Almost every pitcher knows the feeling. The bullpen was sharp. The fastball hit the glove and the slider landed for strikes. Then the game starts and the zone shrinks. Walks pile up, and nobody can explain where the bullpen pitcher went.
Nerves play a part, and we cover pressure training in how pitchers train to perform under pressure. A bigger and more fixable cause sits upstream. Many bullpens are built to look good, and conditions that make practice look good are often the ones that teach the least.
Practice performance vs real learning
Learning scientists separate two things that coaches usually blend together [1]. Performance is what you can do right now, during practice. Learning is the lasting change, what you can still do days later and in a new setting. In their review of decades of research, Soderstrom and Bjork show that the two often move in opposite directions. Conditions that produce fast, impressive gains during practice frequently produce weak retention and transfer, while conditions that slow practice down, such as mixing tasks or spacing them out, often produce more learning.
Look at a typical bullpen through that lens. Twenty fastballs in a row to the same glove, then twenty breaking balls, with the same catcher and no hitter, often indoors with a net behind the plate. Every one of those features makes the pen look better. Few of them resemble a game. On Rubicon's podcast, the host argues that indoor throwing into nets is a hidden command problem: pitchers rarely see the consequences of a wild throw, and wide open space at a real field can look very different [2]. That is a practitioner observation, and it fits the learning research well.
The baseball evidence for harder practice
Hall, Domingues and Cavazos ran one of the few practice-design studies on skilled baseball players [3]. Thirty college hitters were split into three groups. Two groups got twelve extra batting practice sessions over six weeks, each with 45 pitches: 15 fastballs, 15 curveballs and 15 changeups. One group saw the pitches in blocks of one type. The other saw all three mixed at random. On a game-like test with random pitches, the random group improved 56.7% from the pretest, the blocked group 24.8% and the group with no extra practice 6.2%.
The study involved hitters, and the sample was small, so treat it as a strong hint. Still, the principle carries straight to pitchers: practice that asks you to solve a new problem on each rep builds the skill a game demands.
Find your challenge point
Harder is only better up to a point. Guadagnoli and Lee describe a "challenge point" for every learner [4]. A task that is too easy gives the brain little new information, since every outcome is predictable. A task that is too hard overwhelms. The optimal difficulty depends on the person's skill level, and it moves upward as they improve.
Most practice sits far below that point. On the Rubicon podcast, the host, a former Driveline coach, and Chicago Cubs strength coach Steffen Simmons discuss how many coaches get uneasy when players fail more than a few percent of the time in practice [2]. If every pitch hits the glove, the pen was too easy to teach anything.
A practical starting point is a target you hit about half the time. Driveline's Differential Command Program uses a similar rule, advancing a pitcher when they throw at least 50% strikes in a block [5]. That is a program rule, untested in a published study, and it gives you a usable number.
Make command work a habit
Give command its own day
Pitch AI builds dedicated Command days into your throwing program, so harder, game-like command work has a set place in your week.
Make it harder every time you succeed
Rob Gray tested adaptive difficulty with eighty high school hitters [6]. In the adaptive virtual reality group, every hit made the next pitch 2 mph faster and every miss made it 2 mph slower, and pitch location spread was adjusted the same way. After six weeks of training, that group improved more than the others on most hitting measures and posted better league batting statistics the following season. Over the next five years, 8 of its 20 players (40%) reached college or professional baseball, compared with 1 to 3 of 20 (5% to 15%) in the other groups.
Two limits are worth stating. The study was on hitting, and the adaptive training happened in virtual reality. The design principle is simple enough to copy on any mound: success raises the difficulty, failure lowers it, so practice stays near the challenge point.
A bullpen plan that transfers to games
- Shrink the target after success. Start with a target you hit about half the time. Two hits in a row, make it smaller. A miss, widen it back.
- Have the catcher call spots late. The location comes as you start your motion, the way game information arrives under time pressure.
- Mix pitches you can already land. Random order for pitches that are game-ready. Keep brand new pitches in short blocks until they work (see random vs blocked practice).
- Grade it the next day. Chart the first ten pitches of your next session. That number reflects what you learned, which is what Soderstrom and Bjork's retention tests measure [1].
How to score a command pen
Charting balls and strikes throws away most of the information. Driveline's command research uses miss distance and reports that the average MLB pitch misses its intended spot by roughly a foot [7]. That is company data and not peer reviewed, and it is a useful reminder of how demanding command is even at the highest level. A simple scoring scheme for amateurs: two points for a pitch within about six inches of the glove, one point for six to twelve inches, zero beyond that. Track your average score and how it changes from week to week.
Kids need a lighter dose
The benefit of mixed, harder practice is real and smaller than many coaching clinics suggest. Brady's meta-analysis of 61 studies found an average effect of 0.57 in basic lab research and 0.19 in applied sport settings, and the effect for younger learners was much smaller than for adults (0.10 vs 0.50) [8]. A 2024 meta-analysis by Czyż and colleagues confirmed a retention benefit overall, and found it almost negligible in applied settings and not significant in young participants [9]. Wulf and Shea add that when a task already places very high demands on a learner, reducing the load tends to help more than adding to it [10].
For a twelve year old still finding the strike zone, a normal throw is already a hard task. Keep the pen simpler, use bigger targets and add challenge slowly. Our guide to youth pitcher development covers age-appropriate training, and the free 4-week throwing and command program gives a structured starting point. For older pitchers choosing a full plan, see how to choose a throwing program.
Part 2 of this series looks at whether your bullpens should be random or blocked, and part 3 covers training for pressure.
Frequently asked questions
Why can I throw strikes in the bullpen but not in games?
Bullpens are often easier than games: one pitch type at a time, the same catcher, no hitter and no consequences. Practice that looks good in the moment often produces less lasting learning. Make pens more game-like and harder, with targets you hit about half the time and pitch types mixed once they are reliable.
What is the challenge point in pitching practice?
The challenge point is the level of difficulty where practice teaches the most. Too easy gives little new information, too hard overwhelms. A practical rule is to start with a target you hit about half the time and make it harder as you succeed.
How should I measure if a bullpen worked?
Look at your next session. Chart the first ten pitches of the following bullpen for miss distance. That reflects what you retained, which matters more than how the pen looked at the end.
Should youth pitchers practice with harder, random bullpens?
Only lightly. Research finds the benefit of mixed, harder practice is much smaller in young learners. Younger pitchers do better with bigger targets and simpler practice until they throw strikes reliably.
Practice that carries into games
Build bullpens that transfer
Pitch AI builds dedicated Command days into your throwing program and logs every session, so you can keep raising the challenge week by week and see the pattern in your training.
Sources and further reading
- [1] Soderstrom NC, Bjork RA (2015). Learning versus performance: an integrative review. Perspectives on Psychological Science, 10(2), 176-199. https://doi.org/10.1177/1745691615569000
- [2] Rubicon (2026). Casting The Die, Episode 2, with Steffen Simmons (podcast; practitioner opinion). https://youtu.be/PkRCz3dhYn0
- [3] Hall KG, Domingues DA, Cavazos R (1994). Contextual interference effects with skilled baseball players. Perceptual and Motor Skills, 78(3), 835-841. https://doi.org/10.1177/003151259407800331
- [4] Guadagnoli MA, Lee TD (2004). Challenge point: a framework for conceptualizing the effects of various practice conditions in motor learning. Journal of Motor Behavior, 36(2), 212-224. https://doi.org/10.3200/JMBR.36.2.212-224
- [5] Driveline Baseball. Differential Command Program (program notes; no published outcome data).
- [6] Gray R (2017). Transfer of training from virtual to real baseball batting. Frontiers in Psychology, 8, 2183. https://doi.org/10.3389/fpsyg.2017.02183
- [7] Driveline Baseball (2026). The 2 skills that actually make up pitching command (video; internal data, not peer reviewed). https://youtu.be/dXr7jj9ai4A
- [8] Brady F (2004). Contextual interference: a meta-analytic study. Perceptual and Motor Skills, 99(1), 116-126. https://doi.org/10.2466/pms.99.1.116-126
- [9] Czyż SH, Wójcik AM, Solarská P, Kiper P (2024). High contextual interference improves retention in motor learning: systematic review and meta-analysis. Scientific Reports, 14, 15974. https://doi.org/10.1038/s41598-024-65753-3
- [10] Wulf G, Shea CH (2002). Principles derived from the study of simple skills do not generalize to complex skill learning. Psychonomic Bulletin and Review, 9(2), 185-211. https://doi.org/10.3758/BF03196276