24-25 Torvik Projections | The Boneyard

24-25 Torvik Projections

Eh. It's another computer model like Kenpom. Torvik himself probably thinks that seems low but if that's what his model says, there you go. Torvik is a great site though. Definitely worth checking out once he has the full numbers up for 2025. I'm guessing it's hard to adequately account for the roster changes. Tons of production left. Lots of freshmen and transfers, so it's tough to slot them in properly.
 
tried to find the model but don't see much of anything. i'd venture to guess that there's a lot of normalized noise in the usage patterns, and also a bifurcated model choice between recruits and returnees in that recruits are valued by expected value of a given recruit ranking, while returnees are evaluated on past college production.

this means that for a team that is reliant on returning freshmen who did not get much run last season/didn't produce much, they are being evaluated on that sample and not on their recruit rankings. this can be seen by this model giving uconn the highest "talent" but low production and minutes from those returnees.

this is just a case of uconn falling through the cracks of a model designed to approximate the normal college team in terms of progression and development. wouldn't worry too much about it.
 
Poe said:
Garbage in, garbage out
Expand ...
It literally looks like he took a spreadsheet with a whole bunch of data points from past years. Inputted this into the ChatGPT context window. And, asked ChatGPT To come up with a ranking, records, etc.

This, folks, is called AI Hallucinations!
 
B1GEast said:
I have never heard of this guy. Supposedly he influences a lot of writers and when he updated his projections, people made a lot criticism. UConn at number 12.

Expand ...
Odd. If you sort by "Projected Effective Talent", UConn is #1 and in that they're +2.8 over second place. No bigger gap in the entire list.

Add that talent to the best coaching staff in America and you... end up #12 ? Huh ?
 
He ranked us third last year in the preseason, which was one of the highest I saw.

He tweaked things this year he said on Twitter. From reading into his comments, he may have gone from just estimated usage plus offensive rating to a BPM-based player projection model. As @awy says, this is likely underrating our sophomores... hopefully at least. I'm sure the recent years program strength weight is pulling us back up, so the projected production has to be quite low, which as noted is at odds with our talent rating.
 
DRBackup said:
Odd. If you sort by "Projected Effective Talent", UConn is #1 and in that they're +2.8 over second place. No bigger gap in the entire list.

Add that talent to the best coaching staff in America and you... end up #12 ? Huh ?
Expand ...
I would bet some conference-related projected strength of schedule rating bring us down. People will never believe in the Big East.
 
Show me his preseason top-10 for each of the two previous seasons.
 
B1GEast said:
I have never heard of this guy. Supposedly he influences a lot of writers and when he updated his projections, people made a lot criticism. UConn at number 12.

Expand ...
I'm curious. You've been on here since 2011 and have almost 11,000 posts and you've never heard of Bart Torvik? Seems almost impossible. Do you know who Ken Pomeroy is? Similar type stuff. Metrics based projections. There seem to be more of these popping up but those are 2 that are popular and been around a long time.
 
HooperScooper said:
I'm curious. You've been on here since 2011 and have almost 11,000 posts and you've never heard of Bart Torvik? Seems almost impossible. Do you know who Ken Pomeroy is? Similar type stuff. Metrics based projections. There seem to be more of these popping up but those are 2 that are popular and been around a long time.
Expand ...

Never heard of Bart. Kenpom is the standard, but surprisingly until I saw a YouTube post yesterday that was the first time Torvik ever registered.
 
awy said:
tried to find the model but don't see much of anything. i'd venture to guess that there's a lot of normalized noise in the usage patterns, and also a bifurcated model choice between recruits and returnees in that recruits are valued by expected value of a given recruit ranking, while returnees are evaluated on past college production.

this means that for a team that is reliant on returning freshmen who did not get much run last season/didn't produce much, they are being evaluated on that sample and not on their recruit rankings. this can be seen by this model giving uconn the highest "talent" but low production and minutes from those returnees.

this is just a case of uconn falling through the cracks of a model designed to approximate the normal college team in terms of progression and development. wouldn't worry too much about it.
Expand ...
This is spot on - I don’t think a Torvik formula can measure players that haven’t played all that much at the college level, or measure sum of all parts. His formula takes a lot historical data into account, fairly as what other realistic way is there to measure? UConn’s potential is as much art as science given their formula around development and next man up. Really not worth much right now.
 
basically if you take the #1 uconn ranking in talent and try to square it with the projected production of those players, you must also believe that the uconn staff is terrible at developing players and all of the sophomores will be unplayably bad. to patch this error a bit can do a mixed weight of high school projection (recruit ranking) and college production, but doing so would still leave out team level fixed effects from coaching staff. basically by omitting coaching and really the different level of intra-team competition/replacement level production, you leave omitted heterogeneity unaccounted for.
 
B1GEast said:
Never heard of Bart. Kenpom is the standard, but surprisingly until I saw a YouTube post yesterday that was the first time Torvik ever registered.
Expand ...
So after your reply I was thinking well maybe he's not mentioned on here as much as I thought. So I searched for the word "Torvik" on here. It came up 155 separate times, not including this thread, in the last 12 months.
 
You must log in or register to reply here.

Online statistics

Members online
212
Guests online
1,282
Total visitors
1,494
Totals may include hidden visitors.

Share This Page

Forum statistics

Threads
163,962
Messages
4,376,856
Members
10,168
Latest member
CTFan142


.
..
Top Bottom