adlai
Century (over 100 posts)
Posts: 158
|
Post by adlai on Dec 6, 2010 15:15:09 GMT -5
KenPom's ratings this year began earlier based on a Bayesian algorithm. Non-Bayesian ratings will only be put up when enough games have been played that there are links between the teams. I'm not sure if those are up yet or not, but if not they will be soon.
I actually tend to think that KenPom is actually the best set of rankings with one primary exception which is teams that always tend to lose close games. Statistically, there is very little to differentiate between a scoring one more point than the other, yet certain teams tend to do this more than others. There is certainly value in being able to pull it out in the clutch which elevates some teams and lowers others in the human rankings. However, I find that his predictions are generally quite good, especially later in the season.
|
|
|
|
Post by FrazierFanatic on Dec 6, 2010 15:17:02 GMT -5
I do, however, agree with Amassador's main point - too much emphasis is placed on statistical analysis as a pure predictor, especially in college bball, probably the one sport where heart, momentum and adrenaline play the biggest role.
|
|
|
|
Post by TrueHoyaBlue on Dec 6, 2010 15:19:45 GMT -5
what does it say about other methods of evaluating teams? Did ANYONE have the Hoyas losing to Ohio after the Big East Tournament? I'm sure some jokers in NCAA tourney pools did, but they probably also lost out on some of the other long-shot picks they made in their pools. |
|
nathanhm
Golden Hoya (over 1000 posts)
Posts: 1,041
|
Post by nathanhm on Dec 6, 2010 15:19:46 GMT -5
I don't think AP/USA are better - we've just reached a point where we hear less about those because the stats-based stuff is all over the place and preached as gospel (as if it predicts outcomes), which nobody does with the AP/USA rankings any more in the basketball world. Often times, the first thing talked about on here is what Ken Pom says about a game - and that is just crazy IMO. See Ohio 2010. I think there's some value in it, don't get me wrong, but I think it has gone overboard. (And this is not just hoops - some of the MIT-like stuff in baseball is also tiresome. Can we just enjoy the games without having to be wedded to our calculators? ;D) Do I enjoy the irony of the biggest pseudo intellectual on the board attacking the use of advanced statistics? Yes, yes I do. And if the Ohio game is a good example of why KenPom shouldn't be taken as gospel / predictive, what does it say about other methods of evaluating teams? Did ANYONE have the Hoyas losing to Ohio after the Big East Tournament? Actually going back he prob had Georgetown winning the game as like 85-95% likely. Which means over the course of a season a few of those teams with low percentages are going to beat the teams who are heavily favored. |
|
|
|
Post by jerseyhoya34 on Dec 6, 2010 15:23:34 GMT -5
On Ohio, the problem I see is that KenPom won't take the heat but his adherents give heat to the ESPN/AP stuff (which is partly deserved) for the same reasons - failure of predictive analysis. Again, I don't think KenPom is wholly valueless - just that his value is somewhat exaggerated. Nevermind that, but isn't the point that a good predictor should get something right when others don't? If it just tells us what others believe, what's the added value?
Stats don't tell (and never will) the whole picture in a basketball game, especially at a high level. Little plays (like the Wright thing at the end of the WVU BET game) don't show up in a box score but tell you something about the game and why one team lost and won. Stats won't tell you why Roy was good against some teams and not others in 2006.
On Sims, although he had great numbers against NCSU, you can't learn his impact from a box score - you have to watch the game and figure out why the tempo/game turned when he was in.
Can KenPom help out? Sure, but this is not a math/science experiment - it is a sport with these different intangibles. And the tempo free stuff can tell you Sims/Vee have done great things for us, but that's about all you can say IMO - you don't necessarily get the when and why from it - nor would KenPom have predicted HANKAMANIA. ;D
|
|
FLHoya
Diamond Hoya (over 2500 posts)
Proud Member of Generation Burton
Posts: 4,544
|
Post by FLHoya on Dec 6, 2010 15:29:56 GMT -5
Did KenPom bang your mom or something? And what would kenpom project the FanMatch "Thrill Score" to be for this encounter? Just curious.  |
|
kghoya
Platinum Hoya (over 5000 posts)
Posts: 5,049
|
Post by kghoya on Dec 6, 2010 15:30:08 GMT -5
The Hoyas find themselves at #1 in the Yahoo! Users Rankings. I'm surprised that isn't getting more love on Twitter this afternoon.
|
|
theexorcist
Diamond Hoya (over 2500 posts)
Posts: 3,506 
|
Post by theexorcist on Dec 6, 2010 15:35:07 GMT -5
I do, however, agree with Amassador's main point - too much emphasis is placed on statistical analysis as a pure predictor, especially in college bball, probably the one sport where heart, momentum and adrenaline plays the biggest role. I'll actually disagree with that. College basketball has numerous possessions, each of which are worth a small part of the overall score. College football has few possessions, each of which is incredibly more important. A momentum swing in football can be 14 points, which wins most games. And statistical analysis seems a better predictor than some bogus definition of "heart". Two teams with a 19-10 record may be far divorced from each other in terms of quality. Too often, the decision is that "Arizona is a better team than Ball State" just because they're the Wildcats. Statistical analysis ignores the names on the jerseys and simply evaluates performance. It's an awful predictor for single games in the NCAA tournament, which is one-and-done, but no more so than any other predictor due to the fact that regression toward the mean doesn't happen in one game. KenPom continues to rate successful teams in the NCAAs well in general because it demands more rigor than "Kentucky has lots of fans and Baylor has never done well in the tourney". |
|
GUMBA
Silver Hoya (over 500 posts)
Posts: 737
|
Post by GUMBA on Dec 6, 2010 15:43:56 GMT -5
Four Big East Teams in the top ten. Only one ACC in the top 25 - Duke at #1. Didn't we beat them last year?
|
|
|
|
Post by TrueHoyaBlue on Dec 6, 2010 15:50:47 GMT -5
And (in the coaches poll), 8 Big East teams among the top 27 vote-getters.
Clearly a down year for the conference.
|
|
|
|
Post by jerseyhoya34 on Dec 6, 2010 15:50:56 GMT -5
To KenPom's credit, he has a prediction tracker - but his predictions do not seem to be outperforming the rates/probabilities that he's been giving them. More like a wash, but he does seem to be underperforming more in the closer calls - as anybody would.
BE is surprisingly good this year relative to the rest of the country - think SU will come back to earth but UCONN/GU are surprises that have added significantly thus far.
|
|
kchoya
Platinum Hoya (over 5000 posts)
Enter your message here...
Posts: 9,934
|
Post by kchoya on Dec 6, 2010 15:54:52 GMT -5
To KenPom's credit, he has a prediction tracker - but his predictions do not seem to be outperforming the rates/probabilities that he's been giving them. More like a wash, but he does seem to be underperforming more in the closer calls - as anybody would. BE is surprisingly good this year relative to the rest of the country - think SU will come back to earth but UCONN/GU are surprises that have added significantly thus far. What did KenPom ever do to you? Did he decline to read one of your 5,000 word missives on the Blue & Gray board? |
|
kchoya
Platinum Hoya (over 5000 posts)
Enter your message here...
Posts: 9,934
|
Post by kchoya on Dec 6, 2010 16:03:13 GMT -5
|
|
|
|
Post by Coast2CoastHoya on Dec 6, 2010 16:04:43 GMT -5
I'll admit to being a frequent (4+ times/week) viewer of kenpom's site as a disclaimer. I'm also an end-user of stats, as opposed to someone who does more than basic analysis. I use it primarily to scope out our next opponent and to use the score predictor as a starting point for my prediction thread, er, prediction.
I like pom.com not for actually predicting outcomes (though I find the percentage probabilities interesting) so much as being able to compare and contrast key stats and matchups between the Hoyas and the opponents, such as rebounding and TO rates, 3pt% and 2pt%, etc. I think the Luck and Consistency ratings are also interesting . . . and help us explain away a crazy season like 2 years ago when we couldn't buy a close game W. In other words I like the reflective functions of the site. It also loads the fastest on slow connections, so it's good on the road.
I agree that it seems strange to take a prediction -- and this applies to any prediction, really -- about something as filled with unknowns as a college basketball game as "gospel." Most sports fans understand the "Anything Can Happen" Theorem. And I've rarely, if even, seen someone post "well, but kenpom.com said . . . !" in reaction to an unlikely win or loss.
In all, I find it to be a valuable part of the overall ratings/rankings discussion, especially taken in concert with the RPI and Sagarin ratings.
My 2 cents.
Back to the central topic of the thread: good to see the Hoyas rise slowly, as should happen this time of year. Big test on Thurs.
|
|
|
|
Post by FrazierFanatic on Dec 6, 2010 16:13:13 GMT -5
I do, however, agree with Amassador's main point - too much emphasis is placed on statistical analysis as a pure predictor, especially in college bball, probably the one sport where heart, momentum and adrenaline plays the biggest role. I'll actually disagree with that. College basketball has numerous possessions, each of which are worth a small part of the overall score. College football has few possessions, each of which is incredibly more important. A momentum swing in football can be 14 points, which wins most games. And statistical analysis seems a better predictor than some bogus definition of "heart". Two teams with a 19-10 record may be far divorced from each other in terms of quality. Too often, the decision is that "Arizona is a better team than Ball State" just because they're the Wildcats. Statistical analysis ignores the names on the jerseys and simply evaluates performance. It's an awful predictor for single games in the NCAA tournament, which is one-and-done, but no more so than any other predictor due to the fact that regression toward the mean doesn't happen in one game. KenPom continues to rate successful teams in the NCAAs well in general because it demands more rigor than "Kentucky has lots of fans and Baylor has never done well in the tourney". Basketball is played at a much faster pace, and is more prone to momentum swings than football, where the plethora of timeouts after turnovers and virtually every other change of possession can easily deaden the momentum that could otherwise occur. And to characterize "heart" - or whatever you want to call it - as a "bogus"factor ignores reality,especially when you then seem to define it as an element of the name or reputation of the team, rather than a quality of individual players or a team as a whole. I don't disagree that statistics have a place in helping to evaluate and even predict, but I jsut feel thatb they have become a convenient way to try to predict outcome without incorporating the many "intangibles" that not only play a part, but can be the deciding factor. |
|
SFHoya99
Blue & Gray (over 10,000 posts)
Posts: 17,899
|
Post by SFHoya99 on Dec 6, 2010 16:22:20 GMT -5
(And this is not just hoops - some of the MIT-like stuff in baseball is also tiresome. Can we just enjoy the games without having to be wedded to our calculators? ;D) I hate this comment, because it's basically like, "I want to spout stupid comments with no backing, so don't contradict me." You don't have to read it if you don't want to. I do get that as the stats community grows larger, you get a lot of people who a) don't really understand what even the new, better stats are saying b) are overconfident in their accuracy c) tend to be arrogant about it I don't really like them, either. Anyone who pushes analysis with definitives is often posturing, not being honest about the real level of understanding. But I don't really think they've reached the same level of arrogance as Bill Plaschke telling me that Derek Jeter adds fifty wins to the Yankees. |
|
|
|
Post by jerseyhoya34 on Dec 6, 2010 16:29:30 GMT -5
Fair points, SF, and I agree. Admittedly, that is a large part for my problem with the stats stuff. I do read and enjoy the HoyaProspectus stuff in spots and don't mean to attack that kind of thing if that's how it came across. The box score analysis is good if cabined.
The point I'm making is supposed to be fairly limited, and there's a tendency to defend the stats stuff beyond its value.
|
|
|
|
Post by Coast2CoastHoya on Dec 6, 2010 16:55:16 GMT -5
Here's one kenpom prediction that I think we can all agree on: Thursday's game vs. Temple will be the most thrilling college basketball game that day.
|
|
|
|
Post by bigelephant on Dec 6, 2010 17:06:32 GMT -5
There goes Under the Radar.........
KenPom is one of many tools (no pun intended) that help to figure out how good a team is. Usually GT fares pretty well in his system
|
|
|
|
Post by FrazierFanatic on Dec 6, 2010 17:17:31 GMT -5
(And this is not just hoops - some of the MIT-like stuff in baseball is also tiresome. Can we just enjoy the games without having to be wedded to our calculators? ;D) I hate this comment, because it's basically like, "I want to spout stupid comments with no backing, so don't contradict me." You don't have to read it if you don't want to. I do get that as the stats community grows larger, you get a lot of people who a) don't really understand what even the new, better stats are saying b) are overconfident in their accuracy c) tend to be arrogant about it I don't really like them, either. Anyone who pushes analysis with definitives is often posturing, not being honest about the real level of understanding. But I don't really think they've reached the same level of arrogance as Bill Plaschke telling me that Derek Jeter adds fifty wins to the Yankees. Said much more succinctly and accurately than my thoguhts; they are a tool, and can even be a good tool,in evaluating a team or a player, but too often they do not account for critical elements. Derek Jeter is not worth 50 games; however, can you really say that you would have no preference between him and another shortstop with the same specific stats? |
|
