Math/analytics loves offense more than defense, wasnt that what we figured out last year?
2022-2023 MBB computer projections thread
- Thread starter Sigmapolis
- Start date
A team can have a good offensive game plan, but that all changes when they get punched in the face by our D.
I think it's just easier to measure offense. There are a lot more statistics for it so analytics are going to be biased.
I was worried about who would guard Murray but after seeing Jaz guard Nance I'm much less worried.
I did enjoy after Jaz picked up two back to back fouls on him Jaz started guarding him with his hands behind his back. It was interesting to see.
I think a mix of Jaz and Ward will be sufficient
This is incorrect. Sorry to give you the red "X" for "disagree" -- it seems so harsh when a point about the calculation of net efficiency is so dry/nerdy/passionless -- but that is not how those models work.
All these models... Torvik, KenPom, etc., take your pick... essentially work on calculating net efficiency per possession. You can generate net efficiency in three ways: (1.) score, (2.) prevent your opponent from scoring, and (3.) play good opponents so their strength-of-schedule adjustments will reward you.
If in five possessions, a strong defensive team "wins" 4-2 while a strong offensive team "wins" 8-6, assuming it was the same opponent, the model is going to like those outcomes equally as well.
On a net basis, they're +2 either way.
This is true on the level of individual players, though more advanced metrics for defense are being worked on as player tracking and activity data becomes available, but not for teams. It is pretty easy to measure offense (how much did you score?) and defense (what did you give up?) at the team level.
I remember Brockington mostly having (the other) Murray last year.
I think what we determined last year was it loved margin of victory (heavily rewarding Hawkeyes for running up scores with starters remaining in games for example). This aligns with the above.
Maybe thats what i am remembering. There was a big ole thread about it last year.
Cyclonepride
Thought Police
Yeah, Jaz was really, really good last night. I think he guarded Nance a lot, and made him a non-factor
Our fanatical close outs on the 3 point line is my favorite part of this defense. Our help is usually pretty good in that situation. All part of speeding the opponent up.
Agreed. I remember under Prohm we really struggled to guard stretch 4s like Dean Wade and some of those other guys. Definitely don't seem to have that problem under TJ.
I still don't think this is generally the case for any model based around net margin.
In theory, if Iowa State and Iowa play the same schedule (so there would be no worries about strength of schedule adjustments) and the same number of possessions (nor for pace) --
Iowa wins every game 90-80
Iowa State wins every game 60-50
They should both be +10 net per 100. The model would see them as equally-good teams.
There are two factors that might complicate this --
(1.) making strength of schedule adjustments
(2.) making (or not making) garbage-time adjustments for lopsided games
I don't think it would be the first. Doing that is pretty simple and is usually done the same way across these models of adjusting net margins such that the net margin of their opponents is net zero so that, in theory, every team is being treated like it played the same strength of schedule in the end.
The other one is more complicated to analyze. Some models don't try to do this. Others either have some sort of log function to induce diminishing returns to margins (e.g., going from winning by one to winning by nine is a big deal, but going from winning by 20 to winning by 30 doesn't do much for you) with some conditions (e.g., do make adjustments for OOC home games against non-P6 opponents, do not for conference games), and others might just have a "hard" cutoff of a certain margin where winning by more means nothing. Some even try to have some statistical definition of "garbage time" at the end of games depending on the margin and who was playing and discount or eliminate those possessions from the net margin calculations.
Most of these models document these sorts of things only in conceptual terms (if that). It is hard to know exactly what impact those kinds of adjustments would have without seeing the actual code.
I don't think this biases the models towards Iowa, though. Sure, Iowa likes to run up the score against overmatched opponents in Carver early in the season. But TJ didn't call the dogs off against IUPUI and North Carolina A&T until *very* late in those games, and the Cyclone defensive intensity didn't take a break even as they were up 20+ on some bad teams. That is the "culture" and TJ and the team take pride in that. You can "run up the score" on defense if you keep scoring at a modest pace and completely shut the other team down as much as you can on offense by simply scoring a ton in garbage time and kinda sorta slow them down from scoring.
I'd need something more solid to believe that Torvik and its peers enjoy Iowa's offense more than it enjoys Iowa State's defense. In theory, they should be equal -- it is all about the margin of victory.
It loves margin of victory. People to say it’s “efficiency” and not margin of victory, but it absolutely is just another way to package scoring margin.
Also, larger margins of victory are more easy to come by when you score in the 80s rather than the 50s.
Margin of victory per possession (usually per 100) not margin of victory per game.
Winning by 10 in 60 possessions is the same as winning by 12 in 72 possessions.
You're +16.67 per 100 in either outcome.
Correct, but number of possessions is going to be even or plus/minus 1. So it is 100% a margin of victory metric, but just normalized by number of possessions.
Screw that Iowa prediction defense travels and we have a better offense than last yr. Iowa doesn't like teams that play defense. Only downside is that it is in Iowa City and the hostile crowd. I'll call the W now. Screw them g*d d*mn hawkeyes.
