Most new pitches die in the gap between the bullpen and the first real at-bat. The grip worked, the shape looked great on the tracking screen, and then the pitch vanished under pressure or showed up as a hanger. That is normal, and it is predictable. A new pitch has to survive more and more game information before it is ready: a target, pitch mixing, a count, a hitter. Each of those is a constraint the pitch has to learn to live with.
The ladder: six levels from net to live at-bats
This progression borrows from challenge point thinking [1] and Driveline's flat-ground to mound sequence [7]. It is a coaching template, and the levels can be compressed for experienced pitchers adding a small variation of a pitch they already throw.
| Level | Task | Move up when | Step down when |
|---|---|---|---|
| 1 | Flat ground into a net, shape goal only | Shape shows up on most throws | Shape disappears |
| 2 | 60 feet, whole strike zone | About 60% in the zone | Under about 40% |
| 3 | Mound, blocks of the new pitch to one half of the plate | About 60% to the called half | Shape or mechanics break down |
| 4 | Mound, random pitch calls with location | About half to the called location | Well under that, or shape breaks |
| 5 | Counts, stand-in hitter, pitch timer | Competes at a game-like rate | Rushing, big drop-off in quality |
| 6 | Live at-bats with a score | Ready for low-leverage game use | Return to level 4 or 5 |
Those percentages are rules of thumb built on the challenge point idea, which says learning is best when task difficulty fits the athlete's current skill [1]. They are not lab-tested cut-offs. Their job is to keep the pitcher working at a level where he succeeds often enough to learn and misses often enough to have something to learn from.
Is the net a good first rung?
Practitioners disagree here. On Rubicon's "Casting the Die" podcast, the host argues that nets are low-information targets and suggests long tossing breaking balls and mixing grips during catch play instead [3]. If you have the space, an outdoor first rung with real ball flight gives the pitcher more to work with. The net is still a practical option in winter or in small spaces, as long as the pitcher moves off it quickly.
Climb and descend: the two rules
The ladder only works if you move both ways. Succeed at roughly 60% at a level and move up. Drop under roughly 40%, or lose the shape, and step back down. Stepping down is part of the process, and it is far better than grinding through a level where the pitch keeps breaking, which mostly rehearses the miss.
Random pitch calls at level 4 are the step most pitchers skip. Blocked practice (ten sliders in a row) makes a new pitch look better in the moment. Random practice tends to transfer better. In a classic college hitting study, practicing against randomly mixed pitch types improved transfer-test performance far more than blocked practice [8]. That was hitters, and it remains the closest baseball evidence for mixing pitch types in practice (more in random vs blocked practice for pitchers).
Measure the shape
Log your new pitch in Pitch Lab
Pitch AI's Pitch Lab logs velo, IVB, spin, axis and release for every pitch, plots it on a movement chart and flags pitches that overlap, so you can see when a new shape is real.
Periodize the skill like you periodize strength
Skill development can be planned across the year, much like strength. Otte, Millar and Klatt's PoST framework moves through coordination training with simplified tasks and low variety, then adaptability training with more variety and complexity, and finally performance training that is highly game-like, simplified again right before competition [2]. Learning-focused sessions sit earlier in the week, and performance-focused sessions cluster before games [2].
The Rubicon podcast boils this down to two kinds of days: learning days that are hard and varied, and compete days that build confidence [3]. On that view, a blocked "three fastballs, three sliders" pen right before a start is a fine compete-day tool, and a poor main method for learning a pitch [3]. Plan your new-pitch work on learning days, and keep compete days simple. For how those days fit around velocity and strength work, see the weekly training schedule guide.
Why games pull you back to the old pitch
Expect the new pitch to wobble the first time it meets real pressure. Gray describes movement patterns as valleys in a landscape: an old, well-practiced pattern is a deep valley, and a new one starts out shallow [4]. Under pressure or fatigue, the body tends to roll back toward the deeper valley. In a pitching study by Gray, Orn and Woodman, experienced pitchers under pressure hit their targets less often, and some showed changes in their delivery [5].
So when the new pitch hangs in its first outing, treat it as information. Drop a level in practice, rebuild, and climb again. Under pressure, keep game cues simple and outcome-focused, such as "finish at the glove." Technical cues invite the kind of conscious control that tends to break down when it matters.
How long does it take? Shape settles fast, command slowly
Shape tells you early. Command takes much longer to read. FanGraphs found its pitch-quality model (Stuff+) becomes reliable about 80 pitches into a season, while its location model (Location+) needs something like 400 pitches to reach similar stability [6]. Those numbers come from MLB data and full-season samples, so the exact figures do not transfer to an amateur bullpen. The ratio is the lesson: about five times as many pitches to judge location as to judge shape. Our explainer on what Stuff+ is and whether it matters covers these models in detail.
On the practice side, Brian Bannister describes a typical new-pitch block lasting about four weeks, working through the grip first, then finger position, pressure and cues, then fitting the pitch into the arsenal [9]. Pitchers vary a lot in how quickly they pick up new shapes, and variations of a pitch you already throw come faster than a brand-new breaking ball.
The practical upshot: decide whether the shape is real within a few sessions, and give command a full block of weeks before judging it. Do not drop a pitch with good shape because it missed its spots in the first two outings.
Pitch design adds high-intent throws and disrupts a grooved pattern, so count it inside your throwing plan. Avoid stacking new-pitch work on a velocity block or right before a competition peak. Our pitch design workflow covers how to test and log a new pitch.
What to do this week
- Find your level. Be honest about where the new pitch is on the ladder today.
- Pick a move-up rule and a step-down rule before the session, and write them down.
- Log every session with grip notes and shape numbers if you have tracking, so you can see trends across weeks.
- Plan the first game use for a low-leverage spot, such as an early-count pitch against a weaker hitter, and expect a wobble.
For grips and shapes, see our guides on how to throw a slider and how to throw a sweeper. If you missed the earlier parts of this series, start with Part 1 on the constraints-led approach, then Part 2 on task constraints and Part 3 on game-like bullpens.
Frequently asked questions
How long does it take to learn a new pitch?
It varies widely. Pitch designer Brian Bannister describes a typical new-pitch block of about four weeks, but command takes much longer to judge than shape. In MLB data, location metrics need roughly five times as many pitches as stuff metrics to stabilize.
When should I start throwing a new pitch in games?
When it holds its shape and hits called locations at a reasonable rate with random pitch calls and a hitter in the box. Start in low-leverage spots and expect some regression at first.
Why does my new pitch work in the bullpen but not in games?
New movement patterns are less stable than old ones, and pressure tends to pull you back toward the deeper, older pattern. Bullpens also remove game information. Drop a level in practice and climb again.
Should I practice a new pitch in blocks or randomly?
Start with blocks to find the shape, then move to random pitch calls. Blocked practice looks better in the moment, but random practice tends to transfer better to games.
From the pen to the game
Track every step of a new pitch
Pitch AI's Pitch Lab logs velo, IVB, spin, axis and release for every pitch, logs grip changes, ranks what to develop and flags overlapping pitches. Pair it with Command days in your throwing program to climb the ladder on schedule.
References
- Guadagnoli MA, Lee TD. Challenge point: a framework for conceptualizing the effects of various practice conditions in motor learning. Journal of Motor Behavior. 2004;36(2):212-224. doi:10.3200/JMBR.36.2.212-224
- Otte FW, Millar SK, Klatt S. Skill training periodization in "specialist" sports coaching: an introduction of the "PoST" framework for skill development. Frontiers in Sports and Active Living. 2019;1:61. doi:10.3389/fspor.2019.00061
- Rubicon. "Casting The Die | Episode 2 | Steffen Simmons." YouTube, 2026-09-19. youtu.be/PkRCz3dhYn0
- Gray R. How We Learn to Move: A Revolution in the Way We Coach & Practice Sports Skills. 2021. Chapter 6.
- Gray R, Orn A, Woodman T. Ironic and reinvestment effects in baseball pitching: how information about an opponent can influence performance under pressure. Journal of Sport and Exercise Psychology. 2017;39(1):3-12. doi:10.1123/jsep.2016-0035
- McGrattan O. "Stuff+, Location+, and Pitching+ Primer." FanGraphs Library, 2023-03-10. library.fangraphs.com
- Boddy K, Daniels M. Hacking the Kinetic Chain: Advanced Pitching. Driveline Baseball; 2017. Chapter 3.1, Throwing Programming.
- Hall KG, Domingues DA, Cavazos R. Contextual interference effects with skilled baseball players. Perceptual and Motor Skills. 1994;78(3):835-841. doi:10.1177/003151259407800331
- Tread Athletics. "Brian Bannister | Mastering Pitch Design." YouTube, 2022-12-22. youtu.be/E0yoPt42A5Q