This was inspired by one of the FanPosts of the week regarding David Price. Sticky Bandit's post looked into how Price's FIP changed along with varying usage of his different pitches, and it got me thinking along the lines of pitch sequencing and pitch usage in general. Specifically, I started to think about the discussions that seem to keep sparking up about James Shield's changeup, and how often he should be throwing it. One side of the argument states that since it is his best pitch, he should throw it a high percentage of the time. Meanwhile, the other side of the coin is if he shows the change too often, it may fail to surprise the hitter and become less effective.
Most likely, there is some middle ground, which should be the optimal pitch usage for his changeup. But what is that optimal percentage? Due to different matchups and batter strengths or weaknesses, saying his optimal changeup percentage (CH%) is 20% (for example) does not necessarily mean one out of every five pitches has to be a change. Rather, it just means that over the course of a start you would like to see Shields' change usage stabilize close to that mark. Some batters will see far more, some far less; that's just the way pitching works. It is far simpler to look at his aggregate usage as opposed to on a hitter-by-hitter basis, which - in addition to the small sample size problems with such data - is why I chose to look at the overall usage percentage for each of his starts. Here's what I found:
I chose to look into his start-by-start FIP in comparison to his CH% for each start, hopefully to find some sort of correlation, peak, or valley in the data to give some insight into his optimal CH%. Here's a game log for Shields with the relevant information (plus a bit extra), sorted by CH%.
|
Date |
Opp |
IP |
H |
HR |
R |
ER |
SO |
BB |
FB |
GB |
LD |
FIP |
CH% |
ERA |
|
11-Jun |
FLA |
3.33 |
9 |
1 |
10 |
10 |
4 |
3 |
8 |
3 |
3 |
7.40 |
33.8 |
27.03 |
|
28-Apr |
OAK |
7.00 |
6 |
1 |
2 |
1 |
12 |
1 |
4 |
6 |
6 |
2.06 |
30.7 |
1.29 |
|
15-May |
SEA |
8.00 |
6 |
1 |
2 |
2 |
10 |
0 |
7 |
7 |
5 |
2.33 |
29.0 |
2.25 |
|
9-May |
@OAK |
6.00 |
11 |
0 |
4 |
2 |
6 |
1 |
6 |
9 |
8 |
1.70 |
28.7 |
3.00 |
|
17-Jun |
@ATL |
6.00 |
5 |
1 |
3 |
3 |
3 |
0 |
9 |
8 |
2 |
4.37 |
28.4 |
4.50 |
|
22-Apr |
@CHW |
7.00 |
6 |
1 |
2 |
2 |
3 |
3 |
8 |
13 |
2 |
5.49 |
28.4 |
2.57 |
|
5-Jun |
@TEX |
7.00 |
10 |
1 |
6 |
3 |
4 |
1 |
3 |
15 |
8 |
4.34 |
28.3 |
3.86 |
|
1-Aug |
NYY |
7.33 |
4 |
0 |
0 |
0 |
11 |
1 |
4 |
10 |
2 |
0.61 |
26.7 |
0.00 |
|
25-May |
BOS |
8.00 |
4 |
0 |
2 |
2 |
5 |
2 |
10 |
8 |
5 |
2.70 |
25.2 |
2.25 |
|
29-Jun |
@BOS |
5.00 |
7 |
1 |
5 |
5 |
6 |
2 |
7 |
5 |
4 |
4.60 |
22.9 |
9.00 |
|
20-May |
@NYY |
7.33 |
8 |
1 |
4 |
3 |
7 |
1 |
5 |
12 |
3 |
3.47 |
22.2 |
3.68 |
|
30-May |
CHW |
5.33 |
11 |
2 |
7 |
7 |
3 |
1 |
11 |
7 |
4 |
7.52 |
21.5 |
11.82 |
|
26-Jul |
@NYY |
6.00 |
4 |
2 |
3 |
3 |
5 |
3 |
10 |
6 |
1 |
7.37 |
20.8 |
4.50 |
|
11-Apr |
NYY |
5.33 |
4 |
0 |
2 |
2 |
5 |
3 |
6 |
4 |
5 |
3.01 |
20.2 |
3.38 |
|
13-Aug |
BAL |
5.00 |
10 |
0 |
4 |
4 |
2 |
3 |
8 |
6 |
6 |
4.20 |
20.0 |
7.20 |
|
4-May |
@SEA |
8.00 |
8 |
0 |
2 |
2 |
10 |
0 |
3 |
11 |
7 |
0.70 |
18.2 |
2.25 |
|
9-Jul |
CLE |
6.33 |
6 |
2 |
4 |
4 |
9 |
1 |
7 |
5 |
4 |
4.94 |
17.9 |
5.69 |
|
17-Apr |
@BOS |
6.66 |
9 |
2 |
4 |
4 |
7 |
1 |
11 |
7 |
4 |
5.45 |
16.8 |
5.41 |
|
6-Apr |
BAL |
6.00 |
9 |
3 |
3 |
3 |
6 |
2 |
11 |
8 |
2 |
8.70 |
16.8 |
4.50 |
|
7-Aug |
@TOR |
4.00 |
9 |
6 |
8 |
8 |
2 |
4 |
12 |
3 |
4 |
24.70 |
16.0 |
18.00 |
|
23-Jun |
SDP |
7.00 |
6 |
2 |
4 |
4 |
7 |
1 |
9 |
9 |
2 |
5.34 |
15.6 |
5.14 |
|
4-Jul |
@MIN |
6.00 |
8 |
0 |
4 |
4 |
2 |
1 |
8 |
8 |
6 |
3.03 |
15.4 |
6.00 |
|
27-Jul |
DET |
6.66 |
9 |
0 |
2 |
2 |
7 |
1 |
8 |
8 |
5 |
1.55 |
15.0 |
2.70 |
|
21-Jul |
@BAL |
6.33 |
8 |
1 |
4 |
4 |
1 |
1 |
8 |
14 |
4 |
5.41 |
14.0 |
5.69 |
Take note of the starts on Jun 11th and Aug 7th, because for the rest of the charts I created, I chose to consider those starts as outliers and remove them from the graphs. Those starts would have greatly skewed any potential trends or correlations (as this is only 150 innings of data), and I feel fairly confident in saying that those starts do not accurately represent Shields' true talent level. Also, for the sake of argument his batted ball profile on those occasions is not significantly different from other starts in which he was much more successful, indicating some bad luck.
First, I'm going to show the chart of his FIP for each start compared to his changeup usage in each start.
Certainly a lot of ups and downs, most likely due to how volatile FIP can be from start to start. However, it seems to me that there is a slight upward trend in performance (FIP) as Shields uses his change less. To confirm/look at this a little bit closer, I plotted a trend line on the FIP part of this graph:
What strikes me here is not so much the volatility of the graph, but the fact that the trend line climbs all the way from a 3.00 FIP to a 5.00 FIP as Shields' changeup usage drops from over 30% to around 15%. Certainly, he has had bad starts and great starts on both ends of the spectrum, but I think it's reasonable to conclude that Shields is more likely to have a solid performance when he is throwing his changeup more often.
If I had to pick a target percentage from the data above, I would go with 25%; excluding the starts against Florida and Toronto, he has only posted a FIP above five once when throwing at least 25% changeups, while doing so six other times when he throws the change less than 25% of the time.
This is all a relatively small sample, and the pitch sequencing Shields uses is probably just as important as how often he throws his pitches. Unfortunately, that really is not something we can quantify right now, and I feel like this is a decent way to look at how pitchers should blend their pitch repertoire. It would be worthwhile to extend this exercise for at least the past couple years to make the results more reliable, but I believe that there is still something interesting to look at here. For the season Shields has used his change 23.4% of the time, which is reasonably close to my target of 25%, but not as high as I would like to see. I would love to see him throw a bunch more changes down the stretch, and put my little theory to the test.
*A couple notes:
-Stats courtesy of Texasleaguers.com and Fangraphs
-I also graphed ERA against his CH%, but since it was virtually identical to the FIP graph I chose to omit it. If anyone's interested in seeing that addition, let me know.
-I originally intended to look into his Whiff% on the change along with FIP, but that fell by the wayside during my war against Excel. Maybe next time...