Thread: Runs scored
View Single Post
  #34  
Old 08-17-2019, 05:24 AM
ChiTownTrojan ChiTownTrojan is offline
WSI Personality
 
Join Date: Dec 2006
Location: Silver Spring, MD
Posts: 1,110
Default

Quote:
Originally Posted by TDog View Post
I think OPS is not at all a predictor for runs.
You say you "think" OPS is not a predictor of runs. I imagine this is an opinion, based on your own experience and (clearly extensive) knowledge of the game, as well as your personal biases against advanced statistics (if that's what we consider OPS). But I don't think it's based on any hard evidence, that looks at all the data rather than some anecdotal cases (like the Harper one) that stick out in your mind.

What if I told you that I just did it? I was able to use OPS as a predictor of runs, and I got pretty amazingly accurate results. I looked at all the seasons played by all MLB teams since 2000, and I predicted runs scored based only on OPS. On average, I ended up being able to predict the season run total within 19 runs (mean absolute error). That's within 2.6% of the average runs scored during that period, which to me is pretty good.

For context, the margin of error was 39 runs for batting average, 30 runs for OBP, 26 runs for SLG, 20 runs for wOBA, and 47 runs for wRC+. I actually have no idea how those latter two stats are calculated and what they highlight about a player's performance, but they were in the dataset I was able to find. My guess is that wRC+ is designed to fill in some of the gaps that a meta-stat like OPS is missing, which would mean that it isn't that great of a predictor on its own but might be better in combination with OPS. I didn't look at that in my analysis though.

By the way, I was also able to predict team wins with a margin of error of 8 wins, using nothing but team OPS. That margin of error drops to 3.6 wins if I factor in team ERA, which is obviously an important component of winning games that OPS doesn't capture.

To me, this is clear and convincing evidence that OPS is a very strong predictor of runs (and winning). It is also exactly this reason that every team in MLB values OPS so highly.

Some other quick responses:

Quote:
OPS is a number built upon two statistics that are organic to the game. They have overlap, but each only tells you so much.
Nobody, especially sabermetrically-inclined people, think any one stat tells you everything there is about a hitter. That's why there are so many damn sabermetric stats.

Quote:
OPS is a vague, lazy statistical construct that may not tell you what you think it does.
Not sure what's lazy about it. It's pretty clear from the name of the stat ("on base plus slugging") what it means and what's going into it. And as pointed above, it's so often cited because it is the best 1-number summary for predicting runs.

Quote:
Every statistic is taken out of context.
This, I certainly agree with.
Reply With Quote