Pitch Speed/ Pitch Type

by grobbins on Sep 14, 2007 8:26 AM CDT Comment 3 comments

So as I just posted on the PITCH f/x diary, the algorithm to me seems to have a flaw, but that if we just know every pitcher's true top velocity of their fastball, then we can predict the other pitches just by their speed. Over the course of the season, if you call the speed of the fastball X then the other pitches would be:

Cutter = 0.95 X
Splitter = 0.92 X
Slider = 0.92 X
Changeup = 0.90 X
Curveball = 0.82 X

This means that we can predictably order pitches with the exception of a splitter and a slider.

This is based on data that a friend of mine uncovered over 100's of games of analysis, so while I am sure their are flaws....the sample size is pretty good. On a side note, I showed him the type of discussions that occur hear and he was pleasantly surprised by the knowledge of the poster's here.

0 recs | Comment 3 comments

Story-email Email Printer Print

Comments

That is good to hear

I absolutely agree with what you are saying about using the pitch speed to define pitches but the problem comes in with some pitchers like Greg Maddux. He can and will throw his four seam fastball at a large range of speeds. Probably anywhere from .9x to 1x using your notation. Note he is gripping the .9x pitch the same as the 1x pitch and the spin (therefor break) of the ball looks identical. All of these pitches still are four seam fastballs and need to be classified as such. Maddux will also throw a change that will tend to be .9x but with a larger spread than you see with other pitchers as well. You can tell the difference between the two pitches only by looking at the break of the pitch with him, IMO.

The algorithm that I am using does almost exactly what you describe though. It finds the pitcher's fastball and then works everything off that fastball in both speed and break. It then lumps clusters of pitches together and then compares those clusters to clusters from other pitchers to come up with the type of pitch it is. It is far from perfect but as I will be posting in a new diary soon, it is getting better.

Visit my baseball blog ...or else!

by dixieflatline on Sep 14, 2007 9:15 AM CDT reply actions 0 recs

so your DB is built then?

haven't found time to do this myself. :(

Did you use perl and MySQL to put it together? Any tips for someone who might try it soon?

Bring Back The Old Logo!

by jacob on Sep 14, 2007 9:24 AM CDT up reply actions 0 recs

I know nothing about DBs

I use C++ and a statistical package that is basically a C++ add on. That is why my plots look different from all the other excel/R plot you see out there.

But yes my data is basically built. I have stopped adding to it the last couple of days so I can debug the algorithm and such but I can pick up and add more data any time I want.

Visit my baseball blog ...or else!

by dixieflatline on Sep 14, 2007 10:33 AM CDT up reply actions 0 recs

Comments For This Post Are Closed

Search

FanPosts

Community blog posts and discussion.

Recent FanPosts

+ New FanPost All FanPosts >

NL Central Standings

	W	L	PCT	GB	STRK
St. Louis Viva El Birdos	91	71	.561	0	Lost 6
Chicago Bleed Cubbie Blue	83	78	.515	7.5	Lost 1
Milwaukee Brew Crew Ball	80	82	.493	11	Won 3
Cincinnati Red Reporter	78	84	.481	13	Won 2
Houston The Crawfish Boxes	74	88	.456	17	Lost 3
Pittsburgh Bucs Dugout	62	99	.385	28.5	Lost 2