I’m excited to be presenting at the 2026 SABR Analytics Conference in Phoenix later this month, because I’m bringing something I think hitting analytics has needed for a long time and still does not have in a clean, structural way. I’m introducing Causal Analytics, built around Collision Geometry Deviation (CGD) and CGD+.

I created Causal Analytics to solve a specific structural gap in how the sport measures hitters, and the cleanest way to see that gap is to compare how pitching analytics evolved versus hitting. Pitching didn’t outpace hitting because it had better technology. It outpaced hitting because it used technology differently, measuring at the causal source of the event and turning that measurement into a control loop. That’s why pitching analytics looks like engineering. Pitch design starts at release in a stable frame, so teams can tune inputs, iterate with feedback, and tighten execution. Hitting, by contrast, has been evaluated mostly after the collision, where the job becomes explaining outcomes without the collision-context and inferring cause backward from effect.

Here’s what that downstream evaluation looks like in practice. Most modern hitting evaluation still lives downstream. We use descriptive metrics like xwOBA and BABIP to summarize production to predict trends and explain good and bad luck variance. We lean on exit velocity and launch angle, barrels and hard hits to describe contact. That’s useful, but it’s only grading the aftermath. They tell us what happened after the collision, and it routinely fails to answer the question that actually determines whether the result was repeatable: did the hitter solve the collision correctly for the pitch they swung at.

That gap is why baseball keeps tripping over the same contradiction. A hitter can make a perfect geometric decision and still make an out. A hitter can make a wrong geometric decision and still hit a missile. When the model is built around outcomes, execution gets blended with variance, and “luck” becomes the junk drawer for everything the model never measured in the first place.

That is exactly what Causal Analytics is designed to fix. CGD is not trying to be another descriptive metric. It is meant to evaluate the geometric inputs of contact, tied to pitch location, so you can separate directional correctness from outcome noise. In plain language, it is a way to measure whether the swing vector fit the pitch context, instead of letting rare loud outcomes or quiet unlucky outs distort what you think a hitter is actually doing.

Descriptive models are like studying the dashboard after the trade clears. A geometry-first model measures the inputs that created the trade. This is not about playing the stock market and studying the dashboard. It is about controlling the mechanism.

Now, here’s the part that usually gets missed, and to be clear, it’s not a second thesis. In the presentation, I’ll only touch on it briefly, mainly because a stable hitter-relative reference enables automaticity in collision reporting. Without that stability, you can still do the work, but you have to keep compensating for the reference frame instead of letting the measurement speak for itself. If the anchor point moves, the measurement moves with it. If the anchor point is not universal, hitters can still be compared, but the comparison is not clean. Depth and geometry get mixed with setup, and the measurement stops meaning the same thing hitter-to-hitter. That leakage is why comparing swing arcs, comparing collision geometry, building expected geometry for pitch locations, and measuring deviation from it becomes noisy and unstable unless the reference layer is fixed.

It’s the coordinates layer that makes collision measurement comparable across hitters. That’s why I’m also introducing my hitter-centric coordinates alternative. The premise is simple and it is structural. Hitting has been measured against environmental landmarks like the plate, and against shifting anatomical midpoints that describe setup and morphology rather than execution. Pitching did the opposite. Pitch measurement is anchored to an anatomical reference at the causal source, the release point. The rubber is intentionally not the reference frame. At most, it’s a secondary fact you might compare release to, but it is not the origin that defines the pitch.

A universal metric requires a universal origin, and for hitters the origin has to be hitter-centric, not plate-centric. By anchoring measurement to a front foot anatomical tether and standardizing orientation through a Keystone Swing Moment, the framework resolves the instability inherent in existing coordinate systems and makes collision geometry measurable in consistent units.

This work is also supported by a clear IP chain. The patent portfolio is:

• • U.S. Patent No. 10,987,567 (The Anchor): Establishes the foundational prior art for the anatomical origin, the Front Foot, and the orientation framework, the Keystone Swing Moment, removing the instability of plate-relative measurement.
  • U.S. Patent No. 10,994,187 (The Logic): Formalizes the deterministic coordinate transformation algorithms that operationalize the architecture. It serves as the computational bridge defining the equations that convert temporal inputs into precise spatial coordinates, dictating the exact z-depth of contact required for stabilization.
  • U.S. Patent No. 11,596,852 (The Matrix): Establishes the Collision Point Matrix as a multi-point spatial grid of expected contact locations. This transforms the coordinate system into an active guidance and analysis framework for both physical and virtual environments.

Later filings extend the framework further:

• • July 25, 2025 (Patent Pending): Extends the formalized architecture to video-based analysis, ensuring visual diagnostics operate within the hitter-centric frame.
  • December 22, 2025 (Patent Pending): The culmination of the architectural chain. This Computational/Analytical Engine explicitly defines hitter-relative collision geometry formulations, providing the mathematical closure for expected geometry, deviation calculation, and the Causal Engine itself.

That’s what I’m bringing to Phoenix. Pitching development is widely ahead of hitting development, and it isn’t because pitchers got luckier or teams got smarter overnight. Pitch design became engineering because it measures inputs at the source and iterates in context. Hitting still gets graded by aftereffects, or by context-free swing summaries that don’t tell a hitter how to solve the collision for the pitch they swung at. If you’re tired of watching pitching get an engineering stack while hitting gets a report card, you’ll want to be in the room. The goal is straightforward: stop autopsying results and start measuring the mechanism.