Fran losing multiple players!
- Thread starter tomar1
- Start date
This is a fair point. The Torvik model has some wacky things in it right now.
I'll be curious to see how it is doing once the dust has settled and rosters are more-or-less complete.
I think @ChrisMWilliams should have Bart on a podcast once the dust has settled to talk about his site, projections and what he sees from the algo on ISU next year. That's content I'd pay for
Keegan Murray is an elite college player- I really thing they should build around him.
I wanted to second this wholeheartedly.
@ChrisMWilliams, a podcast or interview with Torvik or Pomeroy on ISU and their craft generally would be a fascinating sort of thing to do in the offseason. They don't seem too hard to reach on Twitter.
https://twitter.com/totally_t_bomb?lang=en
https://twitter.com/kenpomeroy?ref_src=twsrc^google|twcamp^serp|twgr^author
The obsession with KenPom and especially Torvik is odd to me. I think KP is good for judging resumes, but the way it's been substituted for cold hard facts about conferences, teams, rosters and individual ability is silly. They are individual tools for assessing/measuring/projecting teams, not the end all be all of college basketball analysis.
Kenpom isn't useful until like 15+ games in either.
Not sure this is really anybody's position.
BT and KP are tools -- ones that might lack some nuance, as you point out, but ones that deal with discreet and observable production and numbers and not the personal feelings of somebody doing an eye test.
I think those tools are better at predictions than you give them credit for, too.
Gonzaga and Baylor were 1-2 on the two of them all last season. The results bore that out.
Human intuition and statistical analysis are at their best when they work together.
And, when the two diverge from each other, figuring out why that is can be extremely revealing.
Could be the computers are missing something and therefore we need to improve that model.
Or it could be the humans are biased. Either way, you're learning something about the sport.
Yes Ken Pom needs that 10-15 game calibration, this past year it did not occur, and I believe they way overrated the Big 10 as a result.
Bart Torvik always uses something like at least 15 games' worth of data.
How do you do that when a season is brand new...?
Use the last 15, 14, 13, and etc. games of the previous season until then.
These rankings aren't really much good until at least mid-January.
Agreed. BT and KP are models. They aren’t going to account properly for some things that a human analyzing the situation would account for better.
That said the human models (polls etc) are about as equally flawed.
Getting butt hurt that a computer model ranked your team low is kinda silly
Also if you think it’s that bad post your own model/ predictions and see if it performs better than BT or KP vs the spread
This gif is the very reason they need to hurry up and bring those fancy moving pictures from Harry Potter to reality. I'd put this on my office wall on constant loop above most anything else.
Is there an NFT of this yet? That's one the I'd pay for just to say I own it.