March 01, 2009
Speed Zones
Yesterday I introduced a probabilistic model of the strike zone using PITCHf/x data. Today, I'd like to use that data to see how speed effects outcomes.
The model is built on three parameters, the X and Z coordinates of where the ball crosses the plate and the handedness of the batter. A positive probability outcome is calculated for each set of parameters, positive from the pitcher's point of view. So any pitch that does not result in a ball or a hit credited to the batter is a positive outcome.
For any set of pitches, the sum of the probabilities provides the expected number of positive outcomes. We can then compare that to the actual number of positive outcomes to see if the set of pitches was effective. Higher than expected positive outcomes indicate an effective set of pitches. Lower than expected positive outcomes indicate the opposite.
For this study, pitches are divided into sets by speed, rounded to the nearest mile per hour (MPH).
The following table shows the effectiveness of speed on pitches. The column definitions:
- Speed, in MPH. There must be 200 pitches at that speed to be included in the table.
- Pitches, total for that speed.
- Positives, the number of pitches with a positive result
- Pct. Pos., the percentage of the total pitches resulting in a positive outcome.
- Exp. Positives, the expected number of positives for those pitches based on the model.
- Exp. Pct., the percentage based on Exp. Positives.
- Ratio, 100*Positives/Exp. Positives. 100 meets expectations, over 100 exceeds expectations.
Effectiveness of Speed
| Speed | Pitches | Positives | Pct. Pos. | Exp. Positives | Exp. Pct. | Ratio |
| 100 | 253 | 156 | 61.7 | 144.6 | 57.2 | 107.88 |
| 99 | 818 | 481 | 58.8 | 480.7 | 58.8 | 100.07 |
| 98 | 2122 | 1290 | 60.8 | 1252.8 | 59.0 | 102.97 |
| 97 | 5093 | 3010 | 59.1 | 2991.1 | 58.7 | 100.63 |
| 96 | 9539 | 5600 | 58.7 | 5616.1 | 58.9 | 99.71 |
| 95 | 15424 | 8998 | 58.3 | 9119.8 | 59.1 | 98.66 |
| 94 | 21580 | 12583 | 58.3 | 12804.1 | 59.3 | 98.27 |
| 93 | 27943 | 16100 | 57.6 | 16487.0 | 59.0 | 97.65 |
| 92 | 32751 | 18917 | 57.8 | 19392.7 | 59.2 | 97.55 |
| 91 | 34993 | 20080 | 57.4 | 20611.1 | 58.9 | 97.42 |
| 90 | 34648 | 19665 | 56.8 | 20367.2 | 58.8 | 96.55 |
| 89 | 31283 | 17793 | 56.9 | 18331.3 | 58.6 | 97.06 |
| 88 | 27000 | 15122 | 56.0 | 15570.7 | 57.7 | 97.12 |
| 87 | 23559 | 13332 | 56.6 | 13400.0 | 56.9 | 99.49 |
| 86 | 21150 | 11890 | 56.2 | 11827.3 | 55.9 | 100.53 |
| 85 | 19903 | 11191 | 56.2 | 10857.8 | 54.6 | 103.07 |
| 84 | 19123 | 10818 | 56.6 | 10289.7 | 53.8 | 105.13 |
| 83 | 18613 | 10337 | 55.5 | 9843.8 | 52.9 | 105.01 |
| 82 | 17530 | 9924 | 56.6 | 9358.9 | 53.4 | 106.04 |
| 81 | 15812 | 8738 | 55.3 | 8399.6 | 53.1 | 104.03 |
| 80 | 13483 | 7453 | 55.3 | 7108.2 | 52.7 | 104.85 |
| 79 | 11136 | 6134 | 55.1 | 5823.2 | 52.3 | 105.34 |
| 78 | 9528 | 5177 | 54.3 | 5027.1 | 52.8 | 102.98 |
| 77 | 8239 | 4488 | 54.5 | 4318.8 | 52.4 | 103.92 |
| 76 | 6903 | 3649 | 52.9 | 3573.7 | 51.8 | 102.11 |
| 75 | 5719 | 3066 | 53.6 | 2996.0 | 52.4 | 102.34 |
| 74 | 4672 | 2419 | 51.8 | 2414.5 | 51.7 | 100.19 |
| 73 | 3687 | 1872 | 50.8 | 1926.1 | 52.2 | 97.19 |
| 72 | 2934 | 1580 | 53.9 | 1557.8 | 53.1 | 101.42 |
| 71 | 2379 | 1241 | 52.2 | 1253.0 | 52.7 | 99.04 |
| 70 | 1808 | 946 | 52.3 | 921.9 | 51.0 | 102.61 |
| 69 | 1401 | 700 | 50.0 | 712.0 | 50.8 | 98.31 |
| 68 | 1234 | 639 | 51.8 | 632.8 | 51.3 | 100.97 |
| 67 | 927 | 463 | 49.9 | 467.3 | 50.4 | 99.08 |
| 66 | 682 | 316 | 46.3 | 328.2 | 48.1 | 96.30 |
| 65 | 434 | 198 | 45.6 | 208.9 | 48.1 | 94.78 |
| 64 | 276 | 123 | 44.6 | 124.3 | 45.0 | 98.97 |
I did not expect this result. In general, I would have guessed that faster pitches would yield better results. This only appears to be true at very high speeds, however. Batters appears to find pitches thrown 97 MPH or better difficult to handle, but they have few problems with pitches in the low 90's. The most effective speeds for pitchers are between 75 and 85 MPH.
Note, however, that in terms of percentage, high speed pitches produce more good outcomes than low speed pitches. Low speed pitches result in a better than expected outcomes for a given location. Getting batters to swing at pitches in the dirt will do that.
This result seems to be in line with the idea that all major league batters can hit a fastball. How they hit the off-speed pitches, however, is what makes them a major league hitter. Twins fans should be very happy, given this data, that Francisco Liriano is bringing back his changeup.
Posted by David Pinto at
04:37 PM
|
Pitchers
|
TrackBack (0)
David - if this is true, will we see some team actually start looking for people who don't simply get a higher percentages of swing and misses?
Does this go beyond comparing Jamie Moyer to a Randy Johnson at his peak? Can you tell pitcher by pitcher whether they are more effective throwing within a specific speed zone - or better throwing more off-speed than hard stuff?
One change you might consider to make your model more sensitive would be to base your dependent value on the run value of each outcome. Strikes and balls would be valued based on the amount each increases/decreases the expected run value of a plate appearance on average (or even better, the specific value based on the actual count when pitch was thrown), while hits, BBs, Ks, and ball in play outs would get their respective linear weights values. That solves the problem of treating both a HR and an 0-2 waste pitch as equally "bad" outcomes, or a 3-0 strike and a GDP as equally good.
Guy,
Yes, I've thought of doing that, but right now I'm trying to keep it simple.
If I understand this right, you are taking speed and location into account, but you are not taking movement/spin into account, correct?
If so, that leaves out a fairly important variable.
Nevertheless, great stuff David. I was hoping you'd get into the Pitch f/x stuff at some point.
That is correct, Doc. I'm working up slowly.