Caution: Your Starting Pitcher Has Been Flagged

When you get around your friends and family who play in your fantasy baseball league and they immediately tell you how sorry they are for you, it means your fantasy team stinks. Hunter Brown had a really bad game and I won’t complain or write bad stuff about him because I’m sure he’s feeling bad enough about it as is. Our ERAs are bad, let’s leave it at that.

Did you see it coming? Did anyone? Apparently, someone in a Paul Sporer chat smelt a stinker coming just by paying attention to the underlying metrics. Brilliant! Making sit/start decisions with starting pitchers is an area of my game that needs improvement. I want to flag and bench a pitcher not because they’re heading to Atlanta or playing the Dodgers, but because I conducted sound analysis and noticed something wasn’t quite right. In this article, I will introduce PINE (Pitchers In Need of Extraction), a new and very basic model to flag starting pitchers who need to be given a little more scrutiny and possibly removed from your starting lineup, set to riding the pine.

Iteration one begins with computing some simple zero/one flags. The most important aspect to note is that a score of one is bad and a score of zero is good. That way when all of these binary flags are totaled up, pitchers with high scores are flagged. In this case, a score of six is the worst possible score and a score of zero is top-notch. A pitcher is given one point (bad) for each of the following:

• Fastball Stuff+ < 95
• Current O-Swing% is lower than their two-year* average
• Current Z-Contact% is higher than their two-year average
• BB% higher than the league average plus an arbitrary two percentage points
• K% lower than the league average minus an arbitrary two percentage points
• BABIP is lower than the league average minus an arbitrary two-points

*If a player does not have two full years’ worth of data, their current plate discipline metric is compared to whatever they accumulated in the past two years, or in the 2022-2023 seasons.

My hope is that this simple model adds in high level scrutiny to starting pitchers I would otherwise start without thinking too much about. But, any model needs to be tested before it’s deployed and this model makes big assumptions. On April 19th, the model output the table of pitchers below. In order to keep the list relevant, I used the “Roster%” which can be found on our new PlayerRater, and am outputting only pitchers whose roster rate is above 60%. Let’s look at who this simple model flagged on April 19th, five or more times:

Since April 19th, only a few of these pitchers have appeared, here’s a look at how they performed since being flagged:

In the case of Jon Gray, his own team flagged him and moved him from the starting rotation to the bullpen. Yes, he subsequently shoved, but hey, that’s baseball. After Gray who would you take? Sure Musgrove and McKenzie collected wins, but they came with earned runs. My grandmother would have told me to bench Hunter Brown and Bassitt faired so-so.

Iteration one of this model is already screaming for improvements, but had it been applied to Hunter Brown before his April 11th rough outing against the Royals, he would have been flagged with a score of six, and fantasy managers, myself included, could have benched him, saving their ERA and WHIP from plunging into an eternal dark hole that sucks all light and happiness out of existence forever:

Next up, I’ll add fastball velocity detection using a system I created and ran last season. Pitchers will be flagged who display decreased fastball velocity. In addition, I’ll change the plate discipline comparisons so that I’m no longer comparing all of 2024 to years past but instead detecting how Z-Contact% and O-Swing% change month to month. Lastly, a hitter version is in the works and should be introduced sometime next week, but the most challenging aspect of that project is finding a comparable acronym. For now, ride the PINE!