the highest correlation I could find before I began examining fastball velocity. The r-squared between fastball velocity and the slider SwStr% was .20. The faster a pitcher throws, their slider is more effective. With so many other factors in play (e.g. deception, spin axis, perceived velocity), I’m surprised I was able to get that high of a correlation.
I started incorporating fastball velocity with the slider metric. I quickly found that the difference between fastball and slider velocity to also be a “major” factor. The change wasn’t linear but instead, it was a parabolic change above and below a 10-mph difference. Simply, throwing as fast of fastball as possible with a slider 10-mph slower is are the keys to the slider’s success.
To simplify the parabolic equation, I used the following equation:
Absolute Value (Fastball Velo – Slider Velo – 10) = ABS Difference
When combining the two factors, the r-squared be .25 and this equation:
Slider SwStr% = -.00643 * ABS Difference + .00918 * Fastball Velo – 0.6778
Going over the projected results, the ABS difference component has a standard deviation of 1% and the Fastball Velo has a 2% standard deviation.
Overall, I was a little disappointed that spin rated didn’t matter. On the other hand, I found a couple of factors to help determine the effectiveness of slider, especially its effectiveness in relation to the fastball. There is a possibility I’m missing something and please let me know in the comments.
Jeff, one of the authors of the fantasy baseball guide,The Process, writes for RotoGraphs, The Hardball Times, Rotowire, Baseball America, and BaseballHQ. He has been nominated for two SABR Analytics Research Award for Contemporary Analysis and won it in 2013 in tandem with Bill Petti. He has won four FSWA Awards including on for his Mining the News series. He's won Tout Wars three times, LABR twice, and got his first NFBC Main Event win in 2021. Follow him on Twitter @jeffwzimmerman.
I looked at spin rates on a bunch of different pitches and it seems it only really matters on fastballs, particularly those thrown up in the zone.
What’s the claim re: sliders? (Forget fastballs/sequencing for a moment, just focus on the slider). Is it A. Sliders don’t break in amounts proportional to spin rate? This is a physics question, no regression needed. (I’m not a scientist but I think it should be proportional.)
Or is it B. Sliders that break more are not harder to hit (in general) than ones that break less? This is a baseball analysis question, and one I would be very hesitant to agree with just based on some quick regression. There could be some meat on the bone here, looking at pitch location for example (I suspect a large break slider should either appear to start as a strike or end up in the zone to be effective).
Logically, I don’t see a 3rd option.
Remember when pitchers could only affect BB/K/HR because Voros’s data said so? People were saying there is no such thing as consistently inducing weak contact, which was soon proved untrue (Maddux always knew). Incredible claims require incredible proof!
Maybe the missing spin axis is a big deal? A slider has to have decent vertical break or it’s just traveling along the bat path for many hitters.
Anyway, interesting stuff. The biggest takeaway for me is about complementary pitches. And hilariously, since people are down on fastballs now, it sounds a little bit like grumpy old announcers going on about “establishing the fastball” to set up the other pitches.
Makes sense. One of the biggest problems with hitting a slider is trying to tell it from a non-breaking fastball in time; the faster the pitch is, the less reaction time is available to the batter to make that determination.
After pitch differentiation the batter must attempt a velocity determination for swing timing, and the greater difference in the velocity between a pitcher’s fastball and his slider the more difficult the latter will be to hit. So better sliders will have both higher velocities and a high velocity differential from the pitcher’s fastball. (Hello, Max Scherzer, with his 95 mph four-seam fastball and his 86 mph slider, both of which have the same trajectory until they’re 15 to 18 feet from the plate, and good luck to batters first trying to figure out whether or not a slider is coming and then having to adjust their swing timing sufficiently to hit that slider.)
All perfectly logical. Thanks for doing the math.
P.S. Sparky Lyle: “Ted Williams told me the slider was the one pitch he couldn’t hit.” If Teddy Ballgame had trouble with, it then you know it’s damned hard to hit.
You technically DO have the spin axis available, but it would take so long to go through brooks baseball and manually input all the data you find. But I think that would help. Just using the pitch f/x leaderboards, the difference in vertical movement (4 seam z minus SL z) between fastball and slider correlated with swinging strikes per swing on sliders to an R-squared of .1187. There was basically no correlation with horizontal movement difference (.0435 R-squared). As always with pitches and spin, useful spin has to matter, right. It’s why Verlander’s slider became more useful again despite its spin rate staying the same. It’s why Shoemaker’s slider lost it’s usefullness. https://www.beyondtheboxscore.com/2017/3/13/14886264/matt-shoemaker-los-angeles-angels-head-splitter-slider
So yeah, if you had an easier way to distribute spin axis amongst your pitchers (maybe you could limit the data to under 100 players (min 200 sliders, min 500 four seamers?) and then research each guys’ spin axis….it would still take a while though.
Sorry guys for being late to the discussion but there are some great ideas here and I look into them at some point in the future.