clock menu more-arrow no yes

Filed under:

Fixing the Rockies, a data driven approach: Part 3, a new way to pitch

New, 19 comments

Invent Driveline Denver, experiment with splitters, and for Pete’s sake stop throwing fastballs down

Fixing the Rockies, a data driven approach: Part 1
Fixing the Rockies, a data driven approach: Part 2

Homegrown talent; or, why the Rockies should never spend money on pitching

When the Rockies were struggling in the early 2010s, we got plenty of high draft picks. And credit to the organization, those players have developed into one of the better starting rotations in baseball. Let’s look at the Rockies pitching staff as a whole and the strategy Jeff Bridich has applied. In short, it has been to develop starters and buy or trade for relievers.

From 2014 to 2016, the Rockies already had many of the key rotation pieces in the organization that would lead to its most recent postseason appearances: Jon Gray, Tyler Anderson, Kyle Freeland, Chad Bettis, Jeff Hoffman, and German Márquez.

Regarding the bullpen, Bridich’s approach has been to sign veteran relievers. Bridich has signed Greg Holland, Wade Davis, Jake McGee, Bryan Shaw, and Mike Dunn, and he’s traded for Pat Neshek and Seunghwan Oh. Some of those worked out in the short-term, but on the whole this strategy hasn’t worked.

The takeaways from Part 2 of this series about Coors Field suggest why there’s a discrepancy between the success of starts compared to relievers. Our pitchers have figured out how to pitch at Coors Field, mostly through experience. It may not be very focused learning, but if you throw enough pitches at altitude your brain and body will adapt over time. This works for the starters, but for a 10 year veteran their whole world gets turned upside down. Nothing moves like it did before and as a reliever that doesn’t train here in the offseason they don’t even throw enough pitches to adapt.

But that’s just a hypothesis that we have to test. For this I want to see if learning to pitch at Coors Field is better than coming here after an established career. As in the team home and away question, I am only trying to compare between the players themselves, not to see of the starters are overall better than the relievers. If this theory is true, then if we take all the veteran relievers and all of the homegrown starters we should see that they veterans have a larger home vs. away park adjusted split, just like the pitchers of opposing teams who also don’t have enough experience at altitude to pitch at altitude.

For this test I am also using 2017-2018 data. 2019 has seen all of the starters go down to injury so that data set isn’t exactly informative. To be considered a veteran reliever, they had to pitch a full season at Coors, not just a trade deadline rental. They also couldn’t be a veteran that used to pitch for the Rockies then came back later (the Rockies love doing this). I considered using more years to prove this, but until the signing of pitchers in 2017-2018, the Rockies hadn’t really done this before.

In addition to the starters named above, the following chart includes Antonio Senzatela and Tyler Chatwood as “homegrown starters” and excludes Oh and Neshek from “Veteran Relievers.”

Homegrown starters vs. veteran relievers

Group Home ERA+ Away ERA+ Difference
Group Home ERA+ Away ERA+ Difference
Homegrown Starters (2017-18) 109 109 0
Veteran Relievers (2017-18) 86 110 -24

Once again the effect of Coors Field reveals itself. Over 1725 total innings between these two groups, the veterans did 24% worse at home than on the road. In fact, the 16% difference between the road and home for the Rockies is almost all attributable to the relievers. Additionally, we see that the veteran pitchers are actually pretty good when they are on the road. They are 10% better than league average. We may give a lot of flack to Bridich for picking these guys, but based on the information he had on them before coming to Colorado it seemed like some solid pickups.

Part of this from our homegrown pitchers could be from pitching mostly at elevation in the minors. With the exception of Double-A Hartford, every full season minor league affiliate plays their home games at an elevation over 2,000 feet. Triple-A Albuquerque is at 5,312 feet. This may be a strange fix, but I’d advocate exchanging minor league teams with someone to get our double-A team closer to mile high elevation.

If we dig even deeper, we see that how early we get the starters into the high altitude system even makes a difference. Three of the starters were acquired via trade: Marquez, Chatwood, and Hoffman. The others all were drafted by the Rockies in the 1st or 2nd round. If we do the same analysis, we see the same issue comes up.

2017-2018

How acquired Home ERA+ Away ERA+ Difference
How acquired Home ERA+ Away ERA+ Difference
Trade 92 121 -19
Homegrown 116 103 13

Does this mean we should never trade for a pitcher? No. Chatwood was already a big league pitcher before coming to us and accounts for most of the difference (41% worse at home). He had no development in our system. Marquez and Hoffman were both Double-A pitchers when they arrived. Hoffman has had a much slower adjustment period but as he has gotten more experience, his gap has gotten smaller over time. In 2019, Hoffman was actually 6% better at home, although that isn’t included in the sample above. All this means is if we are going to get a pitcher from somewhere else, a Class A or Double-A pitcher will have a better pay off specifically for the Rockies.

The solution: the Rockies, as long as they remain in Colorado, should never sign any veteran pitcher that didn’t also come up in the Rockies organization.

From Bridich’s perspective, he saw a problem in the bullpen and he used money the Rockies had to go get that bullpen. He absolutely overpaid, because he had to. But as it turns out, getting any veteran reliever at any price fails.

Offseason work

As we’ve seen so far, simply spending more time pitching at altitude can make a pitcher roughly 20% better. And the data also suggests the more pitches thrown at altitude the better. For both the starters and relievers a regular season gives plenty of experience, but we should be able to cut that learning curve without sacrificing runs when every games counts.

During the offseason, players normally go home where they feel the most comfortable. Maybe they get some home cooking from mom and they workout with friends they may have played with in high school or players that live nearby. This has been the classic model for decades. But unlike every single team in baseball, going home and away from Denver means a completely different environment. Going from Florida to California really isn’t that different from a training perspective, but going from Denver to California is. I absolutely believe the Rockies should be trying to build up their data driven development, but I think they need to take it a step further: they need to do it in Denver. Given how important high altitude experience is for the pitching staff, I feel like all Rockies pitchers from Low A to the major leagues should spend time in Denver during the offseason.

The data driven development should be modeled on Driveline. For those who haven’t heard of it, Driveline is located in a warehouse in industrial Seattle, players from high school, college, the minors, and the majors all go to use data driven techniques to make improvements in their game. It started mostly with pitcher’s velocity increases, then moved into pitch design, and now has expanded into data driven hitting improvements. This is where Trevor Bauer was made and still trains in the offseason. They also have a great Podcast if you were looking for recommendations.

This is also where Adam Ottavino remade his pitching arsenal before purchasing the technology himself and training in a storefront in Brooklyn. After hearing about his experience and seeing the results, several Rockies pitchers took trips to Driveline this offseason, including Jon Gray, Bryan Shaw, Jake McGee, Jeff Hoffman, and Antonio Senzatela. Jon Gray has had a resurgence from 2018, where he couldn’t find his slider. Bryan Shaw and Jake McGee also have been significantly better since their trips.

Creating something like a “Driveline Denver” isn’t an entirely new idea. Driveline isn’t associated with any one team. But there are teams like the Astros whose players don’t go to Driveline. That’s because the Astros already employ similar data driven improvement techniques and keep them under wraps. They even have 75 Edgertronic cameras (high speed cameras originally designed for Aerospace applications) installed in their ballparks.

Regarding offseason work, I propose that

1) The Rockies go all in on data driven development, and
2) base it out of Denver, requiring pitchers to spend at least a few weeks of the 4-5 month offseason making changes at altitude.

This has a few more implications that come with it. Having a Denver based facility also helps concentrate the players from the majors and minors, which creates an organization that openly shares information. Everyone is there to do the same thing: improve. This is a culture that is created at these independent training facilities.

More splitters

While there isn’t any data to say for sure what does or doesn’t work for preventing runs compared to a control group, there is some evidence that one solution might be found in pitch selection. The Rockies should throw more splitters.

For a few years, the Rockies thought they had solved the Coors Field problem by getting sinker-ballers. The pitch naturally moves down which was thought to limit home runs and fly balls. It does (we’ll get into that next), but there was a problem. The pitch has the highest rates of contact of any other pitch, and at Coors field it’s actually a higher rate of contact.

Pitches move due to the Magnus Effect. While the ball spins it creates a pocket of turbulent flow which has less pressure than the other side of the ball. The ball moves in the direction of the turbulent flow due to the pressure differential. This happens due to the seams on the ball, their direction, and the rate of spin. We can calculate how a pitch is going to move based on these factors. Another key factor is air pressure. There is less air at Coors, therefore the pitches move less. This is one reason why it’s hard to pitch here. A fastball has backspin which pushes it up relative to how much gravity would pull it down normally. It still drops, but it looks like it rises because our brains already expect that. A curveball is the opposite. It’s spun very quickly with frontspin and drops way more than normal because the pressure different moves it the other direction. With that out of the way, back to sinkers.

A sinker is very similar to a fastball but doesn’t have as much rise and has some arm side run. Fastball rise less at Coors, but because the sinker already is designed to rise less, the thin air doesn’t work against it. This is good, we’ve eliminated one problem. Unfortunately, the arm side run is another factor of this pressure differential and it doesn’t move as much. This creates a problem.

The Rockies focused on the sinker because 1) it creates more groundballs and 2) Coors Field doesn’t work against its natural sink. But when you remove the run on the ball, they become easier to hit. Any reduction in movement makes the baseball easier to hit. This can be seen in the image below (purple are Coors Field pitches), from 2019. Sinkers are already the easiest ball to hit because they act very similar to fastballs, but you’re less likely to swing under them. Bringing them to Coors means the easiest pitch just gets easier.

Figure from Adam Maahs

This didn’t seem like a bad strategy. But we weren’t really looking at the pitch from a design perspective. Ideally we need a pitch that is hard to hit and isn’t affected by the Coors Field air as much or even better, a pitch whose movement is aided by less air. The study found that sliders move about the same amount at altitude. They move the least out of any pitch, and the amount of spin that contributes to their movement (known as active spin), is very low. In looking at these kinds of studies, I’ve seen an analysis of almost all types of pitches except one: the splitter.

The splitter (sometime called a forkball) is a pretty simple pitch. It is most commonly thrown just like a fastball, but with the fingers on the sides. They come in a little slower than a fastball, but have greater drop. By placing the fingers on the sides, that backspin is reduced. This reduction in backspin causes a smaller Magnus Effect and disappears out of the strike zone at the last second.

The same effect can be seen in a changeup, but it depends which kind. Often when guys throw a changeup it is a circle change. This has arm side run and is normally thrown to opposing handed hitters so that it darts away from them. Also seen in the image above, the changeup actually drops more at Coors Field (like we expect) but has a large reduction in arm side run. The ball drops due to low backspin but the sidespin component is reduced, once again making the pitch just as hittable. There is another pitch called a Vulcan change. This is almost exactly like a splitter but instead the ball goes between the middle and ring fingers.

The splitter isn’t designed to have as much run as a changeup, meaning its function seems to align with our goals. From a physics perspective this seems to be a great candidate for a pitch. In the thin air, the already small Magnus Effect will be reduced even more causing even more drop. And unlike a sinker, it is not the most hittable pitch in baseball.

The reason none of these studies include it is because not many guys throw it. Baseball Savant, which has a nice database of pitch arsenals, says only 46 guys in the MLB threw a splitter in 2019. We want to see if the splitter works at Coors Field. But not only do not many guys throw it, the guys who we can definitely say throw splitters are mostly in the American League. As a result we are going to need to hunt for the right guys.

Of the 46 guys that Statcast says throw a splitter, we are going to narrow it down to pitchers that actually faced the Rockies this season and throw a splitter more than 5% of the time. Then we need to narrow it down to pitchers that actually pitched at Coors Field. This gives us 13 pitchers that we can compare the movement profiles of this specific pitch at altitude compared to how they move normally.

This gives us 300+ splitters at Coors Field this season. Comparing their average movement and the movement seen at other ballparks we see that indeed, the splitter has a very similar movement profile.

Like the changeup, the splitter has a fair amount of drop due to less backspin. However it drops more than an changeup because the spin is even lower. It also has significantly less arm side run which means it isn’t as affected by the Coors Field air. The splitter has a very similar movement difference as a slider, but drops slightly more. Perhaps most importantly, it actually drops slightly more on average at Coors Field because the already small Magnus effect is reduced slightly. The difference isn’t as pronounced as the changeup shown above, but that’s because the changeup has a higher spin rate. The higher the spin in any direction, the more effect seen by the Colorado air. The splitter also is thrown harder on average. This gives a similar amount of movement with less time to react which is why it is a strikeout pitch and the pitch that propelled Tim Lincecum to three consecutive strikeout titles as well as two Cy Young awards.

Does this mean all Rockies pitchers should start throwing splitters? No. If a pitcher has a good breaking pitch they can control, stick with it. We don’t want to succumb to the Aaron Cook fallacy again. Almost all Rockies pitchers have a slider and they also have similar movement profiles at home and on the road.

All I’m suggesting is that if a pitcher is looking for a new breaking pitch, a splitter is simple to throw due to its similarity to a fastball grip and it plays well at Coors Field. Additionally, if a pitcher has a really ineffective pitch this is another option. When we look at our pitchers’ arsenals, we’ll see that there are several with really poor changeups. The splitter has 6” more drop than a changeup and doesn’t lose nearly as much run.

The Rockies had the right idea by focusing on sinkers. Unfortunately the pitch is notoriously hittable and relies on the arm side movement a lot to get strikeouts.

Throw the ball up!

Along with the sinker narrative came the idea that you have to throw the ball down at Coors Field. If you throw it down there will be more groundballs, less home runs, and less runs given up. Simple, right?

In fact, this is Bud Black’s tagline. He loves throwing that ball down. He tells this to every single pitcher that walks into the club house. If a pitcher did well, he was throwing the ball down. If he does poorly, he wasn’t throwing the ball down. I wouldn’t be surprised if every time a pitcher throws a ball up Bud Black makes them put a dollar in a jar on his desk. The thing is, he made all of us buy into it, too. The commentators talk about. Journalists talk about. Here’s an article that both praises throwing the ball down and getting a veteran starter with no Coors Field experience, all in one single data point. Let’s take a look.

The Rockies pitching staff has struggled this year, that much is obvious. If that’s the case, then according to Black, we must not be throwing the ball down. Looking into the numbers we see the Rockies lead the league in groundball percentage. Not only that, in the same sheet you can see we have the lowest fly ball percentage as well. Taking it one step further, if we take all 350,000+ fastballs thrown in the MLB this season and break it down by team, the Rockies ranked 2nd in terms of getting the ball down in 2019.

Fastball height, 2019

Team Avg fastball height (ft)
Team Avg fastball height (ft)
Boston Red Rox 2.843
Tampa Bay Rays 2.822
LA Dodgers 2.756
Miami Marlins 2.731
Houston Astros 2.683
LA Angels 2.667
St. Louis Cardinals 2.664
Cincinnati Reds 2.659
Chicago White Sox 2.651
Pittsburgh Pirates 2.649
New York Mets 2.646
New York Yankees 2.645
Cleveland Indians 2.64
San Diego Padres 2.627
Baltimore Orioles 2.626
San Francisco Giants 2.611
Philadelphia Phillies 2.607
Toronto Blue Jays 2.606
Minnesota Twins 2.602
Oakland Athletics 2.596
Milwaukee Brewers 2.594
Seattle Mariners 2.578
Chicago Cubs 2.562
Detroit Tigers 2.552
Kansas City Royals 2.547
Texas Rangers 2.546
Washington Nationals 2.536
Arizona Diamondbacks 2.523
Colorado Rockies 2.506
Atlanta Braves 2.48

Bud, I think it’s safe to say that you’ve done it — the Rockies have thrown the ball down.

But somehow we still rank 2nd in baseball for HR/9. The only team in front of us are the Orioles who set the record for most home runs allowed in a season this year (4 teams broke that record this year). Part of that is the ballpark yes, but how are we allowing so many home runs if we are getting the ball down? The answer is because fastballs down in the zone are hit hard.

Shown below is a heatmap of all fastballs hit hard (>95mph according to Statcast) as well as the average fastball locations for the best pitching staffs in baseball and us.

The hardest hit balls are mostly in the center of the strike zone and are skewed towards right-handed batters. This is because fastballs in are hit harder and a majority of people are right-handed, although left-handers are included in this sample. You’ll also see that the Rockies, at least height wise, are right in the middle of where the most hard hit balls are. Zooming in we can see it a little clearer.

You can see the best pitching staffs are throwing their fastballs up and trying to get away from the most dangerous part of the plate. As Purple Row has written, the Dodgers are intentional about keeping the ball up. While no teams in baseball (Rockies included) are throwing fastballs too far in, the best are throwing it up. You’ll also see that over hundreds of thousands of fastballs thrown, the difference between teams is really small, but those small differences over time make a huge difference.

The reason these balls are hit hard is because a fastball middle-down follows the natural bat path of a powerful swing. Add into that many hitters are starting to swing up on the ball and those fastballs down turn into home runs. It is true that fastballs down lead to more groundballs, but it’s high risk-high reward. If a batter misses the pitch, he’ll ground out. If he makes solid contact, he’ll lift it for a home run. It might have been true in Bud Black’s day that getting the ball down was a great strategy. At that time guys were instructed to swing down on the ball thinking that enough backspin could solve hitting in the wrong direction. By throwing down and hitting on top of the ball instead of under it it’s possible that home runs were reduced.

Now you might be saying, “But according to this graph, you should throw the ball down even more and then they stop being hit hard.” While this is true in a sense, it works against the natural movement of a fastball. As we just looked at, a 4-seam fastball has is a “rise” compared to what a ball would normally do without spin. You can try to throw fastballs down, but they always want to stay up. To be clear, there is nothing wrong with throwing the ball down, but it has to be the right pitch. Throwing a slider down works because it naturally breaks downwards. That location complements the pitch. Throwing a slider up is a disastrous idea because then it breaks down into the heart of the plate. The same thing happens when a fastball is thrown down: it wants to break to the heart of the plate.

The goal is simple: throw the ball where guys can’t do damage. For a fastball, that place is up. The ball naturally wants to go there and once it’s high enough, it’s very difficult to do damage with it. For a slider or a splitter, that place is down. The ball naturally breaks there, they are pitches with high swing and miss rates, and when hitters do make contact, the ball is already breaking down meaning they rarely hit the ball on the barrel for a home run.

This can even be seen in FanGraphs pitch value tables. This seeks to answer the question of “how well have hitters done against any pitcher’s specific pitch?” Of the 15 Rockies pitchers this season who threw 30 or more innings, 13 of them have a negative value attached to their fastballs. Every hitter is doing well against almost every single pitcher’s fastball on the Rockies.

Pitch value, 2019

PITCHER WFB/C WSL/C
PITCHER WFB/C WSL/C
Scott Oberg 1.51 2.07
German Marquez 0.39 -1.5
Chad Bettis -0.29
Chi Chi Gonzalez -0.4 0.1
Jario Diaz -0.43 0.49
Wade Davis -0.8
Carlos Estevez -0.81 1.5
Kyle Freeland -0.89 -2.15
Jon Gray -1 1.73
Antonio Senzatela -1.14 -0.12
Peter Lambert -1.25 -0.47
Jake McGee -1.64 0.72
Jeff Hoffman -2.47 1.37
Jesus Tinoco -2.8 1.57
Yency Almonte -2.96 2.76

Either the Rockies are wasting their talent by throwing the fastball down, or somehow the Rockies have coincidentally assembled the collection of players with the worst fastballs in all of baseball. I can assure you Jairo Diaz and Carlos Estevez, who both throw 100 mph, do not have negative fastballs. By comparison, 9 out of 13 have a positive value on their sliders. This is because when Bud Black tells them to throw a slider down, that’s where the pitch should be. When he tells them to throw the fastball down, that’s where the pitch gets hit.