Introducing Lumber+

Yes, another 'plus' model

Photo courtesy of AP/Charlie Riedel

It’s been a while since you’ve heard from me. I come bearing gifts.

In the doldrums of winter, my usual tendency to scour Baseball Savant and FanGraphs led me to an idea — what if I created my own skill model to evaluate hitters?

For years, digging up Stuff+ evaluations for pitchers’ arsenals has become more commonplace among even traditional baseball fans. Hitters, however, are a tougher nut to crack in modeling.

There are more variables (playing time, park factors, lineup position, etc.) There are different types of hitters (the contact-oriented slappy, three-true-outcomes mashers and line-drive artists). Condensing all of that into one rate stat is tricky, but I’ve developed a model and tool that makes it possible — I think.

Lumber+ is my new Statcast-based hitter skill model that’s free to use — just click the image above. Designed to measure talent rather than production, it’s scaled so 100 = league average and each standard deviation roughly = 5.

A few things to know before diving in:

• Lumber+ is not a stat; it won’t tell you what a hitter did in the box score. Rather, it measures underlying skills that drive offensive production over time.

• It’s not a counting stat or projection system. It doesn’t measure or factor age, playing time or position in a lineup.

• While the free player lookup app above includes all players with 50+ plate appearances in 2025, there’s more noise than signal before 200+ plate appearances — try not to get too excited about Carter Jensen registering nearly 114 Lumber+ in his 69 plate appearances last season.

Lumber+ is derived from three sub-models I created:

• Swing+: The physical characteristics of a player’s swing.

• Damage+: Relevant Statcast batted-ball data.

• Plate+: Plate discipline metrics that add balance to the model, though not necessarily predictiveness.

The big question you might (or should) ask — does it work and does it mean anything?

The answer: Yes, or at least I believe so. In backtesting 2023 and 2024 data, Lumber+ proved to be quite predictive of next-season success, using wRC+ as the end target.

• 2023>2024: R² = .315

• 2024>2025: R² = .245

• 2-year avg: R² = .280

Compare that to prior-season wRC+, which predicts next-season wRC+ at R² = .223, and we see that Lumber+ clears that bar in both years of backtesting.

Finally, a quick AI disclosure. I want to make it crystal-clear this is a human-made, human-developed model. All of the underlying statistics, inputs and weights are completely my own work. Claude was used for backtesting the 2023 and 2024 samples to make sure the model was predictive, cross-checking for formula errors and writing the HTML for the free player lookup app. In fact, there were instances I told Claude to shove it — some of its re-weighting suggestions made the model actively worse/less predictive.

I have zero coding knowledge or skills, so the HTML is where AI came in especially handy. I don’t want to pretend I had any hand in creating the app beyond guiding how it should look and what should be included.

Enjoy the free tool, which will be updated sometime in the first half of 2026 once the sample is large enough to trust, and please reach out on X (@RoyalsData) with any comments, questions or feedback.