Watch an elite pitcher throw ten fastballs and the deliveries look identical. Put those same ten throws through motion capture and they are not. Joint angles and release points wander a little from pitch to pitch. The real question is what kind of variation separates good pitchers from great ones, and whether your mechanics report is even measuring it.

Why a biomechanics snapshot can mislead you

Most pitching biomechanics reports grade snapshots, such as pelvis rotation at foot plant or shoulder external rotation at its maximum. A snapshot tells you where a body part was at one instant. It cannot show how the whole motion was organized to get there.

A helpful way to think about it: joint angles at key moments are like shadows on a cave wall, the visible result of how the movement is organized underneath. Stergiou and Decker review evidence that movement variability is, in their words, "far from being a source of error," and that its structure over time carries information about health and skill [1]. Nonlinear measures try to describe that structure across the whole delivery instead of at single frames.

What is sample entropy?

Sample entropy asks how predictable a signal is over time. If you know the last few frames of a joint angle, how well can you predict the next ones? Richman and Moorman introduced it in 2000 as an improvement on an older measure, approximate entropy, because it gave more consistent results on short, noisy biological recordings [2]. Low sample entropy means a regular, repeatable signal. High sample entropy means an irregular, less predictable one.

What is spectral entropy?

Spectral entropy asks how spread out a signal's rhythm is. Break a movement signal into its component frequencies. If most of the power sits in one or two frequencies, spectral entropy is low. If power is smeared across many frequencies, it is high. Inouye and colleagues introduced the measure to quantify irregularity in brain-wave recordings [3], and it has since been borrowed for movement analysis.

Illustrative: a signal driven by one rhythm puts its power in one frequency bin (low spectral entropy). A signal mixing many rhythms spreads its power out (high spectral entropy). Source: [3] Inouye et al. 1991
Treat entropy as a research tool

Entropy values depend heavily on analysis settings, recording length, sampling rate and filtering. They are not standardized for pitching, and we found no peer-reviewed spectral entropy study of the pitching delivery. Treat any single entropy number with caution.

Is more variability good or bad for pitchers?

Neither extreme. Stergiou, Harbourne and Cavanaugh proposed that healthy, skilled movement shows an optimal amount of variability with a particular structure [4]. Too little looks rigid and robotic. Too much is noisy and unstable. Both extremes adapt poorly when something unexpected happens, and skilled movement sits in between [1, 4].

Illustrative: too little variability looks robotic, too much is unstable, and skilled movement sits in between. Source: [4] Stergiou, Harbourne and Cavanaugh 2006

Batting research shows what good variability looks like in practice. Gray found that hitters who improved increased the variability that kept the swing on time and reduced the variability that knocked it off [8]. Variability itself never disappeared. The useful question for a pitcher is whether a given wobble keeps the ball on target and the velocity up, or pushes them off.

Do elite pitchers have less variable mechanics?

Somewhat, for some measures. Fleisig and colleagues compared 93 pitchers from youth to MLB, five fastballs each. Variability in several position measures, such as foot placement and trunk tilt, tended to be highest in youth and lower at higher levels. Timing variability did not differ by level, and neither did variability in joint loads [7]. Positional consistency improves with development, but this study did not find that better pitchers have tighter timing.

Release point data tells a similar, nuanced story. In MLB starters, release consistency and pitch speed predicted performance, but together they explained only a minority of it, roughly a quarter or less [9]. A 2024 analysis found MLB starters had smaller release-point variability than minor leaguers, with each centimeter less horizontal variability associated with a modestly better expected FIP [10]. In a small sample of Division I pitchers, release variability did not relate to in-season success at all [11]. Consistency helps. It is far from the whole story.

Look at the whole motion

Get a full-delivery breakdown with Pitch AI

Pitch AI's AI Video Review breaks down your delivery with cues and top fixes. Reviewing several throws from the same session, instead of one frozen frame, gives you a better sense of what is consistent and what is wandering.

Download on the App Store Get it on Google Play

What separates hard throwers? The 2026 entropy preprint

The largest entropy analysis of pitching to date is a 2026 preprint by Adam Bloebaum, an author affiliated with Driveline Baseball. It analyzed 28,307 fastballs from 1,722 athletes, youth to pro, using 300 Hz markerless motion capture, and computed nonlinear features including sample entropy and spectral entropy on joint-angle signals from peak knee height to release [5].

  • Average complexity was about the same at every velocity level. Hard throwers did not move with more or less complex trajectories on average.
  • Throw-to-throw reproducibility separated them. The within-session spread of these complexity measures was much lower in harder throwers, with the gap between pitchers under 85 mph and those at 90 mph or above around a moderate effect size.
  • Nonlinear features predicted velocity better than snapshots. They explained about 41% of velocity variance in cross-validation, against about 21% for standard point-in-time metrics.
  • Early variability may not be a bad sign. In a subset of 473 athletes followed over time, higher baseline pelvic variability predicted more later velocity gain.
In 28,307 fastballs from 1,722 athletes, whole-delivery complexity features explained about twice as much velocity variance as point-in-time metrics. This is a preprint that has not been peer reviewed. Source: [5] Bloebaum 2026 (preprint)

This is an early preprint, not yet peer reviewed, from an author at a commercial training facility, and the design is observational. The author notes causality is untested. Treat it as promising. One plausible reading of the last two findings together is "explore early, converge later," but that is our interpretation and has not been tested. A companion preprint by the same author reached a related conclusion: skilled pitching is characterized by better-structured joint variability, with variation converging in ways that preserve the release [12].

Where consistency matters most for command

For location, the most useful pitching study points early in the delivery. Glanzer and colleagues measured 47 pitchers throwing ten fastballs each at the center of the zone. Variability in five parameters explained 58% of the differences in how consistently they located, and three of those were shoulder and trunk positions at foot contact [6]. The authors suggest that reducing shoulder variability early in the motion may help location consistency.

Three of the five parameters were shoulder and trunk positions at foot contact, early in the delivery. Source: [6] Glanzer et al. 2021

How to use this in training

  • Judge variability by outcome. Ask whether the variation keeps the ball on target and the velocity up. Low variability that misses the target is not a win.
  • Assess with more throws. One or two throws per assessment under-describe a delivery. Capture several game-speed throws, ideally eight or more, and look at the spread as well as the average.
  • Target early consistency for command. Use target and tempo constraints that stabilize the delivery into foot contact, in line with the Glanzer findings [6].
  • Vary speeds on purpose. In MLB data, variation in pitch speed was itself a positive predictor [9]. Repeating your delivery does not mean throwing everything at one speed.

The goal is a set of movement options that converges on stable patterns that get the job done. Organize the variation. You will never erase it. For the practical side of filming that variation, see why video review matters for pitching mechanics, and for the positions that matter most, our guide to pitching mechanics and velocity biomechanics. Earlier parts of this series cover why fixing one flaw doesn't fix your pitching and perception-action coupling. The next series starts with what intent means in pitching.

Frequently asked questions

Do great pitchers have the same mechanics every pitch?

No. Even elite pitchers vary from pitch to pitch. Variability in some positions decreases with level, but timing variability did not differ across levels in one study, and skilled movement shows an optimal, structured amount of variability.

What is sample entropy in pitching?

Sample entropy measures how predictable a movement signal is over time. Low values mean regular and repeatable, high values mean irregular. It is a research tool, sensitive to analysis settings, and not a standardized pitching grade.

Is low movement variability always better?

No. Movement science describes an optimal middle ground: too little variability is rigid, too much is noisy. What matters is whether variation keeps the ball on target and the velocity up.

Do hard throwers move with more complex mechanics?

In a large 2026 preprint, average complexity was about the same across velocity levels. Hard throwers instead reproduced their movement signature more consistently from throw to throw. It has not yet been peer reviewed.

Organize the variation

Review your delivery across many throws

Pitch AI's AI Video Review breaks down your delivery with cues and top fixes, and Pitch AI's Pitch Lab logs velo, IVB, spin, axis and release so you can track whether your release and pitch shapes are converging over time.

Download on the App Store Get it on Google Play

Sources and further reading

  1. Stergiou N, Decker LM (2011). Human movement variability, nonlinear dynamics, and pathology: is there a connection? Human Movement Science 30(5):869-888. https://doi.org/10.1016/j.humov.2011.06.002
  2. Richman JS, Moorman JR (2000). Physiological time-series analysis using approximate entropy and sample entropy. American Journal of Physiology: Heart and Circulatory Physiology 278(6):H2039-H2049. https://doi.org/10.1152/ajpheart.2000.278.6.H2039
  3. Inouye T, Shinosaki K, Sakamoto H, et al. (1991). Quantification of EEG irregularity by use of the entropy of the power spectrum. Electroencephalography and Clinical Neurophysiology 79(3):204-210. https://doi.org/10.1016/0013-4694(91)90138-T
  4. Stergiou N, Harbourne RT, Cavanaugh JT (2006). Optimal movement variability: a new theoretical perspective for neurologic physical therapy. Journal of Neurologic Physical Therapy 30(3):120-129. https://doi.org/10.1097/01.NPT.0000281949.48193.d9
  5. Bloebaum A (2026). Reproducibility of nonlinear movement complexity, not complexity itself, distinguishes elite throwing. SportRxiv preprint, June 8, 2026 (not peer reviewed; author affiliated with Driveline Baseball). https://sportrxiv.org/index.php/server/preprint/view/919
  6. Glanzer JA, Diffendaffer AZ, Slowik JS, Drogosz M, Lo NJ, Fleisig GS (2021). The relationship between variability in baseball pitching kinematics and consistency in pitch location. Sports Biomechanics 20(7):879-886. https://doi.org/10.1080/14763141.2019.1642378
  7. Fleisig G, Chu Y, Weber A, Andrews J (2009). Variability in baseball pitching biomechanics among various levels of competition. Sports Biomechanics 8(1):10-21. https://doi.org/10.1080/14763140802629958
  8. Gray R (2020). Changes in movement coordination associated with skill acquisition in baseball batting: freezing/freeing degrees of freedom and functional variability. Frontiers in Psychology 11:1295. https://doi.org/10.3389/fpsyg.2020.01295
  9. Whiteside D, Martini DN, Zernicke RF, Goulet GC (2016). Ball speed and release consistency predict pitching success in Major League Baseball. Journal of Strength and Conditioning Research 30(7):1787-1795. https://doi.org/10.1519/JSC.0000000000001296
  10. Wakamiya K, Nagamoto H, Yamaguchi R, et al. (2024). Relationship between ball release point variability and pitching performance in Major League Baseball. Frontiers in Sports and Active Living 6. https://doi.org/10.3389/fspor.2024.1447665
  11. Whiteside D, McGinnis RS, Deneweth JM, Zernicke RF, Goulet GC (2016). Ball flight kinematics, release variability and in-season performance in elite baseball pitching. Scandinavian Journal of Medicine & Science in Sports 26(3):256-265. https://doi.org/10.1111/sms.12443
  12. Bloebaum A (2026). Skilled pitching is not less joint variability but better-structured joint variability: a time-resolved uncontrolled-manifold analysis of the baseball pitching delivery. SportRxiv preprint, August 12, 2026 (not peer reviewed). https://sportrxiv.org/index.php/server/preprint/view/1010

Related reading