Revisiting blended WAR

Assessing Guardians starting pitchers using the average of many WAR calculations

Pitchers, however, are valued very differently by the different systems. FIP is a linear weights based system that treats all balls in play as equally valuable and ignores sequencing. Baseball-Reference starts with runs allowed and works backwards. Baseball Prospectus uses a complex modeling system to attempt to derive the value of individual events while controlling for contextual factors. You have to decide which method is the one you prefer, although looking at each site is the best way to get a complete picture of the player.

Piper Slowinski, March 22, 2012, FanGraphs dot com

It continues to be excellent advice, in particular for pitchers. Going one step further, the addition of some simple and easily accessible statistics can help you infer why the various systems disagree with each other.

It is truly up to you whether or not you want to understand starting pitchers with this level of detail. One lesson of this exercise each time I have run it is that the systems tend to agree. When they do not, the difference is not often drastic.

When it is, someone is having an unusual season, and it is often Aaron Nola.

Before we begin, here is how Baseball Prospectus explains its pitching value modeling: “Unlike other public pitching metrics, DRA- focuses on the pitcher’s expected contribution and uses a mixed-model approach to isolate the work of the pitcher from other factors like defense, park, and quality of opponent.”

Shane Bieber - 27 GS

Traditional: 2.91 ERA, 173.1 IP, 25.9 K%, 5.1 BB%
Batted Ball: 3.54 xERA, .290 BABIP, 0.78 HR/9

4.4 fWAR, 3.3 bWAR, 3.4 WARP — 3.7 3WAR

Nothing drastic to see here, folks. Bieber’s 2.91 ERA is quite a bit better than league average, which is right around 4.00 this season. We can also take a quick glance at his ERA and the expected ERA (xERA) to reveal that the Algorithms believe Bieber is a little bit lucky.

Cal Quantrill - 27 GS

Traditional: 3.50 ERA, 157 IP, 16.6 K%, 6.6 BB%
Batted Ball: 4.43 xERA, .277 BABIP, 1.03 HR/9

1.7 fWAR, 1.6 bWAR, 0.6 WARP — 1.3 3WAR

Boy does Baseball Prospectus not believe Quantrill deserves the outcomes he gets. It’s ... kind of funny? Normally when two of the models are neighbors the third is across the street. Prospectus almost lives on the other side of the tracks, here.

Zach Plesac - 23 GS

Traditional: 4.39 ERA, 127 IP, 18.1 K%, 6.8 BB%
Batted Ball: 5.28 xERA, .285 BABIP, 1.35 HR/9

0.8 fWAR, -0.4 bWAR, 0.4 WARP — 0.26 3WAR

Plesac continues to struggle and the numbers bear it out. It is interesting to me that the fielding-independent model prefers Plesac, however mildly.

Triston McKenzie - 26 GS

Traditional: 3.05 ERA, 165.1 IP, 24.8 K%, 6.7 BB%
Batted Ball: 3.85 xERA, .231 BABIP, 1.25 HR/9

2.6 fWAR, 3.4 bWAR, 2.5 WARP — 2.83 3WAR

It makes a certain amount of sense; FanGraphs’ model leans heavily on fielding-independent events. With a BABIP of .228, you can see where a “gap” might open in fWAR’s assessment. Baseball Reference leans on runs allowed per nine and is more interested in outcomes. I am unsure how, exactly, to interpret Baseball Prospectus’s take, but I suspect it does not believe McKenzie deserves his BABIP. I do.

Aaron Civale - 16 GS

Traditional: 5.40 ERA, 75 IP, 23.1 K%, 6.2 BB%
Batted Ball: 4.29 xERA, .321 BABIP, 1.2 HR/9

1.0 fWAR, -0.9 bWAR, 0.7 WARP — 0.26 3WAR

Now we see a bit more of a chasm emerge. While FanGraphs sees Civale’s 2022 to date as slightly above replacement level, Baseball Reference believes it is just plain bad.

Prospectus bridges the gap but leans toward the fWAR side. It is similar to Plesac’s blend of WARs in that way though much more extreme here. When considering the difference in his ERA and xERA, in addition to the slightly elevated BABIP, it begins to become clear why the models are split so dramatically. Keep in mind, this is a two-win difference in about half the number of innings as his peers.

Said another way, Aaron Civale owned very good strikeout and walk rates, a good home run rate, yet suffers from a tough ERA despite a great defense playing behind him. Some of this may be further explained by a slight dip in GB% which has been replaced by additional line drives. I’m not sure that many runs can get created, though. He might be a sneaky regression candidate. If healthy. When healthy? When healthy.

We could review Konnor Pilkington — 9 GS

but I’m not sure WAR means much in fewer than 70 innings for a SP. It’s a bit of a stretch to use in-season WAR at all; single-season WAR is also used a little bit too definitively for my comfort.

WAR also excludes intangibles. Wins lurk there, too. Not many, I suspect, but I’m actually glad there isn’t a metric or model designed to measure this. As far as I know, anyway.

“Can’t sign him, Billy. This model says he’s an absolute jerk.”
“So? He gets on — oh. Oh wow...”

What can we unpack from this?

Let’s look for some fun patterns.

1. The walk rates of every Guardians starter are remarkably similar

For reference, none other than Corey Kluber leads starters with at least 80 IP with a BB% of 3.1%. The worst in the league sits at 13.2%, and league average tends to be near 8%. I cannot find the value in order to be sure, but it feels like it’s closer to 7.5% or even 7.0% this year. I’m not so sure that this indicates hitters getting worse as pitchers continue to outpace hitters. In any case, Cleveland possesses five starters who limit free passes effectively.

2. Every Guardians starter except Civale outperforms xERA by a significant margin

Nearly a full season of Steven Kwan, Myles Straw, and Andrés Giménez seems to be paying dividends. I will not exclude Amed Rosario, who is playing shortstop well this season. All of those defensive runs saved above average represent very real outcomes that are borne out in pitcher ERAs. Even better, the Guardians already limit baserunners better than the average team, and so an unlikely gapper is even less likely lead to runs. Synergies are neat.

3. All of this pretty much confirms what we already knew

Shane Bieber and Triston McKenzie are pitching very well. Cal Quantrill is fine and gets killer run support, while Civale and Plesac aren’t pitching their best for various reasons. By taking the time to tease some of these resources and statistics apart, however, we can begin to see how they all relate. That can represent very real differences in how these players play the game, and keeping an eye on trends and patterns can occasionally yield some interesting information.

For example, Aaron Civale’s overall stats might not indicate the true talent level at which he has been pitching. Anecdotally, this matches with his occasional lights-out performances this season. Sept. 15 vs. Detroit is a fine example.

Nonetheless: he’s been disappointing this season at delivering results, on the whole, and we already knew that.

Baseball.