Top 300 ADP SPs - Stuff Investigation
I look at the SPs going in the top 300 picks with an eye on pitch modeling metrics to see who stands out
Everybody loves shiny new toys and simplicity. I think that’s the main driver behind the popularity of metrics like Stuff+. They are:
Modern. When you start using words like “machine learning” and “algorithm”, people start to think you’re doing things smarter and better.
Simple. We all love when we can boil things down to one number. And it’s even better that even the numbering on these metrics is simple! 100 is the average, and the further you get away from 100 the better or worse the number is.
For further explanations, details, and correlations about Stuff+, Location+, and Pitching+, read this piece I wrote last year.
Very important note here is that I get my pitch modeling metrics from a model designed by Drew Haugen. Drew is an incredibly impressive kid, so check him out if you’re into this kind of stuff. Because of that, my numbers will not match what you may find on FanGraphs, which is where the vast majority of people are getting these numbers from.
I can tell you from experience that some of these words are like magic. I could tweet “I really like PITCHER A this year” and that tweet would do okay, but if I said “My AI-driven machine learning model just is popping majorly on PITCHER A for 2024”, I’ll basically be able to retire from all the ad sharing revenue that tweet generates.
Overall I’m fine with the pitch modeling stuff. My investigation in the piece I linked above convinced me to pay attention to those numbers. But I only find it truly useful in small sample sizes.
For example, there are two ways for me to know that Spencer Strider is an elite starting pitcher
His 143 Stuff+ from 2022-2023
His 29.4% K-BB% from 2022-2023
The Stuff+ suggests/predicts great real-life results, but his 29.4 K-BB% guarantees them.
So if I have a bunch of Major League data on the table, I will just look at K-BB% and SwStr% and other things closer to real-life results than the Stuff+.
I am made even more confident in that decision because of the existence of guys like Graham Ashcraft who in 2023 posted a 121 Stuff+ with a pathetic 9.6% K-BB%. There are plenty of ways to generate a high Stuff+ pitch without the actually being any good. Now, Pitching+ takes location into account as well, which cools things substantially on a guy like Ashcraft, but there’s plenty of noise in a stat like that. Ashcraft’s Pitching+ still shows him being 3% better than the league average SP, which just clearly isn’t the case.
So we want to lean toward results when we have enough data, and incorporate pitch modeling metrics in more and more as we get smaller and smaller samples.
In this post, I just want to take a gander at the SPs in the top 300 to see who seems out of line as far as these pitch modeling metrics go, and then figure out why. The point of this will be to locate possible breakout SPs.
You’ll find full data and data visualizations below along with all of my analysis, but you’ll have to be a paid member to get that. Sign up today to get access to everything happening on this blog, check out this page or this Twitter thread to learn more about all of that.