An independent baseball research project · Satwik Medipalli

Does redesigning a pitch actually work?

Every offseason, MLB pitchers rebuild their pitches: throwing harder, adding movement, changing their arm slot. I tracked 1,054 of those changes across the 2020–2026 seasons to find out which ones actually led to better results.

Get the answer

In plain English

The question

Teams and private pitch labs now redesign pitches every winter, using high-speed cameras and ball tracking. A pitcher might add 2 mph, change how much the ball rises, or drop his arm angle. The pitch looks different on the radar. But does it get hitters out more often?

What I did

I took every pitch thrown in MLB from 2020 to 2026 (4,530,013 pitches, public Statcast data) and found 1,054 cases where a pitcher’s pitch changed a lot from one season to the next. Then I compared each one with similar pitchers whose pitch didn’t change, so the difference can be credited to the redesign and not to luck, age, or a league-wide trend.

What I found

Of 16 kinds of change, only two reliably made pitches better. The clearest is simply throwing harder. Most other changes, including the popular ones like adding “ride” to a fastball, showed no clear effect on results.

The headline number

+0.76

runs saved per 100 pitches, when a pitch gains 2+ mph

A typical pitch is thrown about 367 times a season. At that rate, gaining velocity is worth about 2.8 fewer runs allowed per season, from one pitch alone.

Hitters also swung and missed +4.1 percentage points more often against it.

Based on 109 velocity gains. 95% range +0.51 to +0.99: even the cautious end is clearly above zero.

How to read the numbers on this page

Runs saved per 100 pitches: How many runs a pitch prevents, compared with an average pitch, for every 100 thrown. Positive is good for the pitcher. It judges balls in play by how hard they were hit, not where they landed, to strip out luck.
Redesign: A change to a pitch from one season to the next that is big enough to matter (e.g. 2+ mph faster), bigger than measurement noise, and lasts the whole season.
Look-alikes: Up to 5 pitchers who threw the same pitch, looked similar the year before, and didn’t change it. The redesign gets credit only for improvement beyond theirs.
95% range: The range the true effect very likely falls in. If it includes zero, the data can’t rule out “no effect”.

See it in the data · scroll

1,054 redesigns. One pattern.

Every dot is a redesign

Hundreds of pitches that changed over an offseason, each compared with look-alike pitchers who didn’t change.

Across: how much faster the pitch got. Up: how much better it did than its look-alikes, in runs saved per 100 pitches. Hover any dot.

One of them

That’s Cole Ragans’s fastball.

Most changes

Most redesigns weren’t about speed. They bunch up near zero mph and land above and below the line about equally. Added rise, new arm slots, more spin: no clear effect.

Their plain average gap: 0.00 runs per 100, and the dots scatter widely on both sides.

The exception

Pitches that gained 2+ mph sit above the line. The white line, the average at each speed change, steps up once gains pass about 2 mph.

+0.76

runs saved per 100 pitches after matching and adjustment (95% range +0.51 to +0.99).

One dot = one pitcher’s pitch, one offseason. Single dots are noisy; the averages carry the claim. Values beyond ±5 are pinned to the edge.

01 · What a redesign looks like

One pitcher, then a thousand.

four-seam fastball · 2022 → 2023

Cole Ragans

92.1mph▲ 4.4

Shape moved 3.0 in. Rise 18.5 → 16.1, run 10.7 → 12.4.

Whiff rate10%25%

Cole Ragans came into 2023 throwing his fastball 4.4 mph harder. Hitters swung through it more than twice as often.

But pitchers improve for all kinds of reasons. So the study finds pitchers who looked like Ragans the year before (same pitch, similar shape, similar results, similar age and role) and who didn’t change:

Look-alike, same seasonsSpeed changeRuns change
Andre Pallante+1.0 mph+0.76
Tommy Henry−0.7 mph−0.10
JP Sears+0.1 mph+0.27

Ragans improved by +1.43 runs per 100 pitches. His look-alikes improved by +0.31 on average. The gap, +1.12, is the part credited to the redesign.

One pitcher is an anecdote. He was also traded mid-season, which is exactly why one case proves nothing. The study repeats this comparison for every redesign that had a close match (923 of 1,054) and averages.

02 · The scoreboard

16 kinds of change. Two held up.

Each redesign is named for the thing that changed most. Ranked by effect on runs. A change counts only if it survives a correction for testing 16 things at once and shows up in both early and late seasons.

ChangeRuns saved per 100 pitches, with 95% rangeVerdict

Right of zero means the pitch got better for the pitcher. The line is the 95% range: if it crosses zero, the data can’t rule out no effect. Tap a row for the season-by-season split.

What about the average redesign?

Across every change, the redesigned pitch did a little better than its look-alikes: +0.23 runs per 100 (+0.13 to +0.33). But that number moves from +0.06 to +0.26 depending on how big a change has to be to count. Treat it as “small, maybe”. The velocity result doesn’t have that problem.

03 · Look up a pitcher

Find your guy. See what changed.

Every redesign in the study with a close match, searchable by name. Watch the pitch move from one season to the next, and see how it did against pitchers who didn’t change.

Try

Cole Ragans0 redesigns

“vs look-alikes” is this pitch’s change in runs saved per 100, minus the average change of pitchers who looked like him and didn’t change. One case is noisy: treat it as a story, not a verdict.

04 · Why believe it

Four ways it could have been wrong.

Each one was checked. None of them explain the result.

01 · Fresh seasons

It held up on seasons the rules never saw

Every rule was set using 2021–2024. The later seasons were checked once, at the end. The velocity effect showed up in both.

2021–24+0.65

2024–26+1.06

02 · No head start

Redesigners weren’t already improving

If pitchers who change were already on the way up, the year before the change would look good too. It doesn’t.

Year before−0.09

Year of change+0.23

03 · Not luck

Fake redesigns find nothing

The whole study was rerun 200 times with fake “redesigns” picked from pitchers who changed nothing. 0 of 200 came close to the real result.

04 · Not just healing

It isn’t injured arms getting velocity back

14 of the 109 velocity gains only regained speed lost the year before. Taking them out leaves the effect where it was.

All gains+0.76

Without rebounds+0.76

05 · How it works

Built to be challenged.

01
Find the changes
Compare every pitch to its previous season on speed, movement, spin, and release. A redesign is a coach-sized change that beats measurement noise and lasts all season. Results are never looked at here.
02
Score the pitch, minus luck
Runs saved per 100 pitches, judging balls in play by how hard and at what angle they were hit, not where they landed.
03
Find look-alikes
Match each redesign to up to 5 pitchers with the same pitch who looked similar the year before but didn’t change. Compare how much each side improved.
04
Hold seasons back
Set every rule on early seasons, then check the later ones once. Uncertainty comes from resampling pitchers, and it was tested on simulated leagues with a known answer.

What it can’t tell you

This is observational. It controls for what Statcast measures, not health, intent, grip, or a whole training program changing at once. Pitchers who abandoned a change in spring, or lost their job, never make the sample. And “works on average” doesn’t mean it works for every arm.