MLB Model Notes
The MLB model has evolved a lot over the years. After a stellar 2014, the high inherent variability of the sport sent us on a bit of a trip. I started chasing CLV too much in 2015, and let it negatively shape my model. Then I started denouncing CLV, and that didn't help either; I was in a little too deep to revert to the 2014 model at that time.
Beginning with the 2018 season, the model now resembles what it was in 2014; but with the addition of things like exit velocity data, which has been a revelation.
The NHL and MLB will duke it out for best profitability in the long-run; consistency driving the NHL, while volume drives MLB. NHL will be preferred by novice bankroll managers, while MLB will favor experienced bankroll managers.
Update 9/13/19:
We are in the throes of a million-to-one streak of negative variance that has been nothing short of incredible. I don't hold it against the model; that's just baseball being baseball, especially with the new super-juiced "variance balls" they are using this year.
I made the following minor adjustments, though, because it is hard to endure a run like this lying down:
-
Increased the importance of exit velocity for hitters and pitchers.
-
Increased the weighting factor of the current season exit velocity vs. prior season.
-
Increased the weighting factor of realized results vs. projections because exit velocity numbers are less noisy, and now they are more pronounced in my numbers (see adjustment #1).
-
Decreased recent form adjustments for pitchers and increased recent form adjustments for hitters.
Update 9/22/19:
I addressed a platooning issue that inadvertently made declining hitters better, and improving hitters worse due to platoon figures mistakenly anchored to long-term realized results, diluting the core metrics we use to measure hitters.