What is More Important for a Fastball: Velocity, Location, or Movement?
Velocity, location, and movement are all unquestionably important when we try and compare ‘good’ pitches to ‘bad’ pitches. My particular interest lies in how important each are. I’ve often wondered if a 98 mph cutting fastball can be thrown right down the middle and still have little chance of being hit for a home run? Or conversely, is an 88 mph fastball that’s straight as an arrow still likely to get a swinging strike if it paints the bottom outside corner of the zone?
The approach I decided to take to try and begin answering the question of relative importance between velocity, location, and movement, was to look at the pitchFX data for all pitches thrown during the 2011 MLB season. As a start, I decided to look only at fastballs thrown by right handed pitchers to right handed batters in an attempt to get as homogenous a data set as possible. I included all pitches from the 2011 season that MLB gameday classified as either: fastball, four-seam fastball, two-seam fastball, cut fastball, or split finger fastball. I then narrowed the data set to include only pitches that were either swinging strikes (good) or hit for a home run (bad). I decided to use swinging strikes and home run as the good/bad pitch criteria because ultimately those are the best and worst possible outcomes for any given pitch.
For righty vs. righty matchups in 2011, I found that 1,134 fastballs were hit for home runs, and 11,790 were swinging strikes. This means that approximately 10x more fastballs were swung on and missed than hit for home runs.
Next, I wanted to get a general sense of the location of the pitches for each outcome. I created two separate plots of pitch location, one for swinging strikes (SS) and one for home runs (HR). Figure 1 shows those plots. From the plots I could see that batters swing and miss at fastballs pitched in pretty much any location, so that wasn’t helpful. But, I also noticed that fastballs down and away were not often hit for home runs, and fastballs middle in look more likely to be hit for home runs than fastballs away. Although not particularly useful in answering my question, since these two figures agreed very well with conventional baseball wisdom, I was able to proceed with a more thorough investigation knowing at last my data set was good.
Figure 1 – Location of fastballs pitched by right handed pitchers to right handed batters during the 2011 MLB Season. Red dots represent pitches swung on and missed, blue dots represent pitches hit for a home run. All units are in feet. The rectangle between -0.83 to 0.83 in the horizontal and 1.6 to 3.5 in the vertical represents an approximate strike zone. Figures are from the catchers perspective, therefore negative horizontal numbers represent inside on a right handed batter.
The next thing I decided to do was run a series of independent t-tests to determine if there were any statistically significant differences in the average velocity, location, or movement of the HR group compared to the SS group. Table 1 shows the results of these tests.
Table 1 – Summary for independent t-tests for velocity, horizontal location (px), vertical location (pz), horizontal movement (pfx_x), and vertical movement (pfx_z)
HR Avg |
HR Stdev |
SS Avg |
SS Stdev |
p |
|
Velocity |
90.85 |
2.94 |
91.27 |
3.59 |
<0.05 |
px (ft) |
-0.21 |
0.46 |
0.06 |
0.71 |
<0.05 |
pz (ft) |
2.66 |
0.49 |
2.63 |
0.83 |
0.228 |
pfx_x (inch) |
-4.49 |
3.46 |
-3.99 |
3.85 |
<0.05 |
pfx_z (inch) |
7.76 |
3.00 |
7.08 |
3.78 |
<0.05 |
I found the average velocity difference between HR and SS was statistically significant. However, the difference between the averages was less than half a mile per hour, and the range of velocities for each group (as indicated by the standard deviations) were quite high. What I interpreted from this was, most MLB fastballs are probably between 88 to 94 mph, and in this range, velocity probably doesn’t matter that much.
In terms of location, in the horizontal direction I found the average location difference between HR and SS was statistically significant. Practically, I’m not sure how meaningful the difference really is. The average SS fastball was pretty much right down the middle, and the average HR fastball was only .21 ft (approx. 2.5 inch) from the middle on the inside half. In other words, it looks like hitters swing and miss at pitches everywhere, and hit home runs on pitches near the middle of the plate or middle in. No big surprises here. In the vertical direction, average location was not significantly different between groups. This may be a little bit misleading, because even though the distribution is normal for both HR and SS, the distribution is much wider for SS compared to HR. More on this in a bit.
Statistically, movement in both the horizontal and vertical direction were significantly different. Pitches that are hit for a home run appear to run in on right-handed hitters about half an inch more than pitches swung on and missed. Again, the range in movement was quite high within each group that, practically speaking, I’m not so sure how much that difference in average matters. A similar trend was found with vertical movement in that pitches swung on and missed moved downward about ¾ of an inch more than those hit for home runs. Although this was again statistically significant, I doubt it is practically significant because of the range in movement.
SS to HR ratios
After I decided that the independent t-tests didn’t really help me answer my original question of relative importance between velocity, location, and movement, I decided to take a completely different approach. Instead of looking at the differences in the averages between SS and HR groups, I instead looked within velocity, location, and movement to see what changes within each affected the outcome of each pitch.
I started with velocity. I broke down both the SS and HR groups into 1 mph increments. I then determined the ratio of pitches at each velocity that were swinging strikes compared to home runs. This ratio effectively tells me how much more likely a pitch is to be swung on and missed rather than hit for a home run. Remember, in general a fastball is 10x more likely to be swung on and missed than hit for a home run. A higher SS:HR ratio is better for the pitcher. What I found was that fastballs between 87 to 92 mph were actually less than 10x more likely to be swinging strikes. More interestingly, I found that fastballs 96 mph were more than 20x more likely to be swinging strikes. Ninety-eight mph fastballs are 33.5 times more likely to be swinging strikes. So looking at Figure 2, it appears as though fastballs over 95 mph are considerably less likely to be hit for home runs. Also somewhat interestingly, fastballs less than 86 mph are also less likely to be hit for home runs.
Figure 2 – Swinging strike (SS) to Home run (HR) rate broken down by velocity
As for location, Figure 3 shows the SS:HR ratio for both the horizontal and vertical directions. In the horizontal direction, it looks like most pitches in the strike zone are less than 10x more likely to be swinging strikes than home runs. The only exception is the outer six inches. Pitches about 3-4 inches from the outside edge of the plate are nearly 20x more likely to be swinging strikes, and pitches that paint the outside edge are over 40x more likely. On the other hand, if a right handed pitcher paints the inside part of the plate on a right handed batter with a fastball, he’s only 7x more likely to get a swinging strike than a home run (less than average). In the vertical direction, it looks like both high strikes and low strikes are better at getting swinging strikes than stuff right down the middle. At 1.75 ft from the ground, or typically just above the knees, the pitcher is 35x more likely to get a swinging strike. Interestingly, at 3.5 ft from the ground, or typically about the top of the strike zone, the pitcher is only about 13x more likely to get a swinging strike. This isn’t too much above the 10x average. Only when the pitch gets 3.75 ft above the ground, or out of the strike zone, does the likelihood change to 25x.
Figure 3 – Swinging strike (SS) to Home run (HR) rate broken down by horizontal and vertical location. For the horizontal direction, zero indicates the middle of home plate and negative numbers are inside to a right-handed hitter. For the vertical direction, zero is the ground.
Finally, the most interesting findings to me come from movement (Figure 4). I was very surprised to see how little horizontal movement matters with respect to SS:HR ratio. From the t-tests I found that pitches hit for home runs had slightly more horizontal movement inward than pitches swung on and missed. This finding was again seen here. From the graph, it looks like inward movement causes the SS:HR ratio to remain around 10x. However, both a straight fastball and a slightly tailing fastball have a slightly higher rate of 15x. In the vertical direction, again movement does not seem to matter for the most part. Upward movement greater than 4 inches has a consistent ratio of approximately 10x. Only when upward movement is 4 inches or less does the SS:HR ratio considerably increase. At 2 inches upward movement, the SS:HR ratio is 26x.
Figure 4 – Swinging strike (SS) to Home run (HR) rate broken down by horizontal and vertical movement. For the horizontal direction, negative numbers are inside to a right-handed hitter.
Side note: The vertical movement value given by the pitchFX system can be slightly counter-intuitive. PitchFX defines movement with respect to a pitch thrown with no spin and acted upon by gravity. Gravity pulls all pitches downward, however, spin on the ball can create a small amount of lift. Therefore, most fastballs classified by the system have upward movement. The average upward movement for an MLB fastball is between 7-10 inches.
To summarize, it appears as though velocity and location matter far more than movement when it comes to fastballs. Fastballs over 95 mph are far less likely to be hit for home runs than swung on and missed (over 20x), however, fastballs between 87 to 92 mph were no different with respects to how likely they were to be hit for home runs compared to swung on for strikes (all about 10x). Fastballs on the outer edge of the plate were over 40x more likely to be swung on and missed, and fastballs down in the zone 35x. Fastballs up in the zone were not much different than fastballs in the middle of zone. For a high fastball to be effective, it must be out of the strike zone. Tailing fastballs seem to be better than cutters.
Finally, I think the take home message here is, if a pitcher is gifted enough to have a 95+ mph fastball, it probably doesn’t matter too much where they throw it. As long as they can throw it for strikes, they will be effective. Pitchers that don’t have that 95+ mph fastball can also be effective, however, they need to locate their pitches. There are still places in the strike zone they can throw their fastball and be very effective. The idea that movement on a fastball matters appears to be totally erroneous. A pitcher probably can’t live on a fastball that is 92 mph with a lot of movement if he can’t locate it. A pitcher probably can live on an 88 mph fastball with little movement as long as he can locate it. Location is far more important than movement for fastballs.
Good effort here – thanks for sharing your work.
A few observations:
(1) The first part of your analysis in which you tried to directly compare relative importance may have failed because of your use of averages when looking at location. Using averages, inside and outside locations cancel each other out, leaving you with no useful information.
(2) You may want to broaden your analysis by including called strikes too. From an outcome/run creation standpoint, there’s no difference between a swinging and a called strike, and there are some pitches that we think of whose intent is to produce a called strike — against a right hander, for example, some pitchers throw a cutter meant to start outside the strike zone but breaking in towards the batter into the zone.
This is awesome, and I want to see things this good every day at Fangraphs. Would it be possible to look at hit balls classified as fly’s or line drives vs. grounders?
I’m not sure that you can make a very strong conclusion when only taking swinging strikes and homeruns into account. It’s inherently biased towards strikeout-type pitcher’s fastballs, as not every pitcher’s fastball is intended to produce a swinging strike. It’s still a good start, though.
I think both called strikes and line drives should be taken into consideration. I would guess that called strikes might be more likely to favor location than swinging strikes, and line drive rates might favor movement.
Thanks for the comments thus far.
I chose the outcome measure of SS:HR for this study because I was really interested in what makes a pitch difficult to hit. I wanted to focus only on pitches were the batter had decided he wanted to try and hit the pitch. Swinging strikes represented the pitches most difficult to hit, and home runs represented the easiest pitches to hit. I would have liked to include fly’s, line drives, and grounders but I felt since those classifications are based on the opinion of the mlbgameday stingers, they were too subjective. I couldn’t be confident that the additional value they may have provided would be outweighed by the additional noise they may have introduced.
Thomas, you began with the question, what makes a pitch good or bad? That is a different question than, what makes a pitcher harder or easier to hot once a batter has committed to a swing, and I would suggest to you your original question is the better one.
The reason your original question is better is because in most circumstances a batter does not decide whether to swing until the pitch is thrown. And if a pitch is thrown in such a manner that it tricks the batter into taking what he thinks is a ball or taking a pitch simply because it’s location or type is so unexpected, than such a pitch, assuming it is a called strike, is a key component of what makes a pitch good or bad.
In fact, if we assume an umpires ball-strike calls are less variable than the outcome of swings — a safe assumption — than a thrown strike that is not swung at is a superior pitch to any pitch that is swung at.
And called strikes don’t introduce any noise into the data, at least not more than what would already exist.
I think what you have done is both interesting and helpful. I also think your work would benefit from including call strikes too and seeing what if any difference that has on the quality of a pitch.
rotofan, I’ve done an additional analysis including called strikes and posted the graphs in a google doc here:
https://docs.google.com/document/d/1qBHWzUGlOSUPYQN-I_S9k3474iJj2O6sSR9LOpBFViI/edit
You can see for the most part the trends remain the same.
Just to give some context for the graphs, 32,225 righty vs. righty fastballs were called strikes in 2011. Combine that with the 11,790 swinging strikes and 1,134 home runs, this yields a [called strike + swinging strike]:home run ratio of approx. 40x (38.8 to be exact).
Excellent work, Thomas — thanks for the all your work!
A few observations:
(1) Most interesting change to me was with horizontal movement showing that balls the broke the most towards the batter were the most effective at inducing called strikes, more-so than balls that broke away or broke hardly at all. That’s the opposite of what you found with swinging strikes.
(2) Extreme high velocity matters less with called strikes than with swinging strikes — not a big surprise.
(3) The range of pitches low in the zone to induce called strikes is bigger than to induce swinging strikes.
Keep up the great work!
rotofan, I agree with your first two observations. I’m not really sure I follow what you mean by ‘range of pitches low in the zone’ in your final observation.
Regardless, thanks for all the positive and constructive feedback.
Great analysis. I’m a Rockies fan so I’m always focused on what might work at Coors. Obviously the team itself hasn’t even figured that out – but this reinforces one hypothesis I’ve come up with.
A high heat FB pitcher only really needs one plus fastball pitch to succeed here (see Ubaldo) – looks like 95 mph is that threshhold (maybe 93-94 because Coors itself adds 1-2mph)
A regular FB pitcher is going to need two different FB pitches to succeed here. One as their core pitch – and a different one as their safety pitch for when they get behind in the count – or mixnmatch if they have good enough command of both. I suspect that all of these graphs shift here. And unless a pitcher has elite command (none of them are ever going to want to pitch here since elite is elite), their core-safety FB’s are going to have to work in combination – by creating a bit of deception due to slight differences in velocity and location where both can be thrown. With only one FB in that situation, it’s too easy for batters to key in on it as they see more of them.
jfree, I think your hypothesis might hold true for more ballparks than just coors field.
Elite fastball command or elite velocity on their own are probably good enough for a pitcher to be successful anywhere. With respects to a pitcher that doesn’t have either elite command or velocity, the idea of mixing their fastballs to be successful is something I plan on looking at in the near future. Perhaps variability in fastball movement has a stronger association with successful outcomes than just looking at movement alone.
thanks for the work. it’s very interesting. i have basically the same suggestions as other people (called strikes mostly), except that i would say the element of surprise is the most important element to me. you can throw a 99 mph fastball in the perfect spot, but if a hitter knows it’s coming, it’s going to get hit.
Am I misunderstanding or do you mean cutting when you say tailing. Cutters move away from a righty like a curve would, tailing fastballs tail into a righty. If “Figures are from the catchers perspective, therefore negative horizontal numbers represent inside on a right handed batter”, then isn’t negative movement the tailing fastball and positive movement a cutting fastball?
Therefore the analysis would conclude cutters are lightly better than tailing fastballs. I may just be misunderstanding it, but wanted clarification. Great analysis.
You’re absolutely right McAnderson. The last sentence of the second to last paragraph should read, “Cutting fastballs seem to be slightly better than tailing fastballs.”
I have a humble suggestion to add to your very interesting study. While called balls seems to not be an equally bad thing as a home run, they would be an interesting opposite to called strikes, since neither are really all that good or bad.
And to further that, it seems that called balls and strikes are both not all that extremely good or bad, unless they are called THIRD strikes or called ball FOUR. These pitches do lead to a good or bad outcome. So is there a way to get Swinging strikes with called third strikes VS. Homeruns with called ball fours?
A further thought I have may be resolved by the previous request. In short, a called strike on a 3-0 count shouldn’t be considered good, since most hitters will be asked to not swing. This is more “not bad” than good. In the same way, a called ball on a 0-2 count isn’t a bad pitch. Contrariwise, it’s a good time to see if the hitter will chase something outside the strike zone. If seems like the next level to seeing if a pitch is good or bad is determining the situation of that individual pitch.
Very interesting work, and enjoyed reading through all the thoughtful comments. As a former pitcher and student of the game, I think it’s safe to say that the effectiveness of any pitch is multi-factoral. Having lights out stuff is a huge plus, but in addition to velocity, location and movement, I think some other, perhaps equally important factors need to be considered. Examples: the count, the previous pitch (pitch type, location, outcome), game situation (score, men on base, etc). All of these things factor in. For instance, lets say a pitcher’s avg fastball is 95 mph. But his team is up by 7 runs late in the game and he just threw a 95 mph fastball for a ball to make the count 3-1. If his next pitch is a get-me-over fastball at 92 mph and it gets hit for a HR, was it just the velocity and location that factored in? Would that same 92 mph fastball in an 0-2 count from a pitcher who averaged 88 be just as likely to get hit out? Just some thoughts, again good stuff, and love seeing all of this available data being put to use. Keep up the good work.
Phil R,
Thank you for the positive feedback. I agree with most of your comments. There is more to pitching than just velocity, location, and movement. I think a pitcher’s effectiveness can be broken down into two categories: his ability and his makeup. The sequence of his pitches, how he deals with the game situations he finds himself in, and the pitching decisions he makes, all result from his makeup. For my article, I was not looking at any of these factors. Instead, I was looking at a pitcher’s raw ability. I think ability can be fairly well described in terms of velocity, location, and movement. I wrote this article to try and shed some light on how ability is currently evaluated. Often we hear about a young pitcher’s ability in terms of how good his ‘stuff’ is. Presumably, this is in reference to both velocity and movement. I wanted to show that movement is not nearly as important as some would have you believe, and when evaluating young pitching ability, we should place much more emphasis his ability to locate rather than his ‘stuff’. Again, thanks for the positive comment and I hope you continue reading my work.