COLUMN: Why the analytics still appreciate Iowa State

[ChrisMWilliams]

Most of you probably know this stuff but I wrote this more for the casual fan as I'm getting this question a lot.

https://cyclonefanatic.com/2019/03/...basketball-is-still-appreciated-by-analytics/

 

[coolerifyoudid]

Since this is the first year, it will take some getting used to. In years past, how teams trend at the end of a season definitely factored into their seeding.

Thankfully, we built up a good resume early. It's been hard to find a lot of positive things lately and optimism only takes me so far. Fortunately each game is it's own microcosm and doesn't dictate how you'll do in the next one.

I've been a Cyclone fan forever. No need to stop now.

 

[Messi]

TLDR: team chemistry isn’t an analytic you can measure.

 

[twojman]

I'll be pleasantly surprised if ISU wins another game this season. The team clearly doesn't get along and it is really bubbling over at the worst possible time.

 

[HFCS]

The board has crucified me for years for stating that KenPom and now the similar Net Ranking are horrible tools to use to select a field of teams based on resume strength.

Turns out we're going to benefit from it this year, but I don't see how I've been so wrong. It seems crazy obvious that rankings that minimize Ws and Ls can be ideal for gambling and single game predictions while also sucking as resume evaluators.

 

[SCISUFAN]

Analytics versus reality.

 

[twojman]

Also, when you click on this thread, what does your tab look like up top? I'm using Edge. The words on my tab name make me laugh like an immature teenage boy.

 

[ArgentCy]

Analytics versus reality.

People are many MANY companies are too obsessed with data these days. We know the data doesn't matter because this team has serious mental / chemistry problems. Or your going to have much deeper into that data and look for the problem signs showing up (giving up 54 points to a bad WVU in a half is all you need).

 

[Sigmapolis]

@ChrisMWilliams

I appreciate the article. I think a better name for it should have been something like, "Why ISU still makes the NCAA tournament," which comes down to the Committee beginning to pay attention more to computer analysis instead of just wins/losses.

Being the nerd that I am as the OP in this thread...

https://cyclonefanatic.com/forum/threads/2018-2019-computer-projections-thread.243443/

...I thought to quantify why people are panicking.

For me, Barttovik releasing individual game scores has been a godsend for keeping things in context during an up-and-down season. How did we play tonight? The scores range from 0% (better than 0% of other college basketball teams) to 100% (the best team in the country). Pulling a 97% means you were better than 97% of other D-1 teams that night.

The problem with the team breaks down to...

(1.) They had a so-so start to the season, roughly in-line with expectations.
(2.) The had an awesome January -- seriously -- which raised our expectations.
(3.) They came back to Earth in February and, then...
(4.) Have recently fallen off a cliff.

Barttovik has this data...

http://www.barttorvik.com/team.php?team=Iowa+St.

I took the individual game scores and charted them out by month.



40% = roughly the 212th best team in the country... we belong in the Big South

I charted out the individual scores with a linear fit and a five-game centered moving average. As you can see, neither of them have the program going anywhere good.



The team has been trending down since the last WVU game, and badly since OU.

I know the computers like us because we have played well in the past, especially in January. We were great then. That ultimately becomes kind of meaningless, however, when the team from the last few weeks has very little reasonable shot against Texas Tech, Baylor, and whoever we are going to line up against for an opposite seed in the NCAA tournament.

The panic warnings are in there when you look for them.

 

[Cybyassociation]

This part of the season is why you can't trust analytics all of the time. Analytics don't care if your team is fighting in practice or on national television on the court. Analytics don't care if your team is tired of seeing each other's face every single day.

 

[jsb]

This part of the season is why you can't trust analytics all of the time. Analytics don't care if your team is fighting in practice or on national television on the court. Analytics don't care if your team is tired of seeing each other's face every single day.

Right. If Iowa state played the way they have in February for the entire season, they’d be out. But they didn’t. They were good and deserve the tournament as much as anyone.

 

[alarson]

The board has crucified me for years for stating that KenPom and now the similar Net Ranking are horrible tools to use to select a field of teams based on resume strength.

Turns out we're going to benefit from it this year, but I don't see how I've been so wrong. It seems crazy obvious that rankings that minimize Ws and Ls can be ideal for gambling and single game predictions while also sucking as resume evaluators.

I mean, the 'luck' factor is even specifically there to say 'this is how different your results are from this rating'. That should make it pretty clear kenpom isnt a resume rank at all, and more of a 'this is how good it thinks your team is'.

There are advantages and disadvantages to that. Its always a debate as to whether who belongs in the tournament is the best teams or the teams with the best resumes and those teams arent always the same. Hell, you could conceivably go with a hybrid model where you select the field using a more resume-based criteria and then a metric like kenpom for seeding.

 

[Clonefan32]

Lets just hope the analytics don't become self-aware and generate the capabilities of sight and independent thought.

 

[AuH2O]

I mean, the 'luck' factor is even specifically there to say 'this is how different your results are from this rating'. That should make it pretty clear kenpom isnt a resume rank at all, and more of a 'this is how good it thinks your team is'.

There are advantages and disadvantages to that. Its always a debate as to whether who belongs in the tournament is the best teams or the teams with the best resumes and those teams arent always the same. Hell, you could conceivably go with a hybrid model where you select the field using a more resume-based criteria and then a metric like kenpom for seeding.

Some of the problem is calling that variance "luck." If a team sucks at closing out close games, they are considered unlucky in the KenPom world. You could very easily reverse the rating and call it a "clutch" rating. People have perpetuated that idea that studies show that there really aren't "clutch" players and that it's random. First, that's not really true. What these studies show is that there is a variation in performance, but being "clutch" means avoiding variation to the downside. Second, the studies in basketball focused on gathering data over NBA seasons. Comparing a multi-year, professional 28 year old in game #35 out of 81 is not a relevant comparison to a 19 year old kid playing in one of 30 games. The impact for each game is greater.

In general, the debate of KenPom vs. something like RPI has always been along the lines you present - is it the committee's job to put in teams with the best resumes or to put in the teams they think can do the best in the tournament? While I think the RPI was too SOS heavy, I like the concept of using primarily w/l vs. SOS. I do like some weighting to an efficiency-based system like KenPom.

 

[AuH2O]

Since this is the first year, it will take some getting used to. In years past, how teams trend at the end of a season definitely factored into their seeding.

Thankfully, we built up a good resume early. It's been hard to find a lot of positive things lately and optimism only takes me so far. Fortunately each game is it's own microcosm and doesn't dictate how you'll do in the next one.

I've been a Cyclone fan forever. No need to stop now.

In recent years I believe the committee has explicitly stated there is no momentum factored in, and recent performance is not weighted differently than early season games. I think their selections have backed this up for the most part.

 

[Cydkar]

The selection isn't based on how people think you will do in the tournament. Just because ISU's resume is way better than their current level of play doesn't make the analytics "wrong". The committee isn't supposed to care if a team is hot or cold at selection time. At least they don't claim to care.

They used to have a last 10 game metric but that was shot down because schedules aren't even for all teams at the end of the year. How we are seeded will depend on how much the human element is weighted. Consciously or unconsciously.

 

[Cydkar]

Since this is the first year, it will take some getting used to. In years past, how teams trend at the end of a season definitely factored into their seeding.

Thankfully, we built up a good resume early. It's been hard to find a lot of positive things lately and optimism only takes me so far. Fortunately each game is it's own microcosm and doesn't dictate how you'll do in the next one.

I've been a Cyclone fan forever. No need to stop now.

Not for a while.

 

[coolerifyoudid]

In recent years I believe the committee has explicitly stated there is no momentum factored in, and recent performance is not weighted differently than early season games. I think their selections have backed this up for the most part.

Not for a while.

I stand corrected. I assumed that was still going on. I used to watch the selection show a little more religiously. I guess I've backed off recently.

I'm curious to see how the teams shake out this year. This year, I haven't felt that controversy leading up to the selection show. I suppose it will ramp up over the next week.

 

[FinalFourCy]

The team has been trending down since the last WVU game, and badly since OU.

I know the computers like us because we have played well in the past, especially in January. We were great then. That ultimately becomes kind of meaningless, however, when the team from the last few weeks has very little reasonable shot against Texas Tech, Baylor, and whoever we are going to line up against for an opposite seed in the NCAA tournament.

The panic warnings are in there when you look for them.

Yep. The defensive slide in the last month wasn’t exactly inconspicuous, but the fissures were deeper than realized.

Our last month is more synthetically bad than January was good. Maybe I’m in shock, but perhaps the one and done nature will get us back to playing well.[/user]

 

[EarthIsMan]

The team has been trending down since the last WVU game, and badly since OU.

I know the computers like us because we have played well in the past, especially in January. We were great then. That ultimately becomes kind of meaningless, however, when the team from the last few weeks has very little reasonable shot against Texas Tech, Baylor, and whoever we are going to line up against for an opposite seed in the NCAA tournament.

The panic warnings are in there when you look for them.

I think the best recent comparison (and potential hope for optimism) is 2016 Notre Dame. Great offense, poor defense that crashes in the regular season only to have a surprising NCAA tourney.

Not saying ISU will do the same, but at least there is precedent.