(10-23-2017 03:48 PM)bearcatmark Wrote:  

(10-23-2017 03:40 PM)rath v2.0 Wrote:  So now there are computer rankings before a single shot has been taken on the season? At a certain point we have jumped the sabermetric shark.

There are writers rankings before a single shot has been taken as well. People like predictions. Why wouldn't someone use data driven measures for that?

BTW there are multiple programs that use these kind of numbers for nonconference scheduling (find the lower level teams that should have good RPI numbers) to great affect.

does kenpom have any teams that are significantly different in its rankings than in the others?

(10-23-2017 07:20 PM)bearcatmark Wrote:  

(10-23-2017 06:31 PM)Billy_Bearcat Wrote:  

(10-23-2017 11:48 AM)bearcatmark Wrote:  

(10-23-2017 11:36 AM)CincyBro Wrote:  kenpom.com

UC starts 11th

I'm sure you will be shocked to find out I've been periodically checking kenpom.com waiting for this. Even a bit higher than I expected, though I was thinking around 15. Really pumped for this season.

Shocked? I expect you have the Super Platinum Membership with synchronized Flesh Light

Wait how did you know about the Super Platinum Membership perks? I thought it was a secret... Billy are you a member too?

Card carrying member

(10-23-2017 08:48 PM)Billy_Bearcat Wrote:  Card carrying member

....the groups and organizations of which are too numerous to list.

Quick post on why I value good, computer numbers over human polls.

It's not that watching basketball is irrelevant or you can't understand the game based on the "eye" test. To get a full understanding of a team, their strengths / weaknesses, etc, the eye test is essential. The problem with human polls is there is no reason to believe the humans voting have watched enough of all the teams to really evaluate them. I like computer polls, because though there will always be outliers, things will inevitably be missed, computer polls evaluate teams based on every game...and in the case of kenpom based on every possession in every game.

If the coaches that voted in the coaches poll sat down and watched every single game of every season and really broke down every team undoubtedly they'd ail give us a pretty great picture. Alas the coaches can probably give you an amazing breakdown of their team and any team on their schedule, but it's unlikely they've really evaluated the vast majority of the teams. This is also true of the writers. If I watch UCF twice all season I can learn something, but how in depth is my knowledge really going to be of the team. What I like about computer rankings is though they're imperfect and you don't view them as gospel...i think through sheer amount they evaluate the do a better job than human polls. Most statisticians will tell you that small sample sizes with these computer polls less reliable. Kenpom notes that even halfway through there is a lot of margin for error. Think about how much error you can get if you catch a team you only watch once on a really hot shooting night or catch a decent team against good teams on a couple nights where the other team doesn't make shots.

You think the vast majority of coaches and writers than vote in the polls watched more than a game or two of Wichita State last year? What about SMU? There's plenty you can pick out even in one game (athleticism being probably the easiest), but you catch Troy Caupain on the right night and you might think he's a knock down shooter. I remember when the 2006 team missed the tournament, the head of the committee talked about watching Air Force in the conference tournament final and saying they passed the eye test with their unique style. It was one game, that they lost that essentially got them in the tournament, despite a weak resume. I doubt anyone on the committee (other than the AD from their conference) had watched a single game involving Air Force all year before that. It's a a bad way to decide how good that basketball team is. You can learn some things, but in one game there is too much error.

So I think the computers give a great starting point for conversation about basketball. I like kenpom because it evaluates every team based on every possession of its season. That doesn't mean I don't watch games or don't pick different things up watching games. My guess is I watch more basketball than about everyone here. I love watching hoops. I don't believe because Wichita State finished 8th in kenpom last year they were definitively the eighth best team in the country. I think they were undervalued by the human polls and that number was a better reflection than their seed, but I think there were teams behind them in kenpom that were better.

No poll is ever going to be perfect, but give me the one that uses the most information possible... over the one based on small samples of the random games a writer / coach is able to watch,

(10-23-2017 02:40 PM)MercerCo_BearCat Wrote:  

(10-23-2017 12:12 PM)bearcatmark Wrote:  

(10-23-2017 12:07 PM)Former Lurker Wrote:  How can computer ratings have any validity before games have been played?

How can any rankings have any validity before games have been played?

You account for returning talent, talent development and new players and use what you know to project.

With human rankings this is done subjectively... With computer rankings this is done using numbers from prior season returning players and weighting of incoming players based on how these things have effected teams historically.

kenpom especially. he says his rankings preseason are fundamentally useless. Kenpom's is based of how they played and by the time season's end is usually really good.

His preseason rankings are a trap shoot based off of returning players/coaches.

THIS.

statistical analysis is useless at this point. wake me up in, Jan. Things should be coming into focus at that point. The rankings are purely cosmetic at this point.




mc

Actually, I believe you will find that in most sports, the best indicator of who will have a good team this year is who had a good team last year.

Basing beginning of the year ratings on where teams finished the year before is probably more accurate than any other method.

(10-24-2017 12:54 PM)Bruce Monnin Wrote:  Actually, I believe you will find that in most sports, the best indicator of who will have a good team this year is who had a good team last year.

Basing beginning of the year ratings on where teams finished the year before is probably more accurate than any other method.

Unless they lose a ton of good players, have injuries, or change coaches, etc.

Why lie awake at night unable to sleep with cold sweats? KenPom preseason isn't worth it.

(10-24-2017 10:01 AM)bearcatmark Wrote:  Quick post on why I value good, computer numbers over human polls.

It's not that watching basketball is irrelevant or you can't understand the game based on the "eye" test. To get a full understanding of a team, their strengths / weaknesses, etc, the eye test is essential. The problem with human polls is there is no reason to believe the humans voting have watched enough of all the teams to really evaluate them. I like computer polls, because though there will always be outliers, things will inevitably be missed, computer polls evaluate teams based on every game...and in the case of kenpom based on every possession in every game.

If the coaches that voted in the coaches poll sat down and watched every single game of every season and really broke down every team undoubtedly they'd ail give us a pretty great picture. Alas the coaches can probably give you an amazing breakdown of their team and any team on their schedule, but it's unlikely they've really evaluated the vast majority of the teams. This is also true of the writers. If I watch UCF twice all season I can learn something, but how in depth is my knowledge really going to be of the team. What I like about computer rankings is though they're imperfect and you don't view them as gospel...i think through sheer amount they evaluate the do a better job than human polls. Most statisticians will tell you that small sample sizes with these computer polls less reliable. Kenpom notes that even halfway through there is a lot of margin for error. Think about how much error you can get if you catch a team you only watch once on a really hot shooting night or catch a decent team against good teams on a couple nights where the other team doesn't make shots.

You think the vast majority of coaches and writers than vote in the polls watched more than a game or two of Wichita State last year? What about SMU? There's plenty you can pick out even in one game (athleticism being probably the easiest), but you catch Troy Caupain on the right night and you might think he's a knock down shooter. I remember when the 2006 team missed the tournament, the head of the committee talked about watching Air Force in the conference tournament final and saying they passed the eye test with their unique style. It was one game, that they lost that essentially got them in the tournament, despite a weak resume. I doubt anyone on the committee (other than the AD from their conference) had watched a single game involving Air Force all year before that. It's a a bad way to decide how good that basketball team is. You can learn some things, but in one game there is too much error.

So I think the computers give a great starting point for conversation about basketball. I like kenpom because it evaluates every team based on every possession of its season. That doesn't mean I don't watch games or don't pick different things up watching games. My guess is I watch more basketball than about everyone here. I love watching hoops. I don't believe because Wichita State finished 8th in kenpom last year they were definitively the eighth best team in the country. I think they were undervalued by the human polls and that number was a better reflection than their seed, but I think there were teams behind them in kenpom that were better.

No poll is ever going to be perfect, but give me the one that uses the most information possible... over the one based on small samples of the random games a writer / coach is able to watch,

You said "quick post" in your opening line and the proceeded to write a TodgeRodge level post.... :-)

(10-24-2017 05:20 PM)Bearcats#1 Wrote:  

(10-24-2017 10:01 AM)bearcatmark Wrote:  Quick post on why I value good, computer numbers over human polls.

It's not that watching basketball is irrelevant or you can't understand the game based on the "eye" test. To get a full understanding of a team, their strengths / weaknesses, etc, the eye test is essential. The problem with human polls is there is no reason to believe the humans voting have watched enough of all the teams to really evaluate them. I like computer polls, because though there will always be outliers, things will inevitably be missed, computer polls evaluate teams based on every game...and in the case of kenpom based on every possession in every game.

If the coaches that voted in the coaches poll sat down and watched every single game of every season and really broke down every team undoubtedly they'd ail give us a pretty great picture. Alas the coaches can probably give you an amazing breakdown of their team and any team on their schedule, but it's unlikely they've really evaluated the vast majority of the teams. This is also true of the writers. If I watch UCF twice all season I can learn something, but how in depth is my knowledge really going to be of the team. What I like about computer rankings is though they're imperfect and you don't view them as gospel...i think through sheer amount they evaluate the do a better job than human polls. Most statisticians will tell you that small sample sizes with these computer polls less reliable. Kenpom notes that even halfway through there is a lot of margin for error. Think about how much error you can get if you catch a team you only watch once on a really hot shooting night or catch a decent team against good teams on a couple nights where the other team doesn't make shots.

You think the vast majority of coaches and writers than vote in the polls watched more than a game or two of Wichita State last year? What about SMU? There's plenty you can pick out even in one game (athleticism being probably the easiest), but you catch Troy Caupain on the right night and you might think he's a knock down shooter. I remember when the 2006 team missed the tournament, the head of the committee talked about watching Air Force in the conference tournament final and saying they passed the eye test with their unique style. It was one game, that they lost that essentially got them in the tournament, despite a weak resume. I doubt anyone on the committee (other than the AD from their conference) had watched a single game involving Air Force all year before that. It's a a bad way to decide how good that basketball team is. You can learn some things, but in one game there is too much error.

So I think the computers give a great starting point for conversation about basketball. I like kenpom because it evaluates every team based on every possession of its season. That doesn't mean I don't watch games or don't pick different things up watching games. My guess is I watch more basketball than about everyone here. I love watching hoops. I don't believe because Wichita State finished 8th in kenpom last year they were definitively the eighth best team in the country. I think they were undervalued by the human polls and that number was a better reflection than their seed, but I think there were teams behind them in kenpom that were better.

No poll is ever going to be perfect, but give me the one that uses the most information possible... over the one based on small samples of the random games a writer / coach is able to watch,

You said "quick post" in your opening line and the proceeded to write a TodgeRodge level post.... :-)