Luke Weaver's Inflated K/9

on

Fantasy baseball is my hobby, statistics & analytics is my day job. Over the last couple of days I've been doing some exploratory statistical analysis with 2017 SP results. It began when Eric Cross mentioned on twitter that Luke Weaver could be a top-30 SP in 2018:
I hadn't taken a big look at Weaver before seeing this tweet, but remembered is K/9 prior to getting called up wasn't all that great. In fact, at AAA in 2017 he had a 8.8 K/9 in 77 IP. So where did his improvement to 10.7 K/9 in 60 MLB IP come from? The first places I looked was swinging strike rate (SwStr%) and first pitch strike rate (F-Strike%). Surprisingly, both metrics didn't support the above average K/9 at all. SwStr% and F-Strike% were basically league average for SPs at 9.6% and 59.9% respectively.

 Knowing that SwStr% is a decent predictor of K/9 I decided to do a little research. I pulled data from all 2017 SP with at least 50 IP, which gives us 189 pitchers. A basic linear regression, predicting K/9 from SwStr% resulted an R-squared of 0.65 indicating a moderate linear relationship. Visually that lines up, as you can see how generally the higher the SwStr%, the higher the K/9.

Each point on the graph indicates a different player's combination of SwStr% & K/9 from 2017. With the regression equation, we can get a K/9 projection of each player. Basically, given their SwStr%, where do they end up on the blue line? Looking at the graph, the points above the blue line show the players that performed better than expected given SwStr%. The points below had a lower K/9 than expected. Obviously we don't live in a world where SwStr% perfectly correlates with K/9. So points deviating from the line are expected. What we care about are the points farthest from the line. The"farthest" is measured by residuals, which are simply the difference between a player's actual K/9 and their projected K/9 given SwStr%. The larger the residual, the more a player's K/9 deviated from the projected amount.

 So let's get back to Luke Weaver. How large was the difference between his actual K/9 and expected? Turns out, it was the largest of all 189 SP in 2017, 3.1 additional K/9 than expected with a 9.6% SwStr%.
Adding prediction interval bands to our graph, you'll notice that Weaver and three other pitchers fall outside the expected observed K/9 amounts above the band. With the prediction band, we are 95% certain that an observation will fall within the interval. So we'd consider Luke Weaver, Jose Quintana, Trevor Bauer and Nick Pivetta to be abnormal observations.

 What does this say about Luke Weaver in 2018? I'm taking his 10.7 K/9 from 2017 with a grain of salt and acknowledging that his 2017 AAA K/9 of 8.8 and his 2018 K/9 Steamer projection of 9.1 are better indicators of his true K/9 skills.  Could he still end up being a top 30 pitcher in 2018? Sure, but don't expect that K/9 to be the driving factor.
---
Hello, I'm Smada and I'm just another guy. You can find me on twitter dot com @smada_bb where I have fantasy baseball related tweets and pretend to know what I'm talking about.

Comments

Post a Comment