Completed model
It’s all completed. After hours, days, weeks of work — I have code that represents a feasible model. And it’s incredibly satisfying. It might be bad, it might be good. At this point, I’m not too worried about that. It is done, it works, and it seems to be at least fairly accurate.
When I was previously making posts on it I was working purely on pitching. I was trying to work out all the nuances like dealing with relievers, injuries, etc. I was tweaking it here and there until it felt just right. I then moved on to hitters. Given what I had already experienced earlier with pitchers I think that the hitters portion came to together quicker and with fewer hiccups.
I was able to same the similar plots as previously. Let’s take a look at them:
A few noticeable outliers in the residual plot, but pretty clumped together towards the middle of the residual plot. The regression plot is certainly headed in the right direction. The peaks, while larger, are in the right spots for the histogram. All lines are right up close together for 60%-65% of the ECDF graph. And lastly with the 63,123 observations I’m returning an Rsquared of .484 as well as 74.65% of the predicted values being within +/- 1.0 WAR.
And it all shows that there is still work to be done. I need to check out the data I have and figure out where the disconnect is in order to correct it without overly correcting the data.
I have to keep reminding myself that I don’t need or want it to be perfect. Once you’ve done that, it’s no longer a prediction.
So with that, I want to take a look at my top 25 predicted WAR totals for 2022. They seem to best encompass my feelings and thoughts on my results as a whole.
Player Predicted_WAR
Jacob deGrom 8.4
Trea Turner 8.2
Bryce Harper 7.8
Jose Ramirez 7.5
Corbin Burnes 7.5
Marcus Semien 7.2
Dylan Cease 7.1
Alcides Escobar 7.1
Zack Wheeler 6.9
Carlos Correa 6.9
Walker Buehler 6.5
Vladimir Guerrero Jr. 6.3
Tyler O'Neill 6.2
Steven Matz 6.2
Juan Soto 6.2
Max Scherzer 6.1
Gerrit Cole 6.1
Nathan Eovaldi 5.9
Fernando Tatis Jr. 5.9
Brandon Crawford 5.9
Alex Cobb 5.9
Chris Flexen 5.8
Carlos Rodon 5.6
Will Smith 5.3
Justin Turner 5.3A quick scanning of this list and you could say that a good 80% of this makes at least some sense. Some you could have arguments against (Eovaldi, Smith, Crawford). But it doesn’t take long to see where it likely will get really wrong, really quickly.
I can’t imagine that Escobar will post a 7.1 WAR season when his career WAR is 13.4. He had a decent year last year, posting 1.7 WAR in 349 PA. But let’s be real — no way does he suddenly become MVP caliber at age 35. Fitting in there is Cobb, exactly same deal, though he will “only” be 34.
Steven Matz also had a decent year last year. 2.8 WAR in 29 starts last year. I imagine the odds of him reaching even 4.0 WAR are quite slim — let alone 6.2 WAR. Entering his age 31 season it seems like Matz has a slightly better chance at hitting his target compared to Escobar.
One that I’m really hesitant on believing, but could be more correct than I thought, and I’ve posted about him before, Marcus Semien. 7.2 WAR is very attainable for him. Last two full seasons have been 7.6 and 6.6, so there is no reason for me to think it will not be close. I just… have some weird distrust in his ability.
I also had another issue of not realizing how good of seasons some players had last year. e.g. Chris Flexen, Dylan Cease, Tyler O’Neill. I saw them on the list and really questioned what I had done wrong here. Then I looked up their stats and they were all really good. I can’t say for sure if they will continue to be that good — but at least the predictions are justified for that.
Last up I need to address the elephant in the room. Where is Mike Trout?! Well. I know where the issue is. And this is the next issue that requires some tinkering.
