I encourage you to read this post on Substack for better visibility of the graphics housing the data above each name (click here).
A ton of graduations have pushed a crop of new arms into the newly expanded top 50. Below you’ll find my fresh thoughts with a bunch of data and other goodies along the way. I hope the blurbs are some of the most informative you’ll read.
This is now the seventh edition of my list. I did a brief look back at my preseason 2024 list to see how I did. My big takeaway is that more than 50% of the names had some kind of major arm injury. My approach to injury risk is becoming very simple: discount those who are presently injured and coming off injury until they show prior form. Otherwise, don’t apply discounts to arms that are healthy just because I think they might get injured. They probably will, but so will everybody else.
In case you missed prior lists, here are the links: 2024 preseason, 2024 midseason, 2025 preseason, 2025 midseason, and 2026 preseason.
Graduates
Nolan McLean (NYM), Payton Tolle (BOS), Bubba Chandler (PIT), Trey Yesavage (TOR), Andrew Painter (PHI), Brandon Sproat (MIL), Connelly Early (BOS), Parker Messick (CLE), Didier Fuentes (ATL), Logan Henderson (MIL), Connor Prielipp (MIN)
Presently Injured … but would be inside my t40
Thomas White (MIA - 55 FV), Travis Sykora (WSH - 55FV), Jarlin Susana (WSH - 50FV), Alejandro Rosario (WSH - 50FV), Tekoah Roby (STL - 50FV), Jake Bloss (TOR - 50FV), Juan Valera (BOS - 50FV), Robby Snelling (MIA - 50FV)
Process
My approach to this list was layered. On first glance, I tried to identify fastball quality and release traits. I often did this through the lens of spin efficiency, release height, arm angle (which is a mini-model with +/- 5° error), and extension. Once I got an understanding of those variables, I moved to secondaries, spin rates, and thought a lot about what the pitcher does well. Does he throw breaking balls hard with drop? Is he going to have trouble creating depth on harder breaking ball shapes? How will his tendencies influence repertoire development? For changeup evaluation, I relied heavily on zone and chase rates with less focus on shape.
I then checked my thoughts against my stuff model to square anything I missed and provide new insights. After that came watching the player and incorporating thoughts from past times I’ve watched starts. Finally, there was a round of passing this list around to contacts inside and outside of organizations to see if anything looked dramatically out of place.
I won’t shy away from this list being very data-driven, perhaps to a fault, but that’s what separates it from others in the industry, all of whom do amazing work (Baseball America, MLB Pipeline, FanGraphs, etc).
Tiers
Players are broken into tiers, which you will see on both the cheat sheet and within the blurbs below. These tiers are heavily based off FanGraphs 20-80 scale, which is a good baseline for anybody who doesn’t think they have a grasp on how good it is for a pitcher to amass ~18 WAR in their first 6 seasons. Pitchers are placed into tiers based on what their median outcome is. That's why the projections might appear low.
I split the 50-FV tier into two halves: A and B. A tier has a median outcome that skews just above the ~12 WAR projection. B tier skews slightly below. The tier was too big to keep as one chunk.
Stuff+
I built my first Stuff+ model for this piece. I’m not entirely sure how to feel about the results, but a team analyst once told me: if a model confirms all your beliefs, it’s not useful. Seeing some guys lower than I thought was a good way to check my beliefs and consider alternate perspectives.
A key utility of this model was identifying pitcher splits and which shapes work against which handedness of hitter. Even if the scores weren’t perfectly portable to MLB, I think it’s accurate in directionally pointing out which shapes will work in a more platoon-neutral role versus those that are handedness-dependent. I therefore split this into four models: LHP vs RHH, LHP vs LHH, RHP vs RHH, and RHP vs LHH.
The training data used for the model was 2024-2026 MLB shapes, and the goal was to project MLB run value. 10 points is one standard deviation. The average MiLB Stuff+ is 96. It’s a pretty basic model compared to many of those out there. I hope to improve it over time. It’s heavily reliant on my imputed arm angle, which is a separate model, velocity, and the ball’s movement.
The value in a stuff model like this is that it stabilizes quickly, but once you get results against major-league hitters, layering in those results becomes more powerful than Stuff+ alone. I made adjustments for the MLB ball if a pitcher was below Triple-A. I also corrected for altitude, which applies to a few Texas League parks (Double-A), as well as the Pacific Coast League (Triple-A). For this reason, some of the grades based on the shapes you see on plots might be lower than you think, especially for big, efficient fastballs.