Christopher Weber retweetledi
Hey y'all, I need to chat about our new @PitcherListPLV account.
First of all, we have no idea how this happened. We expected to do a small launch and ramp it up throughout the spring as we ensured everything was working correctly, and in no world did we anticipate climbing from 800 followers to 100k+ in 24 hours. We've reached out to colleagues trying to grasp it, and I don't think anyone in their right mind actually believes I'd purchase bots to pump the numbers (if I actually wanted to, we'd do it for our YouTube channel!).
While inspecting some of the cards during the day, we saw many complaining about the model's grades, and we 100% agree. Something was wrong. How on Earth could Tolle have such a low stuff grade? Triple F for Bergert's outing?!
Let me explain what happened and a little more about what Pitcher List Stats is and our PLV model.
We began developing PLV (Pitch Level Value) in 2019, after four years of work, and launched it in 2023. I personally worked with some brilliant people to get it to the finish line, including Colin Charles, Jeffrey Nicholas, Zach Dobroff, Cameron Grove of Pitching Bot, and now the head of our analytics department Kyle Bland. You can read the award-winning article from 2023 here: pitcherlist.com/what-is-plv-an…
As outlined there, this model is not something whipped up by AI over the last month. It takes into account dozens of factors, from release height, arm angle, movement, velocity, spin rates, extension, and all the good stuff. We've made adjustments here and there over the years and will continue to improve it indefinitely. It's a foundational metric that allows us to assess player skills on a different level, including crafting our Process+ metric for hitters.
"But then why was it so obviously wrong yesterday?" Well, that has to do with a few factors, and Kyle does a fantastic job outlining what went wrong in the tweet below. In short, some overfitting, some small sample problems, and some spring training being spring training.
The next tweet you'll see from @PitcherlistPLV will showcase the fixed model vs. some of the cards you saw yesterday. We're also expediting an article on the Pitcher List site from Kyle outlining all the factors that go into determining the grades you see on these cards, making it the clear reference point for everyone to use these cards to the best of their ability. Remember, we didn't expect for this account to get so much attention so quickly!
Once we have our article live on our website, we will add a link inside these cards so everyone can understand the massive undertaking of making them.
We spent weeks iterating on this design and determining what should go where (no AI involved), and we're incredibly thankful to have so many provide feedback on what metrics they want to see, what needs more clarity, and special requests. Please continue to do so, respectfully, of course.
We've already made some changes that you'll see later today, such as renaming "Start" and "Models" to "Results" and "Skills", as they better represent what those grades are for. We love the idea of a "Start" grade to quickly grasp if a pitcher was good or poor, but we didn't make clear this grade was our version of a Game Score, and not a grade of the pitcher's stuff or PLV. We still like the idea of an all-encompassing "Skills & Results" number (and one for a SP vs. RP) and we'll keep toying with the idea.
And yes, those "Start/Results" grades are low because we intended these cards to be featured mainly during the regular season, when pitchers are properly stretched out and aren't leaving after a few frames. Another artifact of our unexpected attention during the start of spring training.
Lastly, while these cards are currently only for pitchers, we've already begun the process of creating cards for hitters - it's a little trickier given the small sample of a single game.
I hope this post has cleared the air. We're still blown away at the support for our new account and we're excited to share our love for the game with all of you.
Aces gonna ace,
-Nick
Kyle Bland@blandalytics
Hey everyone! Let's talk about our new pitcher cards at @PitcherListPLV. First and foremost, as the Pitcher List Director of Analytics, I want to apologize. The Stuff model isn't accurate to what we want it to be, nor to what our intuition is.
English
