Ever since we acquired Robbie Ray I have been thinking he looks pretty good. Like "no way we acquired this guy for Travis Bergen" good. His numbers haven't been amazing since we got him (4.38 ERA, 5.55 FIP, 4.91 xFIP) but the living room scout in me sees a ton of potential in him. Just a note to start off, however, which I didn't realize at first: his ERA as a Jay has actually been below league average which is 4.52 this year.
His command isn't really that bad and his stuff is electric. I believe that this guy is different from Roark, Anderson, and Walker. My hypothesis is that he looks like a great pitcher who will substantially lower that 4.38 Blue Jays ERA moving forward.
But the reality is we did acquire him for Bergen, so the industry did not value him. And clearly his numbers were terrible (7.84 ERA, 7.29 FIP, 6.47 xFIP) - he wasn't pitching well at all prior to the trade, so what gives? What am I seeing that wasn't there with Arizona?
I went looking for something that has changed with Robbie since he came over to the Jays and it didn't take me long. Turns out that at the start of this season, he decided to stop throwing like David Price and start throwing like a 15 year old girl for some reason:
15 year old girl:
(watch the catcher)
But how can we know if he is for real? It is just three appearances so far (today's start was not included), that is too small of a sample size right? Wrong!
It has been 7 years since I first learned about the wonderful concept of pitch based ERA estimators such as TIPS. This beautiful stat has just 3 inputs: o-swing%, swStr%, and foul%. It doesn't use IP as an input like FIP or xFIP (IP are obviously not independent of defense unless you get every out in the inning via strikeout). Best of all, in a sample of just a few hundred pitches you have a stat that predicts future ERA better than any of the other major players including more complex stats like SIERA.
I was sure that this new idea would revolutionize pitching analysis and discussion, yet here we are 7 years later and we as baseball fans really haven't moved very far past FIP (just look at the start of my post). There was a good post here about the shortcomings of DIPS and I think the clear response should be a shift towards using better ERA estimators in our discussions.
So, using TIPS I will predict Robbie Ray's future success using only the 238 pitches he has thrown so far as a Blue Jay. Spoiler: he is probably gonna be really good.
Warning: Math - feel free to skip to the next dotted line for the conclusion
----------------------------------------------------------------------
Is 238 pitches really enough? In this post it says around 250 pitches is a good sample. I can't find any other references to a stabilization point, just several comments that mention it does stabilize very fast. In many blog posts regarding TIPS, its creator (our boy jFaS) would set the cutoff to 50 pitches. It would be interesting to dive into the math on when TIPS really stabilizes, but for now I will say that since jFaS said around 250 pitches is a good sample we will just trust him that 238 is close enough.
Unfortunately TIPS is kind of annoying to calculate because by definition one of the components is not searchable on fangraphs:
I used Statcast data rather than pitchFX data for classifying Robbie Ray's pitches and for calculating the league average foul% since I couldn't find a reliable source of pitchFX data. jFaS also did this at times and comments on this practice here, mentioning that it will result in slightly different o-swing% values because they calculate the size of the strike zone differently. So the first thing I wanted to do was make sure that I was able to replicate the original results using my method and data sources. To do this I just needed to calculate the TIPS constant for 2013, and make sure it is the same as jFaS's original value of 2.57
From the original blog post:
0.696 o-looking league average from fangraphs
0.094 SwStr% league average from fangraphs
3.87 league average era from fangraphs
But how to calculate foul%? I have different search terms available on Statcast, I can't get Contacts easily but Fouls and BIP are readily searchable. So using the equation from the original blog post and a bit of algebra:
therefore: CONTACTS - BIP = FOULS
rearranging: FOULS + BIP = CONTACTS
therefore: Foul% = Foul/Contacts
substituting:
Foul% = FOULS / FOULS + BIP
Easy enough from here to run a Statcast search on all pitchers and plug the results into Excel to get total Balls in Play and Total Fouls for all pitchers in 2013
total fouls = 119749
total BIP = 131848
foul% = 0.475955596
This allows us to calculate the 2013 TIPS constant as 2.547086859, which I say is damn near close enough to the original value of 2.57
Moving on now, since we know the method is sound:
Using the same logic I calculated the 2020 TIPS constant as 3.499557857, which is way higher than 2013 largely because of the much higher league average ERA
0.696 o-looking league average from fangraphs
0.113 SwStr% league average from fangraphs
0.500376637 foul% league average calculated from Statcast
4.52 league average era from fangraphs
All that's left to do is calculate Robbie Ray's TIPS as a Blue Jay using Statcast search.
use team = Blue Jays; pitchers = Robbie Ray
for fouls use pitch result = foul: 37 total fouls
for BIP use included stats = BIP: 36 total BIP
Therefore foul% = 37/(36+37)= 0.506849315
for o-looking%, use gameday zones 11 12 13 14 and pitch result = ball, ball in dirt, called strike or hbp: 0.3850
for swinging strike% just use pitch result = swinging strike and swinging strike (blocked): 0.131221719
Final calculation:
TIPS = 6.5*O-Looking% - 9.75*SwStr% - 4.8*Foul% + C
= 2.5025 - 1.279411765 - 2.432876712 + 3.50
----------------------------------------------------------------------
As a Blue Jay so far, Robbie Ray's TIPS = 2.29
In conclusion:
Ray started the year off by changing his delivery and had some terrible results. He's changed it back to one that actually makes sense, the one he used when he was an Ace. Given that this is such a substantial change, we should limit ourselves to the sample of games where he has been using the new delivery in order to predict his future success.
Using the best statistics we have available to analyze the sample size following the delivery change, we can predict Robbie Ray's ERA moving forward to be approximately 2.29
Atkins has most likely found our #2 Starter for the playoffs.