The following is an introduction to a new Bayesian projection system, which can be found here.
This project was partly inspired by a recent episode of the Driveline Baseball podcast which aired this offseason in which Kyle Boddy, founder of Driveline, and Mike Rathwell, CEO of Driveline, had a conversation about the topic that overwhelmed baseball for many months – whether teams should sign Manny Machado or Bryce Harper. Their initial reaction was to express disappointment that the debate had started in the first place. Machado and Harper, they argued, had almost nothing in common besides the fact they belonged to the same free agent class. They play different positions (implying different replacement levels and, thus, entirely different markets), and more importantly, have entirely different amounts of uncertainty associated with their projections. Machado has been as reliable as they come, playing premium positions (SS and 3B) and improving his defensive abilities every year. Harper, on the other hand, had just come off of what some have called one of the worst defensive seasons by an outfielder in recent history. However, he also had a 10-WAR season in 2015, a ceiling which Machado hasn’t touched. This led into a broader discussion about how to compare contracts from players with different levels of risk. Specifically, they explain how the sabermetric community’s approach to answering the question of valuing risk in Major League Baseball contracts has fallen short in three areas.
First, while many writers make note of the riskiness of certain assets, they fail to define that risk in precise terms. Most public projection systems output point estimates and public researchers suggest that the output is at the upper limit of predictive accuracy, and hence should be treated as a near certainty. It is worth noting that baseball is not the only field which has grown uncomfortable with uncertainty. Whether it is a decision to buy a certain stock, hire a particular company to ship your goods across the country, or decide who will be our next president, many analysts make the mistake of assuming a binary, discrete outcome is the result of a binary, discrete process. Instead, we posit that once we start to see the world as the outcome of several continuous, probabilistic processes, we can manipulate those processes in ways that give us an extreme competitive advantage (in baseball at the very least). Boddy explains:
“While a tweeter very fairly pointed out that sites like Baseball Prospectus and FanGraphs do mention upside and downside, rarely is it quantitatively actually approached in these articles. Rarely if ever, I should say. It’s very frustrating because just making a note that Tesla’s stock is more volatile than Microsoft’s is not enough. That wouldn’t be enough for a financial planner to be like ‘Oh, ok that’s a very deep analysis.’ It’s also not all downside, which is how a lot of these tweets [go].”
In other words, before describing the optimal mix of risky and safe players on a major league roster, there need to be accurate and reliable methods by which to describe that risk. There are very significant drawbacks to assuming too much downside, so carefully tracking exactly how uncertain you are of your team’s future performance as a whole is imperative. Also, as is the case with all science, precisely measuring levels of uncertainty and tracking resulting performance over time is the most reliable way to gain a deeper understanding of what exactly is uncertain about player projections and perhaps eliminating some of that ambiguity in the future. Read the rest of this entry »
The Home Run Derby (HRD) is one of the most popular MLB events of the year, seemingly as popular among the players as among the fans. Everyone enjoys watching the best players in baseball launch 450-foot home runs while the non-participating All Stars towel the hitters off and cheer wildly for the most spectacular hits as they head over the outfield seats. But it is also one of the most controversial events, since it rewards something that every little leaguer is warned not to do — swing for the fences with every pitch. Some commentators believe that there is a pattern of derby participants exhibiting declining production in the second half of the season, and they argue that participation in the derby is to blame, because, they say, it ruins the swing plane of the participants. If we can put this theory to bed, then, if nothing else, it would take a little bit of stress off of a really fun night. If an effect does exist, however, it would be useful for front offices to know this before sending their stars to their potential demise.
It has become commonplace in the statistically minded baseball community to view the “Home Run Derby Curse” — the decline in productivity for HRD participants — as an example of misguided traditionalist folklore. The statistically savvy point out that people are selected for the derby exactly because they are overperforming their “true” talent level and because they will perform closer to that true talent level in the second half. Considering that, it is reasonable to assume their second-half performance will be worse than their first-half performance — a rather pedestrian example of regression to the mean. However, the argument usually stops here, as if somehow the concept of regression alone is enough to prove the non-existence of a curse.
The fundamental challenge in rigorously exploring whether or not the Home Run Derby caused a decline in production for an individual player is the same as for many arguments about causality — in order to firmly establish (or dismiss) the claim, we would need to imagine a counterfactual world in which that player did not participate in the derby and then we could see the difference in second-half production. That, of course, is impossible. One approach to addressing this challenge is to consider a collection of players whose statistics are similar to the HRD participant but who did not compete in the derby and look at the difference in second-half production. If we do this with all HRD participants, we should be able to see any general effects, if they exist. Read the rest of this entry »
After last week’s deep dive on Gerrit Cole’s release point change and resulting spin increase, I decided it was time to brush off the old physics textbooks and try to identify a causal link between the two. Before I get into the results, I’ll warn you that the second part of this article where I talk about which mechanical changes correspond to the trends we see in the data is almost entirely guesswork. I’m in way over my head on this stuff and you should consider most of it wild speculation in the hopes of provoking the interest of people who can write “biomechanics” without a spell-checker. But as my dad (who happens to be a mathematician himself) has said, “sometimes asking the right questions is more important than finding the answer yourself (Forman, 2018)”.
I think it’s important to explain to readers why I decided to revisit the question of release point and spin. Up to this point, baseball Research and Development departments and private labs like Driveline have learned an incredible amount about the effects of spin on a baseball; however, how to increase one’s own spin rate remains to be understood.
The significance of this research should not come as a surprise to anyone who has been paying attention to baseball since the public dissemination of Trackman data. As noted in last week’s piece, Trevor Bauer has spent five years of his life trying to naturally boost his spin rate and I’m guessing he’s not the only pitcher going down that rabbit hole. If this link between release point and spin truly exists and is widely generalizable, breakout pitchers could be identified long before their true talent level is shown in their ERA and WHIP. Observers could test the sustainability of a pitcher’s success just by looking at changes in their release point. As this summer’s historically slow free-agent market has demonstrated, teams are starting to turn inward to their player development systems for a cheap, alternate talent pool. If this research is confirmed, teams could unlock the true spin potential of their own players, consequently spiking the talent level of the entire field (which fans of the game like myself love to see).
More than anything, this research question makes me excited about the future of baseball. I see a baseball future in which pitchers intentionally vary their fastball spin rate to high and low extremes to get maximum separation on their four-seam lift and sinker drop. One where hitters take batting practice off of virtual reality AI replications of pitchers with realistic spin patterns and pitch physics so their first time facing the pitcher feels like the third time through the order. Harnessing spin rate is not just another tool to which the rest of the league will soon respond. It is an entire framework for understanding the game we all love that changes the nature of the competition itself. Now, how do we get there?
First, data was scraped from Baseball Savant on every pitch Gerrit Cole has thrown in the 2018 and 2017 season. Because we want to examine within-pitch spin variation, a subset was created containing only four-seam fastballs. A simple linear regression was run using all available release point coordinates and release velocity. We use the variables “release_pos_x,” “release_pos_y,” and “release_pos_z” as regressors. X-axis release point is measured from the center of the rubber from the perspective of the catcher, so right-handed pitchers will have negative values. Z-axis release point measures the height of the release point using the bottom of the rubber as a baseline. Y-axis release point tracks the extension of the pitcher. All measurements are in feet.
First, the estimates suggest that there is a positive relationship between an increase in y-axis release point and the spin rate of that pitch. The plot below demonstrates this. Velocity is listed on the x-axis because it is such an important predictor of spin rate. To see the effect of y-axis release point, pick any given velocity value and look at the difference in spin between a point with a relatively small y-value and a large one. The results are pretty jarring:
The color of the points represents how many standard-deviations away from Cole’s mean spin-rate that pitch was. Because spin-rate varies so much from pitcher-to-pitcher, we should look to see how changes in release point affect within-pitcher spin variation.
This same observation between y-axis release (extension) and spin has been documented previously in Nagami et al., as follows:
“The angle at which the fingertips reached forward over the ball during the top-spin phase was highly correlated with ball spin rate. In other words, ball spin rate was greater for the pitchers whose palm was facing more downward at the initiation of the back-spin phase.”
Because the angle between the palm and ground increases as release position along the y-axis increases, we can confirm our intuition: the longer you hold onto the ball, the more spin it has. Can this be used to help transform pitchers with mediocre fastball spin to elite rotation anchors as has been seen with Gerrit Cole this year? To answer that, we need to have a more sophisticated understanding of the biomechanical process of spinning the baseball.
Again, Nagami et al. has an answer,
“The greater the ball speed, the more downward it must travel. To accomplish this, pitchers with a faster speed would need to hold the ball longer, which means that the palm would have to face more downward at the initiation of the back-spin phase. This would result in a longer period for acceleration to produce spin, and thus produce a higher ball spin rate.”
This suggests that because higher velocity pitches have to be thrown at a steeper angle downward [because downward acceleration due to gravity has less time to act on the pitch], the pitcher then holds the ball longer as it is traveling down the y-axis and thus has more time to impart spin on the ball. Work is force times distance. If we want to transfer more energy into an object, we can either increase the magnitude of the force or apply it across a larger distance vector. We already knew that higher velocity pitches have higher spin. The results of our regression, however, suggest that even after controlling for velocity, release position along the y-axis (that is, releasing the ball further in front of the rubber) has a statistically significant effect on the spin rate. This means that for two pitchers with equivalent velocity, a one-foot increase in y-axis release increases the spin rate of that pitch by half a standard deviation. While no pitcher can actually extend his release point by an entire foot, small adjustments in spin can have career-altering results. In combination with a velocity increase and z-axis release point increase, it seems Gerrit Cole has found his optimal release point for maximizing spin. If this isn’t his peak, the MLB better look out.
Next, there is the problem of accuracy. Can an individual pitcher adjust his y-axis release position to improve the spin rate of his fast ball to a significant extent while still throwing strikes? The answer seems to be yes. As the spin rate of a pitch increases with fixed action of rotation, the deflection force increases orthogonal to the velocity vector of the ball. It speeds the air above the ball, which decreases the air pressure relative to the air below it. The air below it travels upward, pushing the ball along with it and generating “lift”. This is referred to as the Magnus effect. Not only does this means pitchers can spin the ball more without sacrificing strikes, but the Magnus effect alone makes pitchers more effective for two reasons. First, because hitters cannot optically track the ball in the last few milliseconds of the pitch, their brain oftentimes has to linearly extrapolate the trajectory and guess where the ball will end up at the point of contact. This means a small amount of lift can create the perception of a “rising fastball” in the batter’s mind. Second, vertical ball movement decreases the area of pitch-plane and bat-plane intersection. More simply put, the ball is harder to hit with upward movement.
Second, and perhaps more surprisingly, a higher z-axis release point was significantly correlated with spin rate. Last week I forgot to mention that clicking on these plots takes you to my official “plotly” page where the graphs are all cool and interactive, so try it out if you’re interested.
I tried to find a convincing explanation for why the estimate for the z-axis was positive without any luck. A few potential explanations come to mind. First, the higher you hold the ball, the more gravitational potential energy it has. Conservation of energy and the fact that the ball is thrown downward suggests that extra potential energy could be transferred to rotational kinetic energy, which is directly proportional to angular velocity. One of the problems with this theory is that, in general, the gravitational potential energy is not large enough to have a significant impact on spin compared to the overwhelming kinetic energy the pitcher is transferring to the ball.
The second (and more likely) potential explanation I came up with is that when pitchers throw with a three-quarters delivery, they decrease the component of force that they exert orthogonal to the moment arm on the ball. This is the only force that matters for torque (and the resulting rotational acceleration). When managers say the pitcher throws “through” the ball instead of “around” it, this is what they’re talking about. The rest gets transferred as translational kinetic energy, which is applied to the center of mass and contributes to what we call “velocity”. However, theoretically the math does not change along with the arm angle. The only thing that would change is the spin axis, which means the Magnus effect would have less of an upward component and would push the ball sideways. Because Trackman calculates spin regardless of the axis, this should not affect our estimate. The change would have to be a mechanical quirk that could be picked up on a high-speed camera.
We have to keep in mind, however, that not all spins are equal. For example, throwing over-the-top has the same transverse spin rate but adds gyro-spin. Gyro spin is the spin of a projectile which is rotating around a spin axis that is parallel with the direction of the velocity vector (as shown in the picture below). This is sometimes referred to by those within the industry as “not useful” spin, due to the fact that it does not trigger the Magnus effect. This change would again have to be due to another mechanical quirk at the release point that are beyond my abilities to track as a college undergrad who has no biomechanical experience and a Khan-academy video’s worth of knowledge. Answering the question of why z-axis release height is correlated with spin rate really should be left to a dedicated biomechanical researcher who has access to a lab.
Our next task is to test whether or not this trend is generalizable. This is a little easier said than done. In order for release point to be a useful regressor, it has to be variable so that we can test the effects of a change. The problem is that release point consistency is also a skill that Major League teams prioritize both for command and tunneling (making two distinct pitch-types seem alike until the very last second). Ideally, we’d have release point data distributed as a Gaussian, but for now we will have to make do with release point varied as a conscious effort by the pitcher. That causes another problem: if our regressors covary with a variable that correlates with spin rate and that variable is erroneously left out of the regression, it will create an endogeneity problem. This is especially prevalent with release point data that is roughly constant until a conscious correction is made, meaning the release point varies with time (along with potentially velocity, a different pitch-mix, stride length, workout regimen, etc.). This means a study of multiple pitchers will have time-variant error. We are using a fixed effects model, meaning that we time de-mean both the regressors and variables of interest (as shown below). Data on every four-seam fastball thrown by this year’s starting pitchers over the last two years was collected and spin was regressed on the release point. For those following along at home, we used the absolute value of the X-axis release position so we get the measure of sideways extension for both left-handed pitchers and right-handed pitchers.
I’m going to give you a taste of one of the applications of this research. We can calculate predicted change in spin rate by using the regression coefficients above. If we weigh changes in release point, multiply them by the standard deviation in spin, and add them together, we should be able to get an idea of which pitchers making mechanical changes and (more importantly) how important those changes are in terms of spin rate. Below is a list of pitchers who rank the highest in “weighted release point change” based on recorded changes in release point from 2017 to 2018.
First thing’s first, while this isn’t the most important thing in the world, it is comforting to see Gerrit Cole’s name near the top of the list in the metric we created with his spin change in mind. Full disclosure, I was only going to show the first ten pitchers but realized he was sitting at 11th. Still pretty good. Second, there are a lot of interesting names accompanying him. Mike Foltynewicz has made drastic strides this year in limiting hard contact, which has been reflected in his ERA and WHIP. I like Daniel Mengden quite a bit this year. He has had flashes of brilliance including his most recent outing where he limited the Red Sox to 1 earned run over 6 innings. Also, it is interesting how a lot of the guys listed here are known as extreme low-spin pitchers (Dallas Keuchel is a great example). This can also have a tactical advantage by exploiting the flip-side of the Magnus effect. The lower your transverse spin, the more drop you have relative to the rest of the league. For them it might be disadvantageous to be on this list. As a result, it might be worth examining the rate of return a pitcher gets from arm angle changes at different ends of the spin spectrum. We note that some pitchers our model predicts would increase spin rate actually experience a decline in spin rate which demonstrates the complexities of the biomechanical process of spinning a baseball. It should be kept in mind that our model is relatively simple, that our model should be used as a general guideline for understanding mechanical changes and not the last word on spin rate, and that release point should not be studied independent of other factors. For example, more complex models might start by examining the interaction effects of release point changes and velocity to determine diminishing or increasing marginal returns to mechanical tweaks as velocity increases.
As mentioned earlier, the study of spin rate and the relationship between spin and release point has wide applications for internal baseball research and development departments along with casual observers wondering if a short-term spike in spin rate is sustainable. While I realize I’m getting into the habit of ending articles by saying smarter people should take a look and see if this is a real thing, the next step is figuring out exactly why we are seeing these trends in the data. Then, we will finally have a strong basis for answering the question of which factors contribute to a pitcher changing his own spin rate.
Hot take: the Astros are good at baseball. This is thanks in no small part to Gerrit Cole’s early success on the mound. After the news broke that the Astros signed the righty, several analysts wondered why they would give up three national top-100 prospects (Musgrove, Moran, and Feliz) for two years of control over what some called a “soft” upgrade at starting pitcher. We found out in a big way. After 8 starts, he leads all qualified starters in FIP (1.56), WAR (2.8), K-BB% (35.6%), and is second only to Max Scherzer in Z-Contact% (75.3%). He has been, to date, (subject to some debate) the most valuable starting pitcher of 2018.
He is not the only bright spot of the Astros’ 2018. 35-year-old Justin Verlander looks to be returning to his 2011 AL Cy Young self. While it is early, he seems to be a promising candidate for this year’s award as long as his teammate Gerrit Cole doesn’t steal it from him. Charlie Morton, once journeyman, has found his home and filed his very own claim to being one of the very best pitchers in the league. I haven’t even mentioned Dallas Keuchel, proud owner of his very own 2015 Cy Young trophy, who is a top-35 pitcher on STATS’ command leaderboard and is finding his way back to his pinpoint control that helped his team win their first franchise World Series in 2017. Oh… also Lance McCullers is filthy. It’s looking like blue skies ahead for the reigning world champs and the beginning of the season only confirmed the rosy outlook. Then, a simple tweet:
If only there was just a really quick way to increase spin rate. Like what if you could trade for a player knowing that you could bump his spin rate a couple hundred rpm overnight…imagine the steals you could get on the trade market! If only that existed…
— Trevor “Tyler” Bauer (@BauerOutage) May 1, 2018
If only there was just a really quick way to increase spin rate. Like what if you could trade for a player knowing that you could bump his spin rate a couple hundred rpm overnight…imagine the steals you could get on the trade market! If only that existed…
— Trevor “Tyler” Bauer (@BauerOutage) May 1, 2018
Ever since Trevor Bauer pointed out the potential source of the increased spin rate of Gerrit Cole, there has been a cloud surrounding the validity of the Astros’ recent pitching success. It’s easy to understand the frustration for someone who claims to have spent a solid five years of his life trying to naturally improve his spin rate. The advantage of a high spin rate has been documented extensively in the literature by people a lot smarter than me, so I’m not going to go into it (plus my last encounter with the subject ended with a B in high school physics). All you need to know is the faster the ball spins, the more it moves in the last few moments before it reaches the hitter, which makes it harder to hit which, as you might guess, a pitcher generally likes.
In the Statcast era, teams are clamoring for every inch of an advantage and detecting small changes in fastball spin-rate is everything. If the Astros really are using some sticky substance (or just training new acquisitions in the art of spinning baseballs), we should be able to detect it in some way. Here are the average spin rates before and after the transition to Houston of some of their finest pitchers:
Just looking at this table, however, can be misleading. Average spin-rates can look a lot different depending on where you split the data. We would never know, for instance, if Gerrit Cole’s spin rate spiked to 2300 the start before moving to the Astros, which got lost in the pure volume of data suggesting a lower overall spin-rate before the move. It is important to understand exactly where this significant change in spin rate occurs.
The key behind detecting significant changes in data is this:
where X is the observed prior data. η here is what’s called a changepoint. Change point detection uses likelihood-based estimation to find the number of different population means (or variances) in a time series. That’s just a mathy way of saying it looks at how likely the data fundamentally changes as some point (or points). Before we can say the Astros are cheating, we should look at if the change in spin rate is really that significant to begin with and determine where that change actually occurs. We’re going to be using Bayesian changepoint detection. The advantages of Bayesian detection as opposed to Binary segmentation are twofold. First, the probability of having a change point is directly proportional to the prior probability of observing the data. This helps prevent overreaction to new information and makes the overall estimation process much more robust, which is especially important in this case. It is tempting to see a big number next to the post-Astros spin-rate chart and jump to conclusions, but it is important to appropriately weight the prior probability of that spike occurring. Second, detecting the changepoint requires a much smaller window of data. This is important in this case as well. If we are correct that the change happens in a 1-game window, i.e. it happens as a result of a game-to-game transition to a different team, predicting changepoints among small data-windows is especially important. Specifically, our algorithm computes the probability of having a particular changepoint configuration as follows:
where π(η) is the prior probability of that configuration and f(X|η) is the likelihood of the observed data given the change point configuration. There’s some other math behind the detection algorithm, but for now we’ll just take a look at the plots. First up: Charlie Morton.
There’s a pretty clear changepoint here. The posterior probability spikes at timepoint 23. That game date is 9/30/2015, a late-September game against the Cardinals, notably while Morton was pitching for the Pittsburgh Pirates. The Astros weren’t even his next team (in November, he was traded to the Phillies). Several analysts weighing in on the spin question have noted that spin rate is positively related to velocity. In an oft-quoted interview with Matt Gelb of the Philly.com:
“For some reason, I just went out there and tried to throw the ball hard one game. I wound up throwing it harder.”
Below is the change-point detection plot for Morton’s velocity. It looks like there was clearly a change during the spin-rate spike.
Regardless of the cause, our likelihood-based examination suggests it would be naïve to attribute Morton’s spike to an organizational conspiracy to increase fastball spin with a foreign substance.
Second, Gerrit Cole:
There is a clear spike at time 86, which is the time of his trade. Something has changed. However, look at the spike in spin rate at time 44. This could provide a hint to a given organization that a player is capable of a spike in spin rate given a change in mechanics. There is a 20% probability that that game contained a change point, which would be higher except Cole’s spin significantly declined right after that start. If spin rate is associated with a specific mechanical quirk, not only could that help us acquit the Astros, but also identify potential steals on the free agent or trade market that have yet to harness the maximum potential of their spin rate.
Some have hinted that high fastballs increase spin-rate by a significant enough margin to where a change in location could be responsible for Cole’s dramatic spike. Below is the graph of Cole’s spin rate broken down by location.
Cole’s spin rate increases by about 100 RPM when he pitches high and inside. That’s a significant jump, but not enough to explain a 300 RPM spike.
In a recent interview with MLB radio, manager A.J. Hinch mentioned that two things could potentially be behind the change in spin rate (without divulging any organizational secrets). First, he said that sinkers have a drastically lower spin rate than four-seam fastballs, and their pitching staff has prioritized the four-seamer.
Below are his pitch-usage charts before and after his transition to the Astros thanks to Brooks baseball:
First thing first, Hinch is right about pitch usage. Sinker rate is way, way down. But, as you can see below, it cannot account for within-pitch spin variation, as his individual sinker and 4-seam spin are both spiking this year.
The second factor that Hinch hinted at was “getting behind the baseball”. We can examine the relationship between release point and spin rate and see if this can really explain such a significant jump. Below is a 3D plot of spin rate by release position.
This seems like it could actually explain a good bit of the change. There are two changes associated with the spike in spin rate. First, he’s releasing the ball much further along the Y-axis than before. Second, it looks like he is releasing the ball higher, but closer to his body than the pitches at the very top left of the distribution. Almost every pitch he has thrown in the neighborhood of (-2.2, 54.4, 5.6) has been at around 2300 RPM. Further research should be done on the physical mechanics that generate that spin and if this could really be a causal relationship.
After a quick look at Cole’s mechanics, it does look like there is a conscious change in release point. Below is a screenshot from Cole’s start against the nationals on 9/29/2017. This looks like one of those pitches further along the x-axis than our sweet-spot that we found on the release point chart. See how his elbow is almost parallel to his shoulder.
Remember that one-game blip in spin rate that showed up on his changepoint plot? Below is a screenshot of a 97-mph fastball from that start. Notice how his elbow is higher than his shoulder. This could be the change that Hinch was talking about when he said “getting behind the ball” could be behind the increase in spin. When watching the video of the previous pitch, it looks almost as if he’s throwing around the ball instead of throwing through it.
Apologies for the blurry screenshot (it was one of two fastballs with media on baseball savant). Lastly, here’s a screenshot from this year. His release point is much, much more vertical than the previous two.
Overall, we should not jump to conclusions on the Gerrit Cole spin question. Just to be perfectly clear, I personally have no idea how mechanical changes actually affect spin rate. I haven’t done the experiments myself and certainly have not spent as much time as Trevor Bauer trying different grips and substances in a controlled setting. However, this article suggests that there is an association, whether it be correlative or causative, in Gerrit Cole’s release point that has come with an increase in spin on both two-seam and four-seam fastballs. If further research can confirm this association, the results could be of incredible use to teams looking for value either in their player development system or trade market.
I went to SunTrust Field on July 4th to see the Astros versus the Braves, and little did I know the fireworks during the game were going to be far more explosive than the ones afterward. I got to see the most productive offense in baseball put up 16 runs against the Braves (and their usually stellar rookie pitcher Sean Newcomb). It’s undeniable how remarkably talented the Astros offense is. So, I set out to answer a seemingly-simple question: “why?”
I got the idea for this article while I curiously flipped through their team statistics to try to see what clicked. My first reaction when looking at batted ball data, hard hit rate, K%, BB%, etc. was “wow…they look entirely human.” It was strange how normal the advanced metrics show the Astros to be. They walk a decent amount and don’t strike out a lot, but they’re tied with the Pirates and behind the Red Sox in BB/K, so while that helps, it can’t be the only reason they’re this good. They lead the league in ISO, which tells us that their slugging percentage is pretty high, but doesn’t really explain why.
The table that jumped out at me was this:
In my article about Brad Peacock, I talked a little bit about the utility (and disutility) of using pitch-type linear weights to predict future pitching success (in short, the variation from year-to-year for pitchers is pretty high). Well, this is certainly more than noise. The Astros hit the absolute crap out of fastballs. For me, the only question is whether this is simply a result or if it can be used as a predictor.
I think there are two important distinctions to draw between the way we’re using pitch-type linear weights here and what we did earlier.
First, these are team-wide as opposed to pitcher-specific. The problem with these is that they capture a lot of context. That’s really the whole point of the stat. If you hit a fastball for a single and score the two runners from scoring position, your wFB is going to be much higher than if you begin the inning with one. However, because we’re looking at team-wide statistics, we should be getting measurements from enough distinct contexts that the noise begins to fade slightly. This is much different than examining Peacock’s statistics after four starts (two of which happened against the same high-strikeout team).
Second, we’re looking at change in offensive run expectancy, not defensive. This could be a metric like BABIP, where pitching is highly dependent on external factors, while it is a bit more focused on skill for hitters. This raises the question “is there such a thing as a fastball hitter?” The linked article was the closest I could find to directly answering the question (I’ve looked all over the place and can’t figure out who the author is, but as far as I can tell he’s decently connected with the baseball analytics community). Either way, if we take this guy at his word, his model does suggest that some hitters do perform significantly (statistically) better against fastballs.
This doesn’t completely solve all the problems with just looking at wFB (like the fact that the Astros are a naturally good-hitting team and will have many at-bats with runners in scoring position where the expected run value will be high), so we will have to devise a way to control for the context and just look at performance on a pitch-by-pitch basis.
Instead of reinventing the wheel with trying to prove the existence of fastball hitters, this article seeks to prove a strong relationship between fastball-hitters and overall success. If so, this could have drastic implications for how to pitch to teams like the Astros, or even which players to target when trying to replicate the “worst-to-first” transition as the Astros have.
Here’s the relationship between fastball linear weights for 2017 teams and their weighted runs above average:
This should be an obvious relationship. For starters, they’re both weighted measures of how many runs a team is expected to produce. So, if you produce more runs on fastballs relative to other teams and hold everything else constant, you by definition are going to increase your total runs relative to other teams. However, the R-squared value does raise some eyebrows. Basically, this means that about 80% of the variation in “weighted runs above average” can be explained only by looking at the wFB statistic of each team. In a sport with as much variation as baseball, that’s actually fairly decent. But, we can do better.
I decided to look at the fastball linear weight for each team and divide that by the total wRAA. This should give us a decent idea of how many of those runs created came on certain pitches. This is kind of a weird comparison, because wRAA calculates the expected runs based off of weighted on-base average (wOBA), while pitch-type linear weights scoop up all of the situational context as explained above. However, it can still give us a general idea of how many expected runs a team generates off of a certain pitch relative to league average.
Here’s the relationship between fastball run percent versus weighted runs above average (please excuse the shaky highlighting):
This surprised me quite a bit. What this says is that about 76% of the variation in wRAA (in 2017) can be explained solely by looking at the wFB/wRAA. I should also mention that even though the dataframe is called “astros” it includes the stats for every major-league team in 2017. In short, the percentage of runs you generate off of fastballs correlates pretty strongly with the total amount of runs you score. Weird stuff.
There are, of course, many problems with latching on to this one observation as the basis for total change in team management. For one thing, some teams may be good at hitting fastballs just because they see them more than every other pitch. A complete 180 in the way teams pitch could bring with it a heavy response by the league’s hitters that mitigates some of the advantage, making this a self-denying prophecy.
That being said, I have a hunch that the Astros are a good fastball-hitting team by design. I think the fact that roughly 60% of pitches seen by hitters are fastballs is an inefficiency that the Astros are effectively exploiting. I also believe this is part of the logic in the recent curveball revolution. Theo Epstein’s short and sweet reaction to the pitch usage of the 2016 World Series between two of the most data-driven organizations in baseball was simply “More breaking balls!”
It’s worth noting, however, the Astros haven’t always been this dominant with respect to hitting fastballs.
The expected run change this year is almost three times as much as last year’s value. If this is by design, why didn’t it happen last year? I’m not entirely certain, but the ridiculousness of the 95.6 value they put up in the first half this year should be a major tip. That doesn’t just happen by accident. If they have a wFB of 0 for the entire second half of the 2017 season, they would still be good for the 26th best fastball-hitting season of all time (…since 2002). That is absolutely absurd. If you’re a fan of the game, please recognize just how good the Astros are. If you’re a pitcher, you better hope they don’t make an example out of you. And for your own good: don’t throw them too many fastballs.
I submitted my last article on Brad Peacock literally 10 hours before Jeff Sullivan published an amazing article on exactly the same topic (using exactly the same GIF of Adrian Beltre looking ridiculous). So just to prove I’m not copying anybody, I decided to write about someone whose shocking appearance on the Statcast leaderboards made me confident this wouldn’t happen again.
In the month of June, Lonnie Chisenhall’s hitting .385 with 4 HR and 15 RBI. By any account, this is a monster three weeks (especially for somebody not named “Cody Bellinger”). Last year, his .328 BABIP (26 points above his career average) cast doubt over his success, however his .374 wOBA and .360 xwOBA (a stat that uses exit velocity and launch angle to predict what “should be” a hit) prove that his success this year is no fluke. There’s something different about this guy that means this inflated BABIP may just be signal and not noise.
Baseball analysts used to believe that BABIP for most hitters regressed to the mean (after you put the ball in play, the rest is just luck) until people started tracking year-to-year differences for players like…Mike Trout, whose BABIPs are high and correlate strongly across years. Now we know that batted-ball data (exit velocity, launch angle, launch spin rate, etc.) hugely influences the results of balls when they get put into play, which means projections have been undervaluing players who consistently make hard contact (Nick Castellanos, Miguel Sano, etc.). Barreling balls isn’t just a fluke; it’s a skill.
The first thing that immediately jumped out at me was how much hard contact this guy makes. Among the league leaders are max effort home-run-or-die-trying type hitters (Khris Davis, Matt Davidson, Joey Gallo), freaks of nature (Aaron Judge and Mike Trout), headline-capturing breakouts (Justin Smoak, Miguel Sano) and…Lonnie Chisenhall. I know FanGraphs readers love player comps, so here y’all go:
Chisenhall closely resembles the batted-ball profile of a Ryan Zimmerman or Giancarlo Stanton when he puts the ball in play. The reason both he and Zimmerman have very similar results in the third column (Barrels/PA) as well is that they have roughly the same batted-ball events per plate appearance (17% K-Rate vs 18% respectively). The major reason for a divergence in the third column with Stanton is his absurd strikeout rate (16.7% K-Rate in 2016 to 25.4% in 2017). The reason I included him is just to give you an idea of the absurd power this guy has. He’s ahead of Paul Goldschmidt, Joey Votto, and Miguel Cabrera in terms of Barrels/PA, mostly because they get walked way more.
What’s more, he’s trending in the right direction. Below is Chisenhall’s distribution of exit velocity over the past two years (red is 2017).
As you can see, his soft contact is slightly more concentrated at about the 70 mph range, with much more hard contact this year in the 95-105 range. Chisenhall has made an adjustment…it’s just a question of what.
We’ve already read a million articles on the “air-ball revolution” detailing how hitters are attempting to elevate the ball — even at the expense of overall contact rates — to produce more home runs in total and thus more offensive production.
While it’s lazy analysis too easy to simply say the words “air-ball revolution” and call it a day — although with Daniel Murphy and Yonder Alonso, it might just be the decisive factor — changes in swing plane can drastically affect batted-ball data.
Two things are incredibly important in this debate that most people misunderstand.
First, it’s the “air-ball” revolution, not the fly-ball revolution. Mike Petriello, Statcast analyst at MLB Network, defines an “air-ball” as any ball hit above 10 degrees of launch angle. Fly balls are good, line drives are good, ground balls are less good. Strikeouts are bad.
Second, putting the ball in the air only works in tandem with exit velocity. There’s no sense in Eric Young Jr. trying to elevate everything, which would result in lazy fly balls, mitigating the benefit of his blazing speed. But, in aggregate, hard-hit balls are more productive when hit above 10-degrees launch angle rather than below. We’ve already established that he’s mashing this year. Now lets take a look at where they’re going.
As you can see from the above chart, a good amount of Chisenhall’s batted balls were directly at the 10-to-30 degree mark, while his hits were almost all at the 10-degree mark. Those were solid line drives, but leave very little room for error. Low-EV balls at the 10-degree mark likely result in ground balls or soft line-outs to infielders. Also, there is a decent amounts of batted-ball events directly below that. If your mean launch angle is 10 degrees, it’s very likely that small misses yield more ground balls than you intend.
This year, the vast majority of his batted-ball events are directly at the 20-degree mark. It seems like not only is he hitting more “air balls,” but they’re solid line drives that afford him more room for error. Even when he doesn’t hit frozen ropes, they still have a shot at becoming base hits. Also, there’s a much more apparent difference between air balls and ground balls, with many more hard-hit balls at 30 and 40 degrees.
The graph above confirms this. Especially on breaking balls, more and more are finding the air.
A lot was written about the importance of plate discipline during the meteoric rise of Eric Thames at the beginning of the year. Petriello details the dramatic difference between swinging at strikes and balls.
“From 2015-16, Major League Baseball hit .292 with an average exit velocity of 89.3 mph on pitches in the zone, and only .168 with an 81.4-mph exit velocity on contact made outside it. Nearly 91 percent of all homers over the past two seasons are on pitches in the strike zone. These are massive differences. Learning plate discipline may be the hardest thing for a batter to do, but it’s also potentially the best thing he can do.”
You might think this is obvious but for major-league hitters geared up to hit a 98mph fastball, laying off of a dirty breaking ball that sweeps through the zone is one of the toughest things to do.
For Chisenhall, this has been one of the more drastic changes in his profile over the past few years. Below is a chart of the change in K-rate over the last six years. The league average has been trending up, starting at 18.5% to about 21.5% at a relatively stable rate (some speculate this increase in K-rate is a result of the move of major-league hitters to elevate more).
As made pretty clear by the above graph, Chisenhall’s strikeout rates have been moving pretty steadily in the opposite direction. A hitter with negatively trending K-rates without sacrificing power is rare. This was the precursor to the Justin Smoak breakout this year.
His walk rate skyrocketed to 9% this year from 5.5%. What’s the culprit?
His O-Swing% is down 10% this year, and his Z-Swing% is down a bit, but nothing crazy. He’s clearly seeing the ball better this year.
Here’s a clip of 2016 Chisenhall on a changeup outside. Pay attention to 1) bat plane and 2) where his front foot lands in relation to the edge of the batters box.
In terms of bat plane, he does go down to get this pitch a little bit, but finishes below his shoulder with one hand on the bat. Just a note about letting go of the bat: it’s fine and the majority of major-league hitters do it, but you have to make sure you’re not cutting your swing short as a result.
As for foot placement, it looks like he’s close to square with the pitcher, maybe a little bit closed, and about two feet from the edge of the plate. On this swing, he produces a fly ball at 88mph that ends in the stands.
This is a heatmap for where pitchers located the ball against him in 2016. Notice a pattern?
It seems like pitchers realized that mistakes inside are what make him the vast majority of his money (as will be shown in a zone map slightly later).
Now take a look at this video detailing a swing on a nearly identical (albeit overhanded) pitch from 2017.
This time, he finishes his swing with both hands on the bat above the shoulder, with the bat traveling on a slightly higher angle (though nothing like the transformation we’ve seen from some of the more absurd poster boys of the elevate-and-celebrate craze mentioned above).
I think the most interesting change for this swing, however, is that he strides fairly substantially closed and his foot lands about a foot from the plate. Instead of a lazy fly ball to the stands next to the left-field line, this produces a 103mph line drive to right-center.
In a recent FanGraphs article by Eno Sarris about the importance of changing where the hitter stands in the batters box, Anthony Rendon (one of my personal heroes who I grew up watching at Rice University) says,
“If a righty dives, we sell ourselves short inside, so if I’m getting crowded, and I’m hitting the ball late and deep, let me scoot back, and so on the same swing, instead of hitting here [on the handle] and fouling it off, I’m hitting it closer to the barrel and hitting into right field.”
It looks like Chisenhall’s position in the back of the box with a slight move toward the plate is designed to correct that issue. His hands are so fast that he can consistently make contact with the inside pitch. His best hitting in 2016 happened on pitches off the plate inside. However, it doesn’t matter how good you are at driving inside pitches if no pitcher is ever willing to give you one. Making an adjustment to put solid swings on pitches on the outer third is the right call in this situation.
I would be a little bit worried about this change from the perspective of hitting the inside pitch, but it looks like a concrete adjustment in response to the outside pitch was warranted from the previous season’s heat map. It seems that pitchers have picked up on this adjustment as well. Below is the distribution of pitches for 2017. Pitchers have made a dedicated effort to react to Chisenhall’s changed hitting style and adjust accordingly.
So? Is this adjustment working? Here’s the map of Chisenhall’s batting average in 2016 broken down by zone provided once again by the lovely Brooks Baseball website.
Anything strange? Um…he’s hitting over .900 on low to middle pitches off the plate inside and .160 on pitches middle-middle and middle-up. What?? He was hitting balls outside of the strike zone way better than meatballs down the middle of the plate. I think this is a big reason for the spike in walk rate we’ve been seeing this year. The videos we saw above show that he’s scooted up to the plate a bit more than he used to, meaning the alarming shade of red (indicating that he’s getting hits with regularity) shifted over the plate and now he can simply take those pitches for balls (and eventually, walks).
Yup! That’s exactly what happened. He’s also getting to pitches off the plate outside a bit more to protect the plate in two-strike counts. This could also help limit strikeout rate. Overall, a better approach toward outside pitches combined with adjusting placement within the batters box have served him well.
So, lets recap what we’ve seen here. First, Chisenhall hits the ball…hard. Second, walk rate is up, strikeout rate is down. Third, more and more balls are being hit in the air (without overly focusing on hitting fly balls). Lastly, all of these changes seem less like a product of batted-ball luck and more like the result of intentional change in approach at the plate. We can expect this success to continue at a current or similar rate.
Some red flags could signal and end to this dominance, however. First, I would keep my eye out for a change in walk rate. While he stopped swinging so much on inside fastballs off the plate, he’s still aggressively swinging off of the low-outside corner. This shouldn’t be a problem as long as he keeps mashing those pitches (as a result of the new approach), but the further he reaches, the less likely he is to make solid contact. Second, I’m incredibly interested to see how he responds to this shift to the inside corner. His 2016 results seem to suggest that his quick hands are capable of getting to these without getting jammed. However, on a pitch with high velocity and cut, we could see much more weak contact.
You’re going to hear a lot of people simply point to his BABIP numbers and say that’s unsustainable, he’s just getting lucky, and that he will inevitably regress. That’s lazy analysis and untrue. Until we see any of the above physical adjustments, just sit back and enjoy the Lonnie Chisenhall show. It’s going to be a good one.
Welcome to the Astros rotation, Brad Peacock. In his first six starts (plus 16.1 innings of relief work), he has a 2.82 ERA with a 1.19 WHIP. Wait, I forgot to mention the 66 strikeouts in 44.2 innings. If you’ve ever seen this guy pitch, you’ve no doubt observed his strangely overpowering repertoire and wondered how All-Star hitters can look so ridiculous on a 93mph fastball. For a former top-100 prospect who has spent 10 years playing professional baseball, one might wonder if there was a fundamental change in stuff, if this is just batted-ball luck, destined to disappear as soon as you pick him up on your fantasy team (#thanksJasonVargas), or if the only missing factor was opportunity. Either way, everybody who has seen this Astros team play knows that something special is happening in H-Town. While many believed that the Astros needed at least one new arm to even compete for a playoff spot, the return of Cy-Young, immaculately-bearded Dallas Keuchel and breakout Lance McCullers mean that there was just one missing piece to seriously contend for the World Series. That piece is Brad Peacock.
Sometimes it’s important to subject pitchers to the eye test, which this guy passes with flying colors (unlike a peacock which, while colorful, cannot fly). For this section, I’ll show you a pitch and then a table showing the league leaders for that pitch type. Because of the problems with pitch values (which I’ll dive into in a bit) it’s important to couple that with other methods of assessing pitch effectiveness, including simply watching the pitcher.
While some of you will no doubt roll your eyes at the next section, I think it’s important to talk about what matters when using film study. The important things to look for when watching a pitcher include:
1. Location (duh). I trust the Astros’ game plan to get the job done on most days. More often than not, the relevant question is if the pitcher can execute that plan.
2. Bite. Is there late movement on the pitch? Arm side tail or cut? This can serve to induce weak contact even on meatballs and serve as a helpful backup in case of missed location.
3. Hitter’s balance. While we now have metrics that calculate the likelihood of each batted ball becoming a hit based off of exit velocity and launch angle, those stats can be misleading as well. Sometimes, a hitter can be completely fooled on a pitch and still get the barrel on the ball. This could change the expected outcome for that batted ball, but not for the same pitch in a slightly different context.
4. Situational pitching. How do they fare in high-pressure situations (3-2, runner on third, battling back from 3-0, etc.)?
OK, sorry for the kiddie stuff. Back to Peacock. Here’s a straight 3-2 fastball that generates a strikeout.
This pitch is indicative of a few things. First, it catches a bit more of the plate than you would like to see, although in a 3-2 count with bases empty and 2 outs, it is more acceptable to challenge the hitter than to give up a two-out walk (as long as it’s not flat and down the middle). Second, note the late arm-side tail. This could be a major factor in his ability to generate swing and misses (as will be explored later).
Before I show you the leaderboards, there’s a major caveat. As with a lot of advanced metrics, pitch-type linear weights are more descriptive than predictive. There’s extremely high variance from year-to-year. On top of that, several variables go into each pitch and it’s extremely hard to differentiate the signal from the noise. The results of certain pitches are not independent from one another, but heavily influenced by the hitter it’s thrown against, the situation it’s thrown in, the pitch it’s following, etc. This all is amplified by the fact that his start of the season in the bullpen has limited his innings total and a good chunk of the results we are seeing are against an Oakland A’s lineup that swings and misses…a LOT.
That being said, if we continue to see major run savings from Peacock as his innings total climbs, the continuation of the trend as more and more context changes, the more we can be confident that this is more a product of skill than anything else.
Here’s how that pitch has fared against the league writ large. Below is a table of the league leaders in weighted pitch value for fastballs. Feel free to peruse the familiar names of his company on the leaderboard.
You might be asking yourselves why I chose to write about Peacock and not Jaime, Walker, Nova, etc. Aside from the amazing name, it’s because this guy has not one, but two All-Star offerings.
Next, his slider (brace yourselves):
Did you see where that pitch starts versus where it ends?? Is he throwing a wiffle ball? The look of frustration on Beltre’s face is not unique to this at bat. That slider has been doing that to hitters all year. Below is the table of weighted pitch value leaders for sliders in the MLB this year.
There’s a good chance that in high-leverage situations, this pitch can generate a swing and miss like it has so many times this season. There’s a reason (which you probably picked up from watching the above gif).
There are five sliders in the entire majors with more break than this pitch.
Overall, while there are problems with looking at pitch values in isolation, we can be slightly more confident that this is the product of solid game plans from the Astros coaching staff, Peacock’s ability to carry it out, and simply a dirty slider.
OK, so we know his pitch mix is fairly strong. What does this mean for his results? When I examined the batted-ball data, three things immediately jumped out:
1. When hitters swing at strikes, they miss.
This much is clear from the section above. This guy truly has overpowering stuff without lighting up the radar gun. Z-Contact is cool because for pitchers with low velocity, it serves as a kind of proxy for movement, sequencing (thanks to the coaching staff for that one), and locating pitches on the corners.
If anyone takes issue with Danny Salazar’s inclusion on this list, it’s important to keep in mind what gave him so many problems was not nastiness of pitch mix; it was mostly a combination of walks and bloated HR/FB ratio. He actually inspires confidence in Peacock, as you can think of him as a Salazar without the walks who keeps the ball in the ballpark (as you’ll see… right now).
2. When they do make contact, it’s soft.
Dickey’s appearance on this list should not be a surprise. Because of the low overall velocity of his pitches, the batter must generate the majority of the exit velocity. While Keuchel isn’t a knuckle-baller, his average fastball velocity of 88.7 mph means the same logic applies, although not to the same extent. Peacock’s appearance on this list is particularly impressive because of his average fastball velocity of 92.3. This means that although he is providing more force than the aforementioned players, he is still inducing roughly as much soft contact.
3. Even with hard contact, the ball stays in the park.
This last table is, I think, what makes Peacock a candidate for greatness.
First, generating swing and misses is one thing, because it intrinsically decreases the likelihood of home runs, RBI, etc. But if you tell me this guy limits the pool of potential home runs AND the likelihood of each individual batted-ball event in that pool turning into a home run, we’ve stumbled on something special.
Second (and most important), he’s in danger of being undervalued. It used to be thought that BABIP (batting average on balls in play) was essentially random for every hitter, so the ones with higher batting averages simply put the ball in play more often. Later, sabermetricians discovered that, in fact, some players just had higher BABIPs. Mike Trout hits the ball harder than Replacement-Level Joe, so Mike Trout is more likely to get base hits when he puts the ball in play. The latest version of this logic is playing out in the debate over HR/FB rates. Baseball analysts for years have been developing models that regress the HR/FB rate in the direction of the league mean. The thinking goes: if someone hits a hard fly ball, several factors have a large sway in determining if it leaves the park (park factors, weather patterns, time of day, etc.) outside of the hitter’s control. These days, however, several pitchers are sustaining remarkably high HR/FB ratios (Gerrit Cole – 18.2%, Lance Lynn – 18.8%). If it turns out that this ratio actually has more to do with the pitcher, then these players are being systemically overvalued in projections. The flip side remains true. Projections tend to undervalue players with low HR/FB rates, because it ignores the skill involved in limiting the amount of hard contact that leaves the yard.
What’s next for this Peacock? We can see clear trends in his pitch usage that suggest this dominance will continue. The below chart from our friends at Brooks Baseball shows what I believe to be the root of his recent success.
Two things to note:
1. He’s throwing the slider much more often as the season progresses. It’s his best pitch and the contact rate hasn’t changed as hitters are seeing it more and more.
2. A sinker is in the works! Several analysts note that to be a major-league starter, you need three effective pitches to keep hitters off balance and be effective the third time through the order. While there are some problems with these studies (because of the preordained conclusion, there are very few two-pitch pitchers in the majors, resulting in self-fulfilling prophecy), it is at the very least comforting to see another pitch in the mix. While Peacock’s curveball generates a high number of whiffs, he has trouble commanding it late in the game relative to his slider, so its usage is limited.
Lastly, for Peacock to be a truly effective starter, he’s going to need to go deeper into games. Strikeout pitchers generally bloat their pitch counts in an attempt to generate swings and misses early on. Since transitioning to a starter, he’s averaging 4 2/3 innings per start. Part of this will be resolved with experience as a starter, but in the meantime this new sinker could mean more ground balls, saving his swing-and-miss stuff for later.
For now, at least, it looks like the sinker is doing just that. In his last start, Oakland slugged .200 against it. With low whiff rate, high ground-ball rate, and low slugging percentage, this could be the tool to get Peacock deeper into games where the curve and hammer of a slider can take the wheel from there. While we’re finally seeing what makes this guy so special, he’s already making the adjustments needed to become an elite starter. This peacock is finally showing his tail feathers, and we haven’t seen anything yet.
To begin the 2017 season, Mike Leake was one of the most cautiously optimistic targets for a breakout season. His low velocity and K-rate had a lot of people worried about how sustainable the success was. But, for a while, he led the league with a 2.03 ERA (5/23). He ended April with 33.1 IP and 5 ER total, good for a 1.35 ERA. While his success came in the face of Jason Vargas stealing all of the low-velocity, soft contact-inducing, ERA-leading thunder, he generated plenty of buzz as a welcome surprise in the Cardinals rotation after a shaky April and beginning of May by resident pitching-staff wizard, Carlos Martinez.
Part of this was certainly soft contact combined with luck to create a stellar (but unsustainable) LOB% of 86.5% of baserunners (warning: that article contains an extended metaphor comparing him to “leek soup”). But even in the midst of a brilliant start of the season, many analysts warned about the impending reversion to normalcy, referring to previous stunts of brilliance at the beginning of the season. Since the beginning of June, he has surrendered 7 ER in 11 IP, for a 5.40 ERA. While this is not a disaster when compared to other pitchers who have flamed out (cough, cough, Kevin Gausman), for those who were hoping this was a turn of the page in the story of a 29-year-old soft thrower with roughly a 4.00 career ERA — what happened?
First, Leake’s sinker velocity has changed in slightly different ways than one might imagine. Below is the chart of his sinker velocity with June in red and the rest of the season in blue:
The most noticeable change is the slight uptick in sinkers for the 92-93 mph and 89-90 mph range, with less of the 90-92 mph variety. For most pitchers, this would correspond to an increase in swings and misses, but for Leake, a pitcher who relies on command and finesse, this has a minimal impact on overall performance. Also, it should be noted that at a certain point, an increase in velocity has higher returns (e.g. a jump from 92 to 95 mph as exhibited by Brewers breakout-ace Jimmy Nelson), but as MLB hitters are used to seeing slightly faster sinkers than Leake’s with less movement, this increase in velocity has small (perhaps even negative) returns on his performance. When I looked at the chart for contact rates broken down by velocity quantile, this phenomena was ever present, although not as prominent for his sinker, but his cutter.
The cutter quantiles were based on splitting the distribution into thirds and were defined as follows: “Slow” (v < 89 mph), “Medium” (89 mph < v < 90 mph), & “Fast” (v > 90 mph). As shown in the above table, the way to miss bats with this cutter is to keep it below 90 mph, and Leake seems to be moving in the opposite direction. The histogram below charts changes in cutter velocity, red being the distribution in June. While he decreased the amount of cutters directly at 90 mph, the number close to 91-92 mph (danger zone) increased, along with the low-velocity 87-88 mph cutter. Also, we can’t rule out the possibility of an injury with a much wider variation in velocity (although there are more reliable metrics for judging injury risk, like variation in spin rate).
With the changeup, we see the same story. The changeup quantiles were: “Slow” (v < 85 mph), “Medium” (85 mph < v < 86.5 mph), & “Fast” (v > 86.5 mph). Again, the way he misses bats with this is to keep the velocity under 85 mph. This histogram below categorizing the change in changeups shows that this may be the culprit.
Many more changeups are being thrown in the 87-88 mph range, which is really dangerous for a pitcher like Leake whose fastball does not get much faster. A major goal of throwing changeups, especially early in the count, is to disrupt the hitter’s timing. Little research has been done on the optimal separation in fastball and changeup velocity, but generally a 3-4 mph difference is insufficient. It is worth noting, however, that the Statcast pitch tracker system is far from perfect and some of these could very well be slow cutters.
Here are some pretty telling gifs (from the same game) demonstrating the two types of changeups. The first is a particularly nasty changeup on the outside corner to strike out Yasmani Grandal. He is not only totally off balance, but uses none of his legs and pokes, trying to stay alive. This change in velocity is exactly what we should be looking for when getting the feel for changeups.
Now, here’s the high-velocity, flat changeup that has been getting him into trouble.
It lacks vertical movement and just sort of slides through the top of the zone. Utley has zero problem keeping his weight back and engaging his hips to launch it over the right-center field fence, which leads me to my next point.
Next, we can note the situational pitching Leake has had to do in June, relative to other months. Below is the bar graph of the change in frequency of counts he has faced in June:
Most of the count distributions are roughly the same, but he’s pitching in a lot more 3-1 and 3-2 counts. Leake has never been one to walk many hitters, which may explain the increased exit-velocity numbers. When Leake falls behind in the count (and loses command of his off-speed pitches) he often times has no choice but to spin a cutter over the middle of the plate. Previous to this last start (6/14) Leake pitched in significantly fewer 3-0 counts, while the amount of 2-0 counts he was in stayed pretty much constant. This could be a sign that he lost confidence to shoot for the corner on 2-0 and would be more likely to catch the middle of the plate. The alternative of a 3-0 count (or subsequent walk) might be the lesser of two evils.
Also, the deeper into the count hitters get against Leake, the more comfortable they are against his variety of offerings. Leake thrives off of keeping hitters off balance and surprising them with variations in movement. The more time hitters have to track these pitches, the less effective they will be at throwing off their timing.
As we can see from the locational charts from this season before June, Leake’s moneymaker is very bottom of the zone:
Compare this with the zone chart from June and you can see Leake’s concerted effort to throw more strikes has resulted in many more pitches middle-in at the expense of the bottom half that he dominated at the beginning of the season:
Establishing the inside fastball is a great tool for pitchers with high velocity, but with Leake’s pitch mix, it can be dangerous to leave a sinker middle-in if right-handed hitters have the ability to catch their hands up.
Overall, I would caution against reading into Leake’s start of the season as an indication of a fundamental change in stuff. Part of it was most certainly batted-ball luck. Even guys who pride themselves as being soft-contact-inducing studs generally cannot sustain a 0.234 BABIP. Whatever adjustments he made at the beginning of the year have faded. However, look for an adjustment in the coming months to move back toward the bottom half of the zone, especially when behind in the count. I would not be deterred by a slight uptick in BB/9 rate if I saw it accompanied by a decrease in exit velocity. If he can find the sweet spot between leveling out the velocity in his pitches a little more to keep hitters off balance and allow for the most movement possible, we could see another go-around of Peak-Leake.
I was planning on writing about Justin Smoak, but Jeff Sullivan stole my thunder and for some reason people like reading articles written by professional baseball analysts more than articles from college undergraduates (but I guess it’s still worth a read). So, I moved on to the next guy on my list.
First of all, if anyone is going to benefit from their environment in a lineup, it’s Adam Duvall. The Reds have turned out to be one of the most productive lineups in baseball (as a Cardinals fan, it hurts to write that). It starts with the best base-stealer in the MLB followed by the player about to overtake Mike Trout as the best of the 2017 season in terms of WAR, followed by one of the best hitters in baseball, followed by Duvall. He’s protected by a surging Eugenio Suarez, a breakout Scott Schebler (who many in baseball refer to affectionately as “this year’s Adam Duvall”), speedy Jose Peraza, and recently-discovered greatest player of all time, Scooter Gennett. Great American Ball Park has the best right-handed home-run factor in baseball. Overall, Adam Duvall has it good in Cincy.
We’ll start with the most obvious factor in what makes Adam Duvall such a force in the Reds lineup: the elite power. Duvall’s .530 slugging percentage and .258 isolated slugging are good for 26th (right behind Kris Bryant) and 28th (behind Paul Goldschmidt and ahead of George Springer) in the majors, respectively. By all accounts, he is one of the top 30 pure power hitters in the league. This much has not changed. What makes him interesting as a hitter is not a major change of swing plane or pitch selection like Alonso or Lowrie. He has always been near the top in FB/GB rate (20th this season with a 1.22 ratio).
The obvious “yes…but” to all of this is his plate discipline. Yeah…fair point. In 2017, he has a weak 24% K rate, and an even worse 6% walk rate, making a 0.26 BB/K ratio (ouch). We can hope for a Justin Smoak-esque transformation in the future where he starts making contact with two strikes without sacrificing any power, but in the meantime, what we should look for is what happens with the balls he does put in play.
When I examined the batted-ball data, it doesn’t look like there’s a major change.
There are very slight adjustments, some that might fall within the range of statistical noise, but interesting nonetheless. It looks like there’s a slight decrease in the number of fly balls, increasing his GB% by 1 and LD% by 2. It also looks like he’s becoming slightly less of a dead-pull hitter and hitting the ball more to center and opposite field. All of this resulted in a slight uptick in his HR/FB rate. This decrease in fly balls is confirmed by the difference in the two years’ launch-angle charts:
2017 Launch Angle Chart
2016 Launch Angle Chart
It seems clear that this year, in terms of launch angle, there’s a much larger difference between his home runs and fly balls. Last year, the majority of his hard-hit balls were square at 20 degrees. This could explain some of the jump in HR/FB rate.
One of the things that jumps out in Duvall’s stats from this year to last is the major transformation in results in his platoon splits.
What is the reason for this sudden transformation against left-handed pitching? Is it just luck?
It looks like there’s a combination of things at play. First, his BABIP in 2016 was right around league average for both right- and left-handed pitchers. His BABIP against righties basically followed the league average while against lefties it rose to almost .050 points higher than the average. It could be luck…or something has really changed for the rising power hitter.
Here’s one of the coolest changes in Duvall’s performance the last few years.
The 2016 data seems like what you would expect from a power hitter. Weak contact with two strikes, watch out when you fall behind in the count to him, and full-count with first base open, it might be worth walking him. However, the 2017 data shows a major difference. He’s averaging 92 mph exit velocity on 0-2?? He’s not getting cheated on any count. This explains some of the change in BABIP over the past two years. Instead of choking up and trying to make contact after falling behind in the count, he’s more consistently driving the ball. This comes with appreciable increases in exit velocity when ahead in the count 1-0 and 2-0.
My next thought was: maybe this is the result of differences in his approach to certain pitches. This is where stuff gets interesting. I looked at the pitch breakdown for the past two years against Duvall and found major differences between years. More than half of the pitches he’s seen this year are fastballs, 138 of them two-seamers (pitchers around the league have decided low and outside sinkers are the only way to get him out). In those 138 pitches, he has a .481 average and is slugging .852…That’s not a typo. Around half of the time his at-bat ends with a two-seam, he gets a hit. Here’s the breakdown of his results against two-seams by year.
He’s always hit them pretty hard when he makes contact (.570 SLG vs .234 AVG in 2016), but the biggest difference is apparent in the last column: he stopped whiffing on the two-seamer. Most of the change in slugging percentage can be explained by the massive .250 point increase in average against what used to be one of the most effective pitches against him. Because his underlying K rates haven’t changed that much, we can assume that it’s not just that he’s putting the sinker in play more, but that he’s driving it.
So we know he can hit the sinker now; what about other pitches? Below are his results on changeups.
He’s whiffing slightly less and still getting on base more often than not, driving the ball a significant amount. While we can expect some of the spike in BABIP to be a result of batted-ball luck (and thus regress in the coming months), some of that change has come from an increase in exit velocity and above-average performance against the pitch that most lefties attempt to put him away with. The lesson here is if I were a DFS player and I saw the Reds facing…I don’t know…a Jason Vargas-type pitcher, it might be worth coughing up the money to buy one of the more-overlooked assets in the Reds lineup.