Let the ball sit deeper.
The ball sits deeper than a fastball, closer to the palm.
Fastball arm, ten miles slower.
The same arm action, the ball deeper in the hand. It arrives late, fading toward the arm and tumbling under the barrel. This is how the pitch fools timing.
Specimen 02
seam-informed schematic
Sourced, not corrected
View
Hand
The deception / 02
It looks exactly like the fastball and arrives a beat slower. It wrecks a hitter’s timing.
Ball depth
Deeper in the hand
Spacing
Wide spacing
Thumb
Thumb curls toward the index to form the inside circle.
Release feel
Keep fastball arm speed and let the deeper grip take speed off.
The ball sits deeper than a fastball, closer to the palm.
Palm gap cue
See it live / via @PitchingNinja
The grip is the cause. This is the effect, the same pitch shot in a real bullpen. Sourced, not corrected.
Fastball arm speed, always. Let the grip take the speed off, never your arm.
The thumb and index finger form an OK circle against the inner side of the ball while the other three fingers lay across the top. The ball rests deeper in the hand, which bleeds velocity, and the off-center grip tilts the axis toward the arm for fade.
Paraphrased from MLB.com Glossary, not quoted.
MLB.com calls the circle change the most common changeup grip: thumb and index in a circle on the inner side, the other three fingers over the rest of the ball. Held farther back in the hand, sometimes toward the palm, which is what takes the speed off.
Paraphrased from the official glossary.
Driveline cues the fade by rolling the hand over the ball or swiping its inside: the more side-spin, the more arm-side run. They place most right-handers between roughly 1:30 and 2:30 on the spin clock.
Mechanical cues paraphrased from Driveline.
The deception is the velocity gap on identical arm action, so the standard cue is to keep fastball arm speed and let the grip take the speed off, not the throw.
The "start the swing before the pitch arrives" deception is the glossary definition; the maintain-arm-speed cue is standard and widely taught.
Grip geometry is schematic and shows a standard circle-change family; individual circle size varies by hand.
Grip shape only matters if the release makes sense. This room keeps the player reading pressure, thumb support, and ball depth before movement numbers show up.
Fastball arm speed, always. Let the grip take the speed off, never your arm.
The ball sits deeper than a fastball, closer to the palm.
Main top-finger control
Keep fastball arm speed and let the deeper grip take speed off.
Film Room / watch a master
The schematic shows the release path. This is the release: Gibson walking through the delivery himself. Real hand, real ball, real finish, no model can fake it.
The arm matches the fastball; the velocity comes off from the deeper grip and the loosened circle, not a softer throw. A slight pronation at release tilts the axis arm-side, so the ball fades toward the throwing hand and tumbles as it crosses.
Paraphrased. Maintaining fastball arm speed is a standard changeup cue echoed across MLB.com and instructional sources.
Buried in the hand, the ball comes out slower with a little arm-side fade. The hitter has already committed to fastball timing, so the swing starts early, and the barrel passes over the top of it.
The velocity gap and the fade work because the arm action is identical to the fastball. The eyes read fastball, the swing starts on fastball timing, then the ball arrives ten miles slower while fading down and toward the arm, and the barrel passes over it. Take the arm speed off and the deception collapses.
Synthesizes the MLB.com deception definition with Driveline mechanics. Elite circle changes (Hamels, Williams) sit near a 10 mph gap.
Spin axis and force direction, drawn in render space
About 1,746 rpm league average since 2015 per MLB.com. Individual versions vary widely; Williams is an extreme outlier.
Active-spin percentages from Savant, 2024. The fade comes from the axis tilt, not from dead spin.
9.8 in is Webb 2024; 19.4 in is Williams 2024, an outlier, not typical.
The 8 to 12 mph separation is the typical fastball-to-changeup gap; elite circle changes sit near 10 mph.
Webb read near 3:00 on Savant in 2024; the exact clock is method-dependent (observed vs spin-based axes differ).
It fades toward the arm and drops under a fastball-timed swing, all on identical arm speed.
“It's just an outlier pitch. The spin I'm able to create makes it different from every other changeup.”
Devin Williams to The Athletic in 2020, re-published in MLB.com's Airbender explainer. It speaks to his own changeup, the extreme tail of the family, not the circle change in general.
Three documented changeups, three answers. The visual is our own seam schematic. Every figure is season-stamped and links to its source.
The textbook circle change and, by run value, the best changeup of the past two decades, a whiff-first pitch built on a ten-mile gap and identical arm action.
2011. Led every pitch in baseball in whiff rate among pitches thrown as often.
2011, first in MLB.
2011, first in baseball.
2011, with +8.4 horizontal and +6.0 vertical movement vs an average changeup.
Held about a 10 mph separation throughout his career.
The extreme tail, the Airbender: the highest-spin changeup on record. We show it as the outer limit of the circle grip while flagging that its movement pushes toward screwball territory.
Off his four-seam in 2024; the precise four-seam figure varies by sample.
Highest-spin changeup on record per MLB.com; other sources put it at 2,827 to 2,852 for other samples. All agree it leads.
2024, wider than the 17-inch plate; the lone arm-side pitch among the season's six biggest movers.
2024, .162 batting average against.
The other archetype: a high-volume movement-and-location change built for weak contact and ground balls, not whiffs, the counterweight to the bat-missers.
2024; ~945 to 990 thrown depending on the cut, one of the highest changeup usages in baseball.
Off his sinker, tighter than the norm; his change wins on movement and location, not velocity separation.
Deliberately shown: a circle change that does not live on whiffs is a legitimate, high-ground-ball profile.
The historical benchmark, and the honest one. Widely called the best changeup of its era, built on identical arm action rather than movement; the data shows the deception, not the velocity gap, did the work. He has his own chapter in the Craftsmen.
PITCHf/x, 2007. The pitch was reputed to have a 15-20 mph gap; the measured separation is closer to a normal MLB changeup, so the deception is what made it elite.
Filed only when the bar is met. A real figure from the wrong tracking system, or a great arm whose signature pitch sits in a different category, gets left off rather than dressed up. The gap is the honesty.
Tier 03 / Field Notes
Every pitcher fiddles. A thumb creeps lower, a seam catches more leather, a cue from a coach suddenly makes the pitch move. Pitch Atlas keeps those experiments visible, labeled, and debated, so the small discoveries stop disappearing into group chats and comment sections.
How notes rank
There is no single correct way to throw a pitch, but there are better and worse claims. A funny line should never outrank a tested grip. So notes do not rise on raw votes. They rise on weighted signals - provenance and adoption first.
Source tier and evidence. A coach-observed note with a link outranks an unverified hunch with none.
How many other pitchers independently tried the same tweak. Real replication, not passive clicks.
Marked useful, measured against views so a small loud crowd cannot stuff the rank.
How close a note is to your level, slot, and velocity. Computed in your session, never stored.
Considered practice: a real sample size behind the claim, not one good bullpen.
The vocabulary
The living layer
Soon you will log your own grip tweak, mark the ones you have tried, and flag anything off, anonymously or under a handle you keep. When the community layer opens, every note will carry a source and confidence label, a content filter will block abusive language, and any note can be flagged. A note hides automatically once enough people report it.
When they open, every community variant will carry the same source and confidence labels as the records above. Nothing appears here unsourced, and no count is shown until it is real.