A new pitch has to prove two things. First, that the shape is real and repeatable. Second, that the shape actually plays against hitters [1]. The first test is quick and cheap. The second is slow and expensive, and it is where most "breakout" pitches quietly disappear. Here is how to run both tests honestly, what numbers to watch, and the decision rules pro coaches use to keep or scrap a pitch.
Test 1: is the shape real?
A shape is real when you can reproduce it across separate blocks and on another day, under the same conditions: same mound, same ball, same tracking device if you have one. Premier Pitching describes this as needing dozens of reps under stable conditions, checking velocity, movement, spin axis and release consistency [1]. Three great pitches in a row is a hint. Thirty pitches whose average and spread both look good is evidence.
Judge the whole cloud of pitches, including the ordinary ones. A new average with a much wider spread can be harder to use than the pitch you already had [2]. Repeatable also means repeatable in the dimension that matters. A splitter that wanders a few inches side to side but always dives may still be doing its job; a changeup that sometimes dives and sometimes floats is a problem.
Test 2: does the shape play against hitters?
Even a perfectly repeatable shape can fail once a batter is in the box. It might arrive in the same lane as your fastball, or you might not be able to land it. Premier Pitching says this stage takes hundreds of pitches of that type, or a model-based estimate, and warns against calling a pitch game-ready from a small cluster of good outcomes [1].
Professional data show why. FanGraphs measured how quickly its pitch models become reliable during a season [3]:
| Metric | What it describes | Pitches to become reliable |
|---|---|---|
| Stuff+ | Pitch shape and speed | About 80 |
| Pitching+ | Shape plus location | About 250 (relievers) to 400 (starters) to beat projections |
| Location+ | Where pitches are thrown | About 400 |
The ratio is the useful part: location evidence takes roughly five times as many pitches as shape evidence. These are MLB numbers from one set of models, so treat them as a guide to proportion. For a high school pitcher, the lesson is the same. The shape tells you early. Command and results take much longer, so judge them over weeks of live work.
Run it like an experiment
Most pitchers "test" a grip by throwing it until a good one comes out, then remembering that one. A better process comes from the Pitch AI Pitch Design Playbook, which pulls together advice from Driveline, Tread, Premier Pitching and others [2]:
- Change one thing per block. Grip, seam orientation, pressure or cue. If you change two, you will not know which one worked.
- Alternate old and new (A-B-A-B). Throwing all of the old version first and all of the new version last mixes the grip change up with fatigue.
- Decide what counts as a pass before you look. For example: "B drops at least 4 more inches than A while losing no more than 1 mph."
- Keep every pitch in the record and retest another day. Exclude only real mistracks. A grip that only works on Tuesdays is not a pitch yet.
Tread adds a useful trick for anything velocity-related: a blind split test. A partner hides the radar readings while you alternate the old and new versions for five or six throws each at equal effort. Faster at the same effort is a keep; slower, or more effort for the same speed, is a drop [4].
See what the grip did
Compare old and new versions in Pitch Lab
Pitch AI's Pitch Lab logs every grip change so you can see what it did to the shape. Log velo, IVB, spin, axis and release for each version, then check the movement chart to see whether the new pitch actually sits apart from the ones you already throw.
Protect velocity first
The most common way a "better" pitch turns out worse is lost speed. A Driveline trainer puts the priority bluntly: "Velocity is king. Then we start thinking about shapes." A slot change that costs two miles an hour usually is not worth it, while one that costs nothing and feels good is worth exploring [5].
The same logic applies to breaking balls. Tread found that for curveball-type pitches, velocity mattered more for whiffs than total depth, with a whiff inflection point near 80 mph [6]. A slow 75 mph curve with huge depth can lose to an 80 mph pitch with less movement. Tread's coaches make the same argument about sweepers: a 12-inch sweeper thrown harder and with better command can beat a 21-inch version that costs six miles an hour [7]. Brian Bannister actually coaches less sweep once it passes about 17 inches, the width of the plate, because command and arsenal balance suffer beyond that [8].
One more velocity check: after a design session, throw a few fastballs and compare them with your normal velocity. If a new grip changed your delivery and slowed your fastball, the experiment cost more than it showed [2].
Is it distinct from what you already throw?
A new pitch that lands on top of an existing one on a movement chart adds a name to your arsenal without adding a problem for the hitter. Ben Brown said his goal for his new two-seamer was a clear distinction from his four-seamer [9]. Driveline's look at Logan Webb and Michael Wacha showed that added fastballs can help a pitcher even when they do not grade as plus pitches on their own, because they change what hitters have to cover [10].
Compare the new pitch with your matchup fastball in plain terms: the speed gap, the difference in vertical movement and the difference in horizontal movement [2]. If all three are small, you probably built a variation of a pitch you already have.
Can you land it?
Driveline coordinator Dylan Gargas put it memorably: a pitch with great metrics that misses by three feet is "an uncompetitive pitch" [10]. A shape you cannot throw for a strike, or at least into a chase zone on purpose, has no job in a game. That is one reason fastball variants are often a smarter first project than a big new breaking ball; they are easier to get in hand and locate.
Tread's changeup coach offers a quick kill rule: if a grip is not working after a day or two, move on, and if you cannot locate it down and to the arm side, it is probably not the right long-term grip for you [11]. Speed matters here too. Shape evidence comes fast, so an honest "no" can come fast as well.
Advancement gates: from bullpen to games
A simple way to organize the whole process is a set of gates. A pitch only moves forward when it has the evidence for the current stage [2].
| Stage | Question | Evidence to advance |
|---|---|---|
| Exploration | Can this grip create the flight I want? | More than one good example, comfortable feel, plausible speed |
| Repeatability | Can I recreate it? | Stable clusters across blocks and on another day |
| Execution | Can I hit a target window with it? | Strike and chase trials with recorded targets, in mixed sequences |
| Hitter exposure | Does it fool a batter? | Swings, takes and contact quality, with count and location recorded |
| Adoption | Is it improving my arsenal in games? | Matchup-specific evidence without losing existing weapons |
The last gate matters more than it looks. A new pitch that steals usage from your best pitch can make your arsenal worse overall. For a full process, see our pitch design workflow; for curveball-type shapes, our guide on how to throw a curveball; and if your candidate is a two-seamer or cutter, read why adding another fastball shape can help.
What to do this week
- Write your pass criteria for your current project before your next bullpen.
- Run one A-B-A-B session with the old and new version, and keep every pitch in the log.
- Recheck your fastball velocity at the end.
- Book the retest on a different day before you decide anything.
If you want to understand the stuff grade many tracking tools report, continue to what Stuff+ measures and ignores. If you are still deciding when to start, go back to when to work on a new pitch.
Frequently asked questions
How many pitches does it take to know if a new pitch works?
Two different numbers. Confirming the shape usually takes dozens of reps under the same conditions. Knowing whether it beats hitters takes hundreds of pitches. In pro data, FanGraphs found its Stuff+ grade becomes reliable after about 80 pitches, while its Location+ grade needs about 400.
Is a new pitch good if it has great movement numbers?
Movement is only the first test. The pitch also needs to hold velocity, look different from your other pitches, and be locatable. Driveline coaches note that a pitch with great metrics that misses by three feet is uncompetitive.
Should I give up velocity for more break on a breaking ball?
Usually no. Tread found that for curveball-type pitches, velocity mattered more for whiffs than total depth, and Bannister coaches less sweep once it passes about 17 inches because command suffers. A harder, smaller shape often wins.
What is the best way to compare two grips?
Change one variable, alternate small blocks of the old and new versions (A-B-A-B), set your pass criteria before looking at results, keep every pitch in the log, and retest the promising version on a different day.
Keep the evidence
Run your next pitch trial with Pitch AI
Pitch AI's Pitch Lab logs every grip change so you can see what it did to the shape, plots your arsenal on a movement chart, and flags pitches that overlap. When a candidate passes your test, it ranks what to develop next.
References
- Premier Pitching Performance Team (2026). The Pitch Design Phase in Baseball: Developmental Fit, Workload-Informed Timing, Biomechanical Determinants of Effectiveness, and Evidence-Based Evaluation Thresholds. Premier Pitching Chronicles. https://premierpitching.com/blogs/premier-pitching-chronicles/the-pitch-design-phase-in-baseball-developmental-fit-workload-informed-timing-biomechanical-determinants-of-effectiveness-and-evidence-based-evaluation-thresholds
- Pitch AI (2026). Pitch Design Playbook, with Pitch Trial Log and Pitch Decision Log templates. Internal coaching synthesis, September 2026.
- McGrattan O (2023). Stuff+, Location+, and Pitching+ Primer. FanGraphs Library, March 10, 2023. https://library.fangraphs.com/pitching/stuff-location-and-pitching-primer/
- Tread Athletics (Brewster B) (2026). Throw Harder By Changing Your Pitching Mechanics [The Right Way]. YouTube, September 24, 2026. https://youtu.be/_EilR25ydwY
- Sawchik T (2026). Pitchers are going low, and for good reason. But it's not for everyone. Driveline Baseball, March 12, 2026. https://www.drivelinebaseball.com/2026/03/pitchers-are-going-low-and-for-good-reason-but-its-not-for-everyone/
- Tread Athletics (Zombro T) (2024). What Is The 'Death Ball' And How Do You Throw It? YouTube, October 31, 2024. https://youtu.be/THd6Ch2VMks
- Tread Athletics (2025). Top Pitching Velocity Hacks To Try. YouTube, November 23, 2025. https://youtu.be/SLZe8s4uSkY
- Tread Athletics (2022). Brian Bannister: Mastering Pitch Design (podcast). YouTube, December 22, 2022. https://youtu.be/E0yoPt42A5Q
- Laurila D (2026). Ben Brown Pitch-Designed a Sinker Over the Offseason. FanGraphs, February 27, 2026. https://blogs.fangraphs.com/ben-brown-pitch-designed-a-sinker-over-the-offseason/
- Sawchik T (2025). How many fastballs should a pitcher have on the menu? Driveline Baseball, November 25, 2025. https://www.drivelinebaseball.com/2025/11/how-many-fastballs-should-a-pitcher-have-on-the-menu/
- Tread Athletics (Filippo R) (2026). Complete Guide To Mastering The Changeup [Grips, Cues, & Sequencing]. YouTube, September 6, 2026. https://youtu.be/fBpo461Q7y8