tashoya
Blue & Gray (over 10,000 posts)
Posts: 12,331
|
Post by tashoya on Mar 5, 2013 23:03:29 GMT -5
SF... thank you for the insight. Much appreciated.
|
|
SFHoya99
Blue & Gray (over 10,000 posts)
Posts: 17,791
|
Post by SFHoya99 on Mar 5, 2013 23:07:33 GMT -5
SF... thank you for the insight. Much appreciated. I should note that I'm not a statistician, though I've had some training. But I work with a whole slew of them on the business side (big data and all that). And I see where things break down constantly. But I find this stuff fun, too.
|
|
bmartin
Golden Hoya (over 1000 posts)
Posts: 2,459
|
Post by bmartin on Mar 5, 2013 23:38:35 GMT -5
Wisconsin had early season home wins by 46, 45, 40, & 33 over teams in the 250 to 320 range. Those blowout margins have not prevented them from losing 9 games against real teams. Because of the distorted efficiency numbers, they were ranked 5th last week by Pomeroy with an 19-8 record and then dropped only to 8th after getting a 9th loss to a 15-loss Purdue team by 13 at home. They seem to do this every year - put up several 80 - 40 home wins against overmatched teams that distort the ratings.
|
|
tashoya
Blue & Gray (over 10,000 posts)
Posts: 12,331
|
Post by tashoya on Mar 5, 2013 23:44:03 GMT -5
How very Syracuse of Wisconsin.
|
|
tashoya
Blue & Gray (over 10,000 posts)
Posts: 12,331
|
Post by tashoya on Mar 5, 2013 23:45:54 GMT -5
I wonder if the Dolphins said the Carrier Dome was closed....
|
|
johnnysnowplow
Golden Hoya (over 1000 posts)
Posts: 2,004
Member is Online
|
Post by johnnysnowplow on Mar 6, 2013 0:25:45 GMT -5
Well I walked away from this thread this afternoon and came back just now to find a few more pages of pretty in-depth statistical analysis. Many thanks to SF and others for the contributions here. For all the arguing and namecalling and such that goes on within these Hoyatalk walls, this was a very informative and pleasant conversation to follow. Kind of refreshing actually.
I tend to follow the advanved stats more than most I would say, but I've mentioned before, I usually like them as a tool for analyzing past performance, not for predictive value. I'm not saying they don't have any predictive value, but I'm far less versed in general statistical analysis than some here are and so I tend to shy away from it. Just my personal preference.
That being said, the key for me in this whole "will the Hoyas regress" discussion is that we all watch every minute of every game (or at least try to) so we have both stats and eyes to use as our "data". When we talk about other teams, how much have we really watched those teams? Probably not that much, at least not even close to as much as we've watched the Hoyas. So the stats are really the driving factor there. And as SF pointed out, statistical models like KenPom tend to lag true performance. With our guys, I think it's obvious at this point from watching them game in and game out that they are clearly not the same team they were in the ooc schedule. These guys have improved significantly, both individually and as a team. So when you combine the fact that the stats would say we are much better, and our eyes tell us we are much better, there's a very good chance that we are actually much better. Basically, I guess what I'm getting at is to me 3-4-5-maybe even 6 games could be considered a hot streak and a regression to the mean is probably likely. 11 games is an absurdly long hot streak. I think we have to acknowledge that the mean has shifted. I tend to think we're probably somewhere between the AP's 5 and KenPom's 16. Will we lose again? Almost definitely. But should we be expected to regress to whatever our mean was back in January? I personally don't think so.
|
|
gujake
Silver Hoya (over 500 posts)
Posts: 831
|
Post by gujake on Mar 6, 2013 2:05:27 GMT -5
I'll add one thing to the stat discussion - out of ease/habit, most people tend to look at only one model. Pomeroy tends to be the default model that people look at for NCAA basketball. But it may be better to look at a consensus of different ranking systems. In most fields, consensus forecasts perform a lot better over the long run than individual forecasts. Note that this is not always true, and I'm not sure if it's true in college basketball. But it's true in most fields. For example, there is a website here that lists the current rankings from 50+ models/polls: masseyratings.com/cb/compare.htm . By the average of these 50+ rankings, we are currently 7th. If you're only looking at Pomeroy, it's easy to conclude that the stat nerds say that Georgetown is overrated. But not all of them say that. One stat guy (Pugh) even has the Hoyas as #1 in his model. On average, we don't look all that overrated.
|
|
MCIGuy
Platinum Hoya (over 5000 posts)
Anyone here? What am I supposed to update?
Posts: 9,427
|
Post by MCIGuy on Mar 6, 2013 7:06:29 GMT -5
Why do these stat geeks even matter in the first place? Do they routinely predict how the Hoyas will do in conference play or how they will do in the post season? Aren't you guys giving them a little too much credence?
|
|
SirSaxa
Blue & Gray (over 10,000 posts)
Posts: 15,620
|
Post by SirSaxa on Mar 6, 2013 8:25:32 GMT -5
All I know is our Hoyas have outperformed even my personal, optimistic, pre-season expectations. And topping all of that has been OTTO! I expected a big sophomore jump in his performance, but WOWOWEE!
Go Hoyas!
|
|
jgalt
Diamond Hoya (over 2500 posts)
Posts: 4,380
|
Post by jgalt on Mar 6, 2013 12:52:09 GMT -5
And that may be true, on average, over lots of teams over lots of years. Pomeroy builds his system to be right like a gambler would build his system, over lots and lots of bets. But we have more information about Georgetown. Information not in the model. Information that in the aggregate might look like noise but isn't. What is the information that we can add to the model that would enhance it? I personally dont see any information that makes me believe Georgetown will buck the trend that many other teams have demonstrated over many years. I guess I am mostly confused because the point I was trying to communicate (which I may easily have failed at doing) was rather broad, not radical, and based on a large amount of data. I didnt mean it to be predictive in the way that Pomeroy tries to be, but predictive more in the sense that it outlined the unlikelihood of future extreme success (i.e a final four or farther). Out of the range of possible out comes I think it is unlikely that Georgetown lives up to the lofty expectations of its current ranking and its seed. That was the question posed by the thread and that was my answer. Also Signal and Noise is the next book on my list. It is on my night stand right now and I am very excited to read it (just after I finish the current Ayn Rand book I am reading ;D [this is not a joke!])
|
|
bmartin
Golden Hoya (over 1000 posts)
Posts: 2,459
|
Post by bmartin on Mar 6, 2013 13:23:36 GMT -5
But there are outliers in any system. It just is not true that all teams with a similar profile have had identical outcomes. Their average outcome is the average. It is not the universal experience.
|
|
SFHoya99
Blue & Gray (over 10,000 posts)
Posts: 17,791
|
Post by SFHoya99 on Mar 6, 2013 13:29:51 GMT -5
And that may be true, on average, over lots of teams over lots of years. Pomeroy builds his system to be right like a gambler would build his system, over lots and lots of bets. But we have more information about Georgetown. Information not in the model. Information that in the aggregate might look like noise but isn't. What is the information that we can add to the model that would enhance it? I personally dont see any information that makes me believe Georgetown will buck the trend that many other teams have demonstrated over many years. I guess I am mostly confused because the point I was trying to communicate (which I may easily have failed at doing) was rather broad, not radical, and based on a large amount of data. I didnt mean it to be predictive in the way that Pomeroy tries to be, but predictive more in the sense that it outlined the unlikelihood of future extreme success (i.e a final four or farther). Out of the range of possible out comes I think it is unlikely that Georgetown lives up to the lofty expectations of its current ranking and its seed. That was the question posed by the thread and that was my answer. Also Signal and Noise is the next book on my list. It is on my night stand right now and I am very excited to read it (just after I finish the current Ayn Rand book I am reading ;D [this is not a joke!]) Pomeroy uses the full season of performance. So there's still a large percentage of the Hoyas' performance where DSR is limited in minutes and usage. Where Hopkins is our primary offensive player, not Porter. Where Porter and DSR are not playing at this level of performance. Of these, only the last two seem possible to revert. Systems like Pomeroy use the average level of regression to the mean but there's a lot out there when you watch Porter and DSR to believe they will regress less than average. But these are all things not in his assumptions. Neither was the knowledge in 06-07 when Thompson changed his D to extend out and let Roy defend the rim. The defense vastly improved with this decision -- and it took forever for systems like Pomeroy to understand that improvement was not random variance. Fans tend to ignore things as well. That's why these systems are useful. We tend to ignore that maybe Nate or Markel may regress.
|
|
bmartin
Golden Hoya (over 1000 posts)
Posts: 2,459
|
Post by bmartin on Mar 6, 2013 18:53:50 GMT -5
There are always details and mysteries that are not included in any stats. There are outliers that get significantly better or worse as the season progresses. No model would predict that the 2009 Georgetown team with Monroe, Summers, Wright, Freeman, and Sapp would go from 10-1, Top 10 after winning easily at UConn, to 16-15, losing 12 of their last 16 games. Pomeroy still had them 27th after all that.
Certain types of teams tend to be overrated by computer models - middling majors that run up big margins against very weak teams but then can't defend teams in their conference. Pomeroy wrote a blog piece about that a few years ago to explain why his system overrated Oklahoma that season but then rejecting an adjustment to the model.
|
|
|
Post by Problem of Dog on Mar 23, 2013 19:00:09 GMT -5
If we're a 1, we will lose to a 16. It's just gonna happen. Georgetown being Georgetown. Told you so.
|
|
DanMcQ
Moderator
Posts: 30,607
|
Post by DanMcQ on Mar 23, 2013 19:48:30 GMT -5
If we're a 1, we will lose to a 16. It's just gonna happen. Georgetown being Georgetown. Told you so. Congrats. Please PM us your address so we can put this in the mail for you.
|
|
|
Post by detmut on Mar 23, 2013 20:47:04 GMT -5
If we're a 1, we will lose to a 16. It's just gonna happen. Georgetown being Georgetown. Told you so. and get rid of your stupid avatar
|
|
|
Post by Problem of Dog on Mar 23, 2013 23:40:54 GMT -5
and get rid of your stupid avatar Well I think that is probably my one redeeming quality for many on here.
|
|
AvantGuardHoya
Golden Hoya (over 1000 posts)
"It was when I found out I could make mistakes that I knew I was on to something."
Posts: 1,481
|
Post by AvantGuardHoya on Mar 24, 2013 7:37:15 GMT -5
Not.
|
|