Stephen Strasburg and Early Season Velocities

With the real Opening Day behind us — sorry Astros and Yankees, you’re just late — we now have real 2014 data from almost every team on the leaderboards. Of course, besides answering trivia questions, we all know that there’s really nothing insightful to be learned from one day’s performance, and we’re not going to find useful information to be analyzed there until the samples get a lot bigger.

But there are some numbers that became useful in very short order. Strikeout rate, for instance, only has to be regressed 50% to the mean at a much lower number of batters faced than most other pitching metrics, and big changes in K% over even a few starts can prove somewhat meaningful. It’s hard to fluke your way into getting a bunch of Major League hitters to swing through your pitches, and if you’re consistently throwing pitches by people, it’s a pretty good sign for the future.

That’s still a results-based metric, though, and by definition, those have to include significantly more variables than things that do not require a response from either the opposition or the teammate. The things that we can measure the quickest are the things that are affected by as few players as possible, and ideally, only one player. For instance, a catcher’s caught stealing rate will take more time to tell us about his arm strength than simply measuring his pop time — baseball lingo for the length of time it takes him to throw down to a base on a steal attempt — because CS% also includes the pitcher and the runner as variables. With enough CS attempts, we can infer things about a catcher’s arm strength from the results, but if we have pop time, we can measure that arm strength directly, and do so much faster.

That’s one of the reasons why we love PITCHF/x data, and specifically, the velocity readings they give us. We can infer quality of stuff from strikeout rate — guys who throw hard get more whiffs, generally — or swinging strike rate, but we don’t have to anymore, since there’s a direct measurement that takes place for every pitch of every Major League game. Velocity readings don’t care about quality of opposition or the umpire’s strike zone that day, with the pitcher himself being responsible for almost 100% of the calculation. Ballpark effects — the system is not calibrated exactly the same in every single city, and there are parks that run a little “hot” or “cold” — and weather have some impact, but relative to other metrics, the outside factors have very little impact on a pitcher’s velocity readings.

How quickly do fastball rates mean something? Jeff Zimmerman noted that a starting pitcher’s velocity over three starts will usually be within 1 mph of his seasonal average velocity. Not surprisingly, velocity can be meaningful very quickly. If any stat from the first day of the season means anything, it’s probably fastball velocity. But, even knowing that, we still have to use caution when trying to draw any kind of strong conclusion from limited samples of data.

For instance, you’ve probably read today about Stephen Strasburg’s lower Opening Day velocity; his pitches that were classified as four seam fastballs averaged just 92.7 mph, down from 95.2 mph a year ago. When asked about it after the game, Strasburg said this.

“It felt pretty good. I guess radar guns have offseasons, too. I don’t know,”

The nice thing about PITCHF/x is that the cameras don’t have to get into game shape, and there usually aren’t gross measurements errors. But that doesn’t mean the system is perfect either, so when you see a pitcher post a significantly different velocity than you’re used to, the first thing I’d suggest doing is checking the velocities of every other pitcher in the game too. If the system is producing a systematic bias, it will show up in the other pitcher’s numbers, and should be easy to spot.

Yesterday, Strasburg was opposed by Dillon Gee of the Mets, and the same system that clocked Strasburg’s fastball down a few ticks had Gee sitting at 88.3 mph with his four seam fastball, a 1 mph decline from his 2013 average. So, immediately, we know we should keep investigating, because the first flag for systematic PITCHF/x calibration error has been raised. So, let’s go to the relievers who recorded at least three outs (and pitched significant innings in the majors last year), although with the qualification that now we’re taking small sample data even smaller, as we’re looking at 10-30 pitches in most cases.

Bobby Parnell: 92.8 mph yesterday, 95.1 mph last year
Jose Valverde: 91.9 mph yesterday, 92.8 mph last year
Tyler Clippard: 92.4 mph yesterday, 92.0 mph last year
Drew Storen: 92.0 mph yeserday, 93.7 mph last year
Jerry Blevins: 88.7 mph yesterday, 89.8 mph last year

Outside of Tyler Clippard, everyone was down yesterday, and Bobby Parnell had as big a drop as Strasburg. That would normally be encouraging, as it would be a decent indicator of the system in New York just running a bit slow, except as I write this, the Mets just announced that Parnell was pitching with a partial tear in his MCL, and he’s headed to the disabled list. So, yeah, that explains Parnell’s velocity drop.

But, keep in mind, Parnell isn’t the only guy who threw yesterday and had lower than last year’s average readings, and they aren’t all headed for surgery. And that’s because it’s April, and velocity is at its lowest point in the first month of the season. Last year, by month, average four seam velocity from PITCHF/x:

April: 91.5
May: 91.6
June: 92.0
July: 92.2
August: 92.2
September: 92.3

Comparing April velocity to seasonal average velocity will often make it look like a pitcher’s not throwing as hard, because pitchers just don’t throw as hard at the beginning of the year as they do at the end of the year. Instead, what you want to do is compare a pitcher’s velocity from this year to the same time period last year. And, while we haven’t advertised this feature enough, you can actually do that right from our game logs.

For instance, here is Stephen Strasburg’s game log for the last 365 days, in chronological order. You can see the huge cliff between his final start velocity of 2013 and his first start velocity of 2014, but you can also go back and see what his numbers were on Opening Day last year. In this case, they’re not particularly encouraging, because Strasburg’s four seam fastball sat at 96.0 last year, the highest average he’d post all year. But, it was still worth checking, and often, you will see that the difference in a pitcher’s velocity from the same time period last year will be less than the difference between his early starts and his seasonal average from the year before.

Also, it’s worth keeping in mind that nearly every pitcher loses velocity as they age. Strasburg throwing slower shouldn’t actually be a shock; it would be a surprise if he wasn’t ticking downwards. This is just how pitcher’s age. Throwing hard is a young man’s game, which is why a 21-year-old was the hardest throwing starter in MLB yesterday. Velocity peaks incredibly early, and it is unlikely that Strasburg will ever throw as hard as he used to. That’s not necessarily a sign of injury; it’s a sign of Strasburg being older.

Now, it can be a sign of injury, as Parnell’s MCL tear shows. Lowered velocity does have a correlation with higher injury rates, and it is an early warning sign that something might be wrong. But before you draw too many conclusions from early season velocity readings, remember that park effects do exist on these readings, and that April velocities are often lower than seasonal averages. Use the tools available to check and see if a park was running hot or cold on a given day. Use the game logs to see how a pitcher’s velocity changed throughout prior seasons. And don’t overreact to one start.

Stephen Strasburg’s missing velocity might be something to worry about. Or it might be nothing. As we know with every other metric, one day’s numbers mean basically nothing. That’s less true for fastball velocity, but it’s still mostly true. Let’s give it a few more starts, and if Strasburg is still sitting at 93, then we can start to wonder what that kind of change might mean for the future.