What Fastball Velocity Value is Most Predictive?

Yesterday, I published an article on a few pitchers whose fastball velocity changed over the course of last season. And then my old buddy MGL showed up.

He’s right. I have so much on my plate right now, mainly my first book and a 2019 player previews, that I didn’t take it another step forward. Here are most of his answers.

First of all, I should have linked Mike Fast’s article from about 10 years ago when he first examined pitch velocity in which I re-examined a few years later. The article has held up fairly well over the years expect maybe this quote:

From both studies, a 1 mph loss in velocity means a drop in (E)RA of .28 for starters and ~.50 for relievers. The strikeout rate drops 1% point for starters and 2% points for relievers.

For the second part of MGL’s question, he wondered on the predictive value. I ran a simple test on starters who threw at least 50 fastballs in two consecutive seasons. I used this subset as I was just trying to make sure the pitcher was a starter in each season. Moving from starter to reliever usually adds a least 1 mph to a pitcher’s average fastball velocity and I didn’t want this variable added in.

From this data, I compared the pitcher’s velocity full season, first half, second half, and September velocities in season 1 to the next full season’s velocity.

First off, a pitcher’s velocity holds up no matter which value is used.

Season 1 variable to Season 2: r-squared
Full Season: .87
First Half: .86
Second Half: .87
September: .86
80% 2H & 20% Sept: .88

It’s best to weigh the late-season values more but the difference isn’t that much.

But how about pitchers who experience the biggest drops.

I grouped all the starters who saw the velocity change by 0.75 mph (~ 1 SD above the median change) from the first half to the second half and here the correlations.

Season 1 variable to Season 2: r-squared
Full Season: .88
First Half: .87
Second Half: .89
September: .87
40% 1H & 60% 2H: .90

Second half velocity seems to be the best individual indicator as its individual value is the highest and is always part of the combined variables.

Instead of a change from the first half to the second half, here are the biggest changes from the first half to September. The value was a 1 mph changed (~ 1 SD above the median change) was used.

Season 1 variable to Season 2: r-squared
Full Season: .88
First Half: .87
Second Half: .88
September: .85
55% 1H & 45% Sept: .91

Overall, the results are the same with second-half velocity holding up the best on its own. When I start adding multi-variables, an interesting mix of first half and September numbers take precedence.

For a simple rule to cover all velocity readings, use second-half velocity values to project a starter’s velocity next season. If an owner wants to go that extra step, they can use a mix of values for those pitchers who have seen a late-season drop.