Is Big Game’s Game Breaking Down?

James Shields was traded this off season from the Rays to the Royals. He has been known for his durability over the years. Spanning the last two seasons, he is first in complete games with 14. Also, he is second to Justin Verlander in innings thrown. The durability and consistency he is known for may be coming to an end. At the end of the last season, he showed signs of breaking down because he was not able to throw strikes and wasn’t able to maintain a consistent release point.

Before, I get into the meat of the analysis, I need to back up a bit. After several years of wishing to recreate Josh Kalk’s injury zone work (reproduced by Kyle Boddy), I finally have a publicly available model which looks for three pitcher injury traits(link): velocity loss, low Zone% and late game inconsistencies (more detail in the Appendix). By using all three traits, a better picture of a pitcher’s health can be revealed.

Just by looking at James Shield’s velocity, which is the only trait most people use to tell if a pitcher is injured, he looked healthy. Here are his two and four-finger fastball speeds from the last 4 years:

Shields fastball was faster than it was in any previous season. All seemed great.

The first sign something was wrong, was a significant drop in his Zone%. A low Zone% indicates the pitcher is having problems throwing strikes. Pitchers who have problems finding the strike zone are more injury prone. It is not known what comes first yet, an inconsistent release leads to an injury or an injury leads to an inconsistent release, but the two seem to go hand in hand. Here are his Zone%’s since entering the league (Pitchf/x values):

Season: Zone%
2007: 52%
2008: 51%
2009: 53%
2010: 51%
2011: 51%
2012: 45%

Besides the yearly values, here is a graph of his game-by-game values from the last 3 seasons with a 5 game moving average (large image).

He had a Zone% over 50% in his first 5 seasons and then in 2012 it dropped like a rock. The lack of pitches in the strike zone didn’t have an impact on his BB% which was down from 2011 (6.7% to 6.1%). The lack of throwing strikes, really hurt him in the number of innings he was able to throw. From 2011 to 2012, the number of pitches he threw went from 3576 to 3617. In the mean time his innings pitched went from 249 down to 227 while his pitchers per batter increases from 3.67 to 3.83.

The second factor, which points to Shields being hurt, was his inability to maintain a consistent late game delivery. Basically, was he not able to keep the same delivery as he wore down. Here are his game-by-game late game consistency scores over the past 3 years (a score of 100 is very inconsistent and 0 is very consistent – large image):

In 2010 and 2011, his 5 game average peaked around 40 with an average value at 29.4. In 2012, the 5 game average was over 50 through May and June with an average yearly value of 45.1.

Besides having a higher average inconsistency score, he was seeing more and more scores at or above 50. In the games where his score was less than 48, he had an 3.37 ERA (3.75 R/9), 23% K%, 5% BB%. In the games where the value was at or above 48, his ERA was 3.77 ERA (4.61 R/9), 24% K% and 8% BB%. When consistent, his results were better.

To find the cause of the inconsistencies, a little more work needs to be done. The tool determines if the player was throwing inconsistently (speed, break and release points) late in the game, but it doesn’t say in what way. The data needs to be looked at in more detail.

After looking at the data, Shields was just not able to maintain a constant release point. To show this problem, I am went back to look at his last game of the season. Normally, when a pitcher is hurt, they vary their arm height and therefore their horizontal release point as seen by Kyle Drabek’s release point two games before he had Tommy John Surgery (his release points are at an angle).

The inconsistency with Shields was only in the horizontal direction as seen here.

Going back to the game, I looked at fastballs from the different horizontal locations. Here are four .gifs from when he faced Chris Davis and Adam Jonesin the 7th inning on 10/2/2012. Try to find the differences:

The difference is his foot placement on the rubber. In the 1st and 3rd image, his drive foot is noticeably closer to 1B. In each case he threw to the 1B side. In the other two images, his foot was in the middle of the rubber and he threw to the 3B side. Pitchers should not need to move on the rubber to throw to different sides of the plate, but Shields was needing to move.

Early in the game, he threw mostly from the middle of rubber, but as the game wore on he needed to position himself on the 1B side of the rubber to throw to the 1B side of the plate. Here is a graph showing how he eventually needed to move around to the inside and outside. The graph is his horizontal release point vs the horizontal point the ball crossed home plate (positive values mean the 1B side and negative numbers are the 3B side).

The big question which needs to be asked is why was he repositioning himself on the rubber. Was he doing it unknowingly and was tipping his pitch locations? Did it hurt him to throw to both sides of the plate with the same motion? Was the inconsistent release points a cause of the low Zone% over the course of his season? Hopefully James and the Royals will be able to figure out differences. We will be able to find out soon because the Royals spring training facility in Surprise has a Pitchf/x system installed.

The old method of just using fastball speed to see if a pitcher is injured is limited. When looking at James Shields, his fastball velocity indicated he was healthy. By additionally looking at his 2012 Zone% and Late Game Consistency, he showed signs he was inconsistent in his release and couldn’t get pitches over the plate. His problems could be the indication of an injury or the possible causes of one. He was signed by the Royals to stabilize their pitching staff, but missed time may be in store for Big Game James.

.
Huge kudo’s to my brother Darrell for setting of the web page.
.
.
.

Appendix

Information on the pitcher injury finder application.

• It is slow, almost painfully slow at times. With everyone possibly using it right after this article is released, it may be even slower. Right now, it takes over 1 minute to process one pitcher for just one year. If it is too slow over the next day or so, try it again when less people may be using it.

• Predicting possible injures is an imprecise science. Two pitchers could have almost identical values, but one may need Tommy John surgery and the other one wouldn’t. All I have done is make data available on traits which have been previously known to lead to injuries.

• Directions: Select a starting pitcher (tool only really works with starters because a minimum number of pitches needs to be thrown) and date range. Press submit. Next will come up a most common pitches chart for the pitcher. Pick the most common fastball (to use FA if going back to 2009). Press Submit and have a drink or take a nap because this may take a while. The results will eventually appear. 

• The first graph is pretty simple. An average velocity graph for the selected pitch with a 5 game average curve.

• The second graph measures late game consistency. A 100 value is an inconsistent pitcher and 0 value is a consistent pitcher. I got the values by looking at pitchers with major arm issues in a season and pitchers without arm issues in a season. Then, I compared one group of pitchers to the others over the last 10 fastballs thrown in each game. Velocity, release points and break were examined using logistic regression. In the end, I got a formula which detects inconsistencies. The exact cause of the inconsistent is not outputted. The user will need to go look at the game data to find the pitcher’s exact issue. Note: Pitchers, like Bruce Chen, who have two distinct release points, will have all their values near 100. I have not been able to work out this problem yet.

• The final graph is the pitcher’s Zone%. A value under 47% (baseline value on graph) means the pitcher had issues throwing strikes and is more likely to be injured.

• In the near future, I will be going back and looking at how the values can be applied to other pitchers and situations.