A Survey of FanGraphs Preseason Projected Win Totals vs Actual Results 2016-23

[formerlybis]

2 hours ago, Cool Hand Lucroy said:

KenPom did this relatively recently with the Badgers. They put up a couple of middling years in that Ryan/Gard transition (the Davison years), and yet the algo had them rated very highly. I think he tweaked things to bump them down a bit because he thought something about Wisconsin's per possession strategy was gaming the system.

Might not be a bad idea to adjust MLB projection in the same way, although I think you might start to run into funkier and less quantifiable variables given the nature of the sport. Good thought.

Oh you hit another favorite topic (maybe you knew that, as I do a regular recap of Badger hoops games by the analytics on one of their sites). I won't derail things too much here, only to say that KenPom and others always have a catch-all category called "luck" (or "F.U.N." on Torvik) which captures either randomness or variation that is (currently) unexplained. I think those guys are always searching for more explanatory data (and I would guess upwards of 95% of teams now use some form of PPP as a strategy from high school on up, too, so that's no longer unique). 

In baseball, especially before the season starts, I just can't fathom how you'd account for something like the acquisitions of Priester and Vaughn, and their subsequent effectiveness beyond their historical performance.  I just don't think there *is* data. Maybe you can give some adjuster to the Dodgers and Yankees always being able to bolster themselves, but not MKE.

[Cool Hand Lucroy]

54 minutes ago, formerlybis said:

Oh you hit another favorite topic (maybe you knew that, as I do a regular recap of Badger hoops games by the analytics on one of their sites). I won't derail things too much here, only to say that KenPom and others always have a catch-all category called "luck" (or "F.U.N." on Torvik) which captures either randomness or variation that is (currently) unexplained. I think those guys are always searching for more explanatory data (and I would guess upwards of 95% of teams now use some form of PPP as a strategy from high school on up, too, so that's no longer unique). 

In baseball, especially before the season starts, I just can't fathom how you'd account for something like the acquisitions of Priester and Vaughn, and their subsequent effectiveness beyond their historical performance.  I just don't think there *is* data. Maybe you can give some adjuster to the Dodgers and Yankees always being able to bolster themselves, but not MKE.

This is great. I'd love a link to your Badger hoops analytics recap. Always fascinated by that stuff, and by the "luck" category on KenPom/Torvik. 

[Underachiever]

7 hours ago, Cool Hand Lucroy said:

This is great. I'd love a link to your Badger hoops analytics recap. Always fascinated by that stuff, and by the "luck" category on KenPom/Torvik. 

I read them all. They are excellent. Buckyville is the site.

[markedman5]

9 hours ago, Cool Hand Lucroy said:

This is great. I'd love a link to your Badger hoops analytics recap. Always fascinated by that stuff, and by the "luck" category on KenPom/Torvik. 

I’m pretty sure the luck category is just the difference between actual win loss pct and expected WL % based on his adjusted points per possession system.

Luck seems like a poor term to use……..but not a big deal……

[formerlybis]

14 hours ago, markedman5 said:

I’m pretty sure the luck category is just the difference between actual win loss pct and expected WL % based on his adjusted points per possession system.

Luck seems like a poor term to use……..but not a big deal……

Yeah, I think that's basically it (and Torvik's FUN is "fortune unexplained by numbers or failure unexplained by numbers") and is very similar.

I'll leave it at that - all models are wrong, but some are useful.

[Oxy]

My theory is that the Brewers are far better at understanding all the nuances of situational baseball and sequencing game relevant tactics than most other teams.  Everybody uses analytics these days to some degree, and historically (since the OBP/moneyball revolution) the majority of these numbers are used for large samples and "average" situations in order to make projections over the course of 162 games--and it assumes that everything will "average" out. 

Winning a ballgame is NOT a sequence of average situations, however.  Accumulating wins is the goal, NOT accumulating the best averages for average situations.

[Samurai Bucky]

20 hours ago, Underachiever said:

I read them all. They are excellent. Buckyville is the site.

Be afraid... be very afraid 😉