Is an arm-side 12-6 curveball physically possible?

Abstract

In 2025 and 2026 Statcast data, 279 non-fastball pitcher-pitch averages combined at least 6 inches of glove-side break with at least 8 inches of drop, and none combined the same drop with 6 inches of arm-side break [1]. We propose a "reverse hammer" that pairs a slightly less supinated over-the-top curveball with a seam held on the glove-side spin pole, which Utah State University measurements suggest would add a non-Magnus push toward the arm side [2]. Our calibrated model predicts about 15 inches of drop and 8.7 to 11.4 inches of arm-side break at 79 mph [3]. No real pitch sits within two distance units of that shape, so every effectiveness estimate here is an extrapolation.

1. Motivation and gap

Our movement-direction analysis of 5,102 pitcher-pitch-seasons found 15 rows in the "arm side and drop" sector against 475 in its mirror, "glove side and drop", where curveballs and slurves live. At 16 inches or more of total break the counts are zero and 149. The cleanest version of the count, non-fastballs only, is 279 pitcher-pitch averages with at least 6 inches of glove-side break and 8 of drop, and zero with at least 6 inches of arm-side break and 8 of drop [1]. The true screwball zone at breaking-ball speed (IVB of -4 inches or less, 8 or more inches of arm-side break, under 82 mph) holds two rows, both changeups (Hoby Milner and Craig Yoho), and among over-the-top arms above 60 degrees no non-fastball moved arm side with drop [1].

The shape has a history. Adair describes the screwball as "a kind of reverse curve" used to break away from opposite-handed batters [4], Brent Honeywell's version was described as "literally a curveball in reverse" [5], and Hector Santiago described getting over the top like a reversed 12-6 [6]. Travis Sawchik of Driveline found only 7 pitches classified as screwballs in 2025, and our 2026 data contain none [7][1]. Sawchik quotes the Royals' Zach Bove calling it a "screwball-ish area" that "not many guys can accomplish" [7]. The same article notes that the screwball's injury reputation is contested, citing Warren Corbett's SABR argument, drawing on Glenn Fleisig and Mike Marshall, that the pitch is harder to throw but not shown to be more stressful [7]. The nearest real precedent for the shape is Alan Rangel's 2026 curveball, 73.2 mph with -17.6 inches of IVB and 2.2 inches of arm-side break from a 69 degree slot [1].

2. Physical basis

2.1 Separation and the seam-shifted wake

At 79 mph a baseball flies at a Reynolds number near 1.7 × 10⁵ (about 2,180 times the speed in mph) [8]. Its boundary layer stays attached over the front of the ball and separates behind the widest section, where pressure rises again. Roughness changes where that happens, which is why a trip wire on a sphere moves separation rearward at a Reynolds number of only 30,000 [9]. On a baseball, seams near the hemisphere plane (the plane through the center perpendicular to the velocity) can advance separation on their side, the wake tilts toward that side and the ball is pushed the other way. Seams on the front or inside the wake do little, and the location of separation varies shot to shot [10][11][12]. This is a boundary-layer effect, not the airfoil story of air travelling farther over one side.

2.2 Why a seam on the pole persists

Most seams rotate through the active band near the hemisphere plane and back out, so their effect averages down. The exception is the spin pole, the point where the axis leaves the ball. When the spin is fully transverse the axis is perpendicular to the velocity, so both poles lie on the hemisphere plane and a seam there stays in the active band for the whole flight. Utah State's "Scuffball" exploited this. With a seam on one pole, launcher pitches at 90 mph and 1,100 to 1,200 rpm moved about 4 inches away from that seam, with no gyro required, and Smith wrote that "there is no reason you could not bend it just as much to the other side by switching which pole has the seam" [2][13]. A real scuff on the pole produced more than 2 feet of break in the same direction, away from the scuff [14]. That direction contradicts Adair\'s prediction that a scuffed ball veers toward the scuffed side [4], and we follow the measurement. Ball scuffing is illegal and appears here only as a physics reference.

Over the top, a 12-6 curveball spins with a nearly horizontal axis, so its poles sit at the glove-side and arm-side edges of the ball. A seam on the glove-side pole should therefore push the ball toward the arm side. The design adds a second, Magnus contribution. A hand set slightly less supinated than a normal curve tilts the topspin from about 6:30 to about 5:15 on the pitcher's clock, so the Magnus force itself gains an arm-side component.

2.3 The efficiency constraint

Smith notes that gyro spin moves the pole off the separation line [13]. Geometrically, the pole's angular offset from the hemisphere plane equals the arc cosine of spin efficiency.

θ_pole = arccos(ε),   ε = |ω̂ × v̂|ε = 0.80 gives about 37°, 0.90 about 26°, 0.95 about 18°. Efficiency definition from Nathan (2020).

Our simulation assumes 80% efficiency, which places the pole roughly 37 degrees from the plane [15]. Utah State's separation maps for non-spinning balls found that a seam between the hemisphere plane and 10 degrees behind it nearly always fixes separation, while a seam 6 to 11 degrees in front does so only about a third of the time [16]. We therefore infer that the pole seam would work best on curveballs above about 90% active spin. Over-the-top curveballs averaged 77% active spin in 2025 and 2026, and Rangel's curveball registered 99% [1], so suitable spin exists but is not typical.

3. Simulated shape

We integrated each pitch with a fourth-order Runge-Kutta scheme using Nathan's pitched-ball coefficients, a spin vector fixed in space and an air density of 1.194 kg/m³ [17][3]. Against 2025 Savant pitcher averages the model overpredicted curveball movement by about 13%, so curveball lift is scaled by 0.87. The seam force is a side-force coefficient C_S, where 1 inch of movement corresponds to about 0.0112, run at a low, middle and high value because pole-seam magnitudes at game spin are unmeasured [3]. Inputs are a right-handed pitcher at a 65 degree arm angle, 79 mph, 2,650 rpm, 80% efficiency and a 5:15 tilt.

The middle case is about 18 inches of total break in a sector with no pitches, arriving at 72.5 mph after 0.479 seconds. Relative to the normal 12-6 curve, the two "hammers" differ by about 15 inches in horizontal break with almost the same drop. That is a difference in movement; how far apart they cross the plate depends on where each is aimed. When each pitch is aimed through the four-seamer's position at Baseball Prospectus's tunnel point, 23.8 feet from the plate [18], the middle-case reverse hammer finishes about 20.4 inches below the four-seamer and 1.9 inches to its arm side, 77 milliseconds later. When the reverse hammer and the 12-6 curve are aimed at the same spot, they are only 3.4 to 4.0 inches apart at the tunnel point across the three cases [3].

The figure runs the same model in your browser. The slider sets the seam-shifted wake strength, and its shaded band marks the range we consider plausible (C_S 0.035 to 0.065). Setting it to zero shows what the 5:15 tilt does alone, about 5.5 inches of arm-side break in our runs, so about 4.5 of the 10 middle-case inches depend on the unmeasured seam effect. With the tunnel box ticked, every pitch passes through the four-seamer's tunnel point, so the catcher's view shows only the divergence after the swing decision. Untick it to aim each pitch at the middle of the zone, and use slow motion to replay at a quarter speed. Live values can differ from the table by about 0.1 inch, and approach angles change with the aiming mode.

The apparent late break in the catcher's view is a property of constant acceleration, since displacement grows with the square of time. Adair noted that half of a curveball's deflection occurs in the last 15 feet [4], and Bahill and Baldwin attribute perceived sharp breaks to errors in the batter's speed estimate rather than to any change in the force [19].

4. Expected effectiveness

Our gradient-boosting shape model of whiff rate and xwOBA (velocity, IVB, HB, arm angle, extension and differences from the primary fastball; five-fold cross-validation grouped by pitcher) explains 56% of out-of-sample variance in whiff rate (RMSE 6.6 points) and 24% in xwOBA (RMSE 0.058) [3]. For the reverse hammer it predicts 30.6%, 30.4% and 29.7% whiff and xwOBA of .247, .256 and .263 in the low, middle and high cases, against a league curveball baseline of 30.8% and .279.

These numbers should carry almost no weight. There are zero real pitcher-pitch rows within two distance units of any case, and the nearest neighbours are more than three units away. They are Will Dion's 2026 curveball (81.0 mph, -9.2 inches IVB, 2.2 arm side, 13.7% whiff, .383 xwOBA over 157 pitches), Rangel's 2026 curveball (7.7% whiff, .455 over 59) and Nick Martinez's 2026 curveball (80.0 mph, -14.5, 1.0 glove side, 16.7%, .463 over 103) [3][1]. Tree-based models cannot extrapolate. Their predictions in an empty region are anchored to the nearest populated region, so the estimate mostly reflects deep, slow curveballs with a large velocity gap from the fastball, not arm-side movement.

Two mechanisms could make the pitch valuable, and neither can yet be quantified. The first is handedness. Adair reports players saying that curves breaking away are harder to hit, and that right-handed screwball pitchers threw screwballs to left-handed batters so that every batter saw a ball breaking away [4]. A reverse hammer would break away from opposite-handed hitters like a changeup, with curveball depth. The second is rarity. Langin argues that Drew Smyly's left-handed reverse breaking ball was graded poorly by stuff models because it had no peer group, and that the league never mounted a coordinated response to it [20]. A shape with no neighbours may also have no prepared hitters, but that cuts both ways for any model.

5. Complementary pitches and sequencing

The natural partner is the riding over-the-top four-seamer, about 35 inches of IVB above the reverse hammer (19.4 versus -15.3 inches) on a nearly vertical tunnel [3]. The second partner is the pitcher's own 12-6 curve, so that two hammers from the same slot differ by about 15 inches in horizontal break. Over-the-top pitchers rarely throw arm-side off-speed at all (one arm-side non-fastball row among 78 above 60 degrees) and instead throw riding changeups [1]. A riding changeup at 85 mph, 16.3 inches of IVB and 6.6 of arm-side break in our arsenal fills the space between. Against opposite-handed hitters the reverse hammer breaks away, and against same-handed hitters it dives in.

6. Execution

Grip and arm action

Use the normal curveball grip, with the middle finger along a seam, and rotate the ball so that a seam runs through the glove-side pole, the point on the index-finger side where the topspin axis exits. Pre-set the hand slightly less supinated than for the normal curve. Mark the pole and film from behind at 1,000 frames per second or more to confirm where the seam sits. Throw from a high slot.

The curveball's orientation is pre-set rather than snapped. Sakurai and colleagues (1993) found peak forearm supination of 39.9 degrees in the curveball against 19.4 in the fastball, about 72 to 76 milliseconds before release, with both pitches pronating through release [21]. Shibata and colleagues (2022) found fastball-like wrist flexion in the curveballs of six of eight pitchers and more finger adduction torque in the curveball [22]. Jinji, Sakurai and Hirano (2011) showed that the spin axis follows the hand plane just before release [23], so a slightly less supinated preset is the plausible lever for the 5:15 tilt.

What anatomy permits and what it costs

Pronation and supination happen entirely in the forearm, by rotation of the radius at the elbow and over the ulna [24]. Over-the-top slots make backspin easy, but a strongly arm-side axis from that slot would require pronation against the slot, which is why Langin's precondition for arm-side breaking balls is a high-pronation arm [20] and why Mario Delgado Genzor's pronator and supinator typology matters for who should try this [25]. The reverse hammer asks for a smaller change, less supination rather than active pronation, with the seam supplying much of the arm-side force. That distinction is our inference.

On load, curveballs carried no more elbow varus torque than fastballs in college pitchers (Fleisig et al. 2006) and youth pitchers (Dun et al. 2008), and in adolescents Nissen and colleagues (2009) measured 59.6 versus 54.1 N·m for fastball and curveball, although the curveball carried a higher wrist ulnar moment (4.9 versus 3.2 N·m) [26][27][28]. Hall summarises the literature as fastballs producing significantly greater valgus torque than curveballs [29]. Solomito and colleagues (2018) found that a larger supination moment goes with a larger varus moment [30], and this design asks for less supination. The slot is a separate cost. If a pitcher reaches the high slot by tilting the trunk toward the glove side, Solomito and colleagues (2015) found about 0.5 m/s more ball speed but 3.7 N·m more varus per 10 degrees of tilt [31]. No kinetic data exist on screwballs or on this pitch, so introduce it inside existing pitch counts and build capacity before volume.

For learning, define the target by outcome (spin readout, tilt and movement direction) and use ball-flight data as transition feedback, the constraints-led approach Rob Gray describes, rather than prescribing joint angles [32].

7. How to test it

• Use a pitching machine or athletes, and film every throw from behind at 1,000 frames per second or more to confirm that the seam sits on the glove-side pole.
• Track spin and movement with Rapsodo, TrackMan or Hawk-Eye, and compare the spin-based axis with the axis implied by the movement. Their difference, axis deviation, isolates the seam force (Nathan 2023; Driveline 2020) [33][34].
• Throw at least 30 pitches with the seam on the pole and 30 with the same curve and the seam off the pole, because seam effects vary shot to shot [12].
• Judge shape over dozens of repetitions and value over hundreds of pitches [35].

The prediction is that both versions will show matching spin rate, tilt and efficiency, while the pole-seam version moves 3 to 6 inches farther to the arm side, with its movement direction deviating from the spin-based direction. Testing at two efficiencies, about 80% and above 90%, would check the pole-offset argument in Section 2.3. If the pole-seam version shows no consistent arm-side shift, the seam half of the design fails and only the tilt remains.

Know your slot first

See which pitch fits your arm angle

Pitch AI's Pitch Lab estimates your arm angle and maps your movement profile from TrackMan or Rapsodo CSV imports, then recommends pitch designs that fit your slot.

8. Limitations and open questions

• Pure extrapolation. No real pitch sits within two distance units, so the shape model's predictions are anchored elsewhere.
• Magnitude at game spin. The Scuffball was measured at 1,100 to 1,200 rpm and full efficiency. A curveball at 2,650 rpm and 80% efficiency may behave differently, and Smith reported that one side of the Scuffball worked less well than the other, possibly because of alignment [2].
• Seam height and ball lot. Utah State found about 20% more seam-induced break for seams about 35% taller [36], so the effect will vary with the ball.
• Lift calibration. The model needed a 0.87 scaling for curveball lift, which adds uncertainty to the drop.
• Command and hitters. A high-slot curve with a precise seam position may be harder to command, and hitter responses to a shape with no precedent are unknown.

Frequently asked questions

Start from your own data

Build your next pitch in Pitch Lab

Pitch AI's Pitch Lab turns your pitch data into a movement profile and pitch design recommendations, with grip library videos for the pitches you already throw.

References

1. Baseball Savant, MLB Advanced Media. Statcast search exports grouped by pitcher and pitch type, plus the spin-direction and active-spin leaderboards, 2025 and 2026 regular seasons (2026 season to date). Retrieved October 2026 and analysed by the author (5,102 pitcher-pitch-seasons with at least 50 pitches). https://baseballsavant.mlb.com/statcast_search
2. Smith BL (2020). How to legally scuff a baseball (Post 45). Baseball Aerodynamics, Utah State University, January 30, 2020. https://www.baseballaero.com/2020/01/30/how-to-legally-scuff-a-baseball-post-45/
3. Osnacz A (2026). Pitch AI new-pitch trajectory and shape models: RK4 integration of Nathan's pitched-ball coefficients calibrated to 2025 Baseball Savant pitcher averages, and a gradient-boosting shape model of whiff% and xwOBA (5-fold cross-validation grouped by pitcher; whiff R² 0.56, RMSE 6.6 points; xwOBA R² 0.24, RMSE 0.058). Unpublished analysis, October 2026. Method summary in New pitch types from first principles.
4. Adair RK. The Physics of Baseball, 3rd edition. HarperCollins (Harper Perennial; 2015 ebook printing). Chapter 4, Pitching.
5. Moore J (2015). Prospect Profile: Brent Honeywell. Baseball Prospectus, August 13, 2015. https://www.baseballprospectus.com/prospects/article/27209/prospect-profile-brent-honeywell
6. Laurila D (2012). Q&A: Hector Santiago, Screwball Artist. FanGraphs, November 16, 2012. https://blogs.fangraphs.com/qa-hector-santiago-screwball-artist
7. Sawchik T (2025). The search for missing pitches: Has pitch design given the game all the shapes it needs, or are we at the end of the beginning? Driveline Baseball, September 22, 2025. https://www.drivelinebaseball.com/2025/09/the-search-for-missing-pitches-has-pitch-design-given-the-game-all-the-shapes-it-needs-or-are-we-at-the-end-of-the-beginning/
8. Nathan AM (2008). The effect of spin on the flight of a baseball. American Journal of Physics 76(2), 119-124. https://doi.org/10.1119/1.2805242
9. Anderson JD, Cadou CP (2023). Fundamentals of Aerodynamics, 7th edition. McGraw Hill. Section 6.6, The flow over a sphere, the real case, pp. 513-515 (Figures 6.9 and 6.10).
10. Smith AW, Smith BL (2021). Using baseball seams to alter a pitch direction: The seam shifted wake. Proceedings of the Institution of Mechanical Engineers, Part P: Journal of Sports Engineering and Technology 235(1), 21-28. https://doi.org/10.1177/1754337120961609
11. Smith BL (2020). Cliff notes seam shifted wake (Post 65). Baseball Aerodynamics, Utah State University, December 23, 2020. https://www.baseballaero.com/2020/12/23/cliff-notes-seam-shifted-wake-post-65/
12. Smith BL, Garrett J, Dufour P, Francis E (2024). How seams alter boundary layer separation points on baseballs. Experiments in Fluids 65, 27. https://doi.org/10.1007/s00348-023-03751-3
13. Smith BL (2020). The state of baseball aerodynamics (Post 47). Baseball Aerodynamics, Utah State University, February 10, 2020. https://www.baseballaero.com/2020/02/10/the-state-of-baseball-aerodynamics-post-47/
14. Smith BL (2020). Scuffed baseballs (Post 44). Baseball Aerodynamics, Utah State University, January 17, 2020. https://www.baseballaero.com/2020/01/17/scuffed-baseballs-post-44/
15. Nathan AM (2020). The 3D Spin Axis: An Addendum. University of Illinois, August 30, 2020. https://baseball.physics.illinois.edu/trackman/SpinAxisAddendum.pdf
16. Smith BL (2022). New SSW results: baseballs these days don't know who they want to be, part 1 of 5 (Post 72). Baseball Aerodynamics, Utah State University, January 11, 2022. https://www.baseballaero.com/2022/01/11/new-ssw-results-baseballs-these-days-dont-know-who-they-want-to-be-part-1-of-5-post-72/
17. Nathan AM (2026). Trajectory Calculator, new 3D version (updated June 26, 2026). University of Illinois. https://baseball.physics.illinois.edu/trajectory-calculator-new3D.html
18. Long J, Judge J, Pavlidis H (2017). Introducing Pitch Tunnels. Baseball Prospectus, January 24, 2017. https://www.baseballprospectus.com/news/article/31030/prospectus-feature-introducing-pitch-tunnels/
19. Bahill AT, Baldwin DG (2004). The rising fastball and other perceptual illusions of batters. In Hung GK, Pallis JM (eds.), Biomedical Engineering Principles in Sports. Kluwer/Springer, chapter 10, pp. 275-276.
20. Langin C (2026). Tatsuya Imai Owns a Pitch Type. Unfiltered Labs. https://unfilteredlabs.com/blog/tatsuya-imai-owns-a-pitch-type
21. Sakurai S, Ikegami Y, Okamoto A, Yabe K, Toyoshima S (1993). A three-dimensional cinematographic analysis of upper limb movement during fastball and curveball baseball pitches. Journal of Applied Biomechanics 9(1), 47-65. https://doi.org/10.1123/jab.9.1.47
22. Shibata S, Kageyama M, Inaba Y, Yoshioka S, Fukashiro S (2022). Kinetic analysis of the wrist and fingers during fastball and curveball pitches. European Journal of Sport Science 22(2), 136-145.
23. Jinji T, Sakurai S, Hirano Y (2011). Factors determining the spin axis of a pitched fastball in baseball. Journal of Sports Sciences 29(7), 761-767. https://doi.org/10.1080/02640414.2011.553963
24. Drake RL, Vogl AW, Mitchell AWM (2022). Gray's Basic Anatomy, 2nd edition. Elsevier. Chapter 7, Upper Limb, pronation and supination pp. 412-413; intrinsic hand muscles p. 430.
25. Delgado Genzor M (2024). Luis Castillo and the Pronator's Triangle. Baseball Prospectus, January 5, 2024. https://baseballprospectus.com/?p=87514
26. Fleisig GS, et al. (2006). Kinetic comparison among the fastball, curveball, change-up, and slider in collegiate baseball pitchers. American Journal of Sports Medicine 34(3), 423-430. https://doi.org/10.1177/0363546505280431
27. Dun S, Loftice J, Fleisig GS, Kingsley D, Andrews JR (2008). A biomechanical comparison of youth baseball pitches: is the curveball potentially harmful? American Journal of Sports Medicine 36(4), 686-692.
28. Nissen CW, et al. (2009). A biomechanical comparison of the fastball and curveball in adolescent baseball pitchers. American Journal of Sports Medicine 37(8), 1492-1498.
29. Hall SJ (2023). Basic Biomechanics, 9th edition. McGraw Hill. Chapter 7, The Biomechanics of the Human Upper Extremity, p. 192.
30. Solomito MJ, et al. (2018). [Forearm supination moment and elbow varus moment in pitchers.] American Journal of Sports Medicine 46(1), 52-57.
31. Solomito MJ, et al. (2015). Lateral trunk lean in pitchers affects both ball velocity and upper extremity joint moments. American Journal of Sports Medicine 43(5), 1235-1240.
32. Gray R (2021). How We Learn to Move: A Revolution in the Way We Coach & Practice Sports Skills. Perception Action Consulting & Education. Chapter 7 (constraints-led approach and transition feedback), chapter 8 (differential learning), chapter 14 (representative practice).
33. Nathan AM (2023). What Is The Hawkeye Spin Data Teaching Us? III: Searching for the Seam-Shifted Wake. University of Illinois, March 1, 2023. https://baseball.physics.illinois.edu/Hawkeye-4S-2S.pdf
34. Driveline Baseball (Aucoin D, Caravan A, eds.) (2020). More than what it seams: An introduction to seam-shifted wakes and their effect on sinkers. Driveline Baseball, November 1, 2020. https://www.drivelinebaseball.com/2020/11/more-than-what-it-seams-an-introduction-to-seam-shifted-wakes-and-their-effect-on-sinkers/
35. Premier Pitching Performance (2026). The Pitch Design Phase in Baseball: Developmental Fit, Workload-Informed Timing, Biomechanical Determinants of Effectiveness, and Evidence-Based Evaluation Thresholds. Premier Pitching Performance blog, February 2, 2026. 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
36. Smith BL (2020). A zero-gyro seam shifted wake pitch, the Looper (Post 49). Baseball Aerodynamics, Utah State University, February 17, 2020. https://www.baseballaero.com/2020/02/17/a-zero-gyro-seam-shifted-wake-pitch-the-looper-post-49/

Related reading

• New pitch types from first principlesThe hub. Methods, the gap map and all six proposals.
• The pole-seam riserA low-slot four-seam that rides like a higher slot.
• The float cutterA riding cutter in the mid 80s.
• Mirror sweepersOne spin, two breaks, set by seam orientation.
• The bullet fastballA gyro pitch at fastball speed.
• The two-face splitterOne split grip, two lifts.
• How to throw a curveballGrips and cues for the parent pitch.
• Pitch design by arm slotWhy slot decides which shapes come easily.