Dork Polls

[hurleyfor3]

Want to start a dicsussion on the difference between what the computer polls think and how this differs from the public (pollsters, commentators and us) perception of us.

We are not only #2 Pomeroy, but also we have maintained that ranking through the Georgetown and NC State losses. The distance between us and #3 (Bucky) is far greater than that between any other two teams.

We are also #2 Sagarin, although the Cuse is quite close to us and Kansas is way north of everyone. Our "predictor" subrating is #1, but year after year we tend to do better there.

So who thinks we're really the second-best team out there? On the one hand I think after multiple early-round exits more of us have become more sceptical of mid-February rankings than ever. I became jaded to this phenomenon earlier than most, so I'm on board here. But I stand by my earlier implications that no one is really that good or that consistent, and that Kansas isn't even that dominant. We could see a lot of early round losses by 1-2-3 seeds.

The dork polls really like the ACC. The ACC is either #1 or #2 sagarin (behind the Televen by one measure) and #1 Pomeroy. I think the middle/bottom of the ACC, eliminating ncsu and possibly unc, is pretty strong in fact.

So what are the computers thinking that the human pollsters (and we) disagree with? Is it just recollections of past tournaments? Again, I don't feel like we should be #2 either, but right now what's a better idea and why?

[dukebluelemur]

I think the formula poles overvalue (to many human rankings) the value of all those decent-but-not-great ACC teams. This punishes the BE and its top teams for playing the utter crap bottom teams, while it helps Duke.

[Kedsy]

The public doesn't see any significant difference between playing Charlotte and playing Stetson. The computers see a huge difference. IMO, this explains much of the discrepancy.

The computers can't see a player actually play, so they rely on their creators being able to find the proper statistics that tell them as much as seeing the games tell us. In many ways the computer way is more accurate, because they can't be fooled by flash and high-flying entertainment. In other ways the computers miss things because there are no statistics to measure certain aspects of the game.

As far as why the pollsters, media, and public, don't completely agree with the computers, it is human nature to discount things we don't entirely understand, especially when our eyes tell us something different.

As far as why Duke doesn't get the same respect from humans as from computers, you have to remember that according to the computers we've been similarly dominant almost every season. For example, according to Sagarin, we were #4 in 2005, #1 in 2006, #9 in 2008, and #6 in 2009. And the funny part is we may really have earned those numbers in each of those years. Because here's the last word on computer ratings: they're not so good at predicting the outcome of a single game. As long as NCAAT performance is the primary factor in the perception of who is the best, the computer ranking is not the bottom line.

[tbyers11]

I think as knowledgeable Duke fans that (mostly ) populate this board, we are undervaluing, or not putting a lot of stock in, the computer polls (Pomeroy, Sagarin and the like) because of our recent poor showings in the tournament. I agree that I am not sold that we are the 2nd best team in the country, but I am rather optimistic about our chances due to improved interior defense and will be really psyched if our interior offense continues to improve.

On a national poll/commentator level, I think that Duke is not given the perception of a #2 or even #6 (AP) team partly because of 1) recent tourney performance, 2) laying an egg in our highest profile non-con game @ Georgetown and 3) the lack of ranked teams in the ACC this season (the Duke only has a gaudy record b/c the ACC sucks camp).

UNC's awful season (that is fun to type) has really lowered national perceptions of the strength of the ACC. While the ACC is not as top heavy as the Big East (having 4 teams in basically the top 10 all season long has helped the Big East rep), I think that the ACC depth is better than the Big East. The top 4 Big East teams as a whole may be stronger than the top 3 ACC teams, but I think teams 4-8 in the ACC are much better than 5-10 in the Big East.


Right on topic with this thread, Vegas Watch has written a really good piece with a simulated 2010 tourney (h/t Rush the Court), based on their modified version of the Pomeroy rankings which factor in recent point spreads and performance relative to the point spreads. The simulation results with Duke as the second most likely to win at 16.8% aren't that different from Pomeroy. They do bump up Purdue and Kentucky a couple of notches and take Wisconsin and Syracuse down a couple.

The interesting part to me is comparing the sim results to current Vegas odds to win the tourney and figuring the return on investment (ROI) a bet would get. Duke has the second highest ROI at 170% of the 15 teams with greater than 1% chance to win. Only Missouri has a higher ROI and they only have a 1.6% chance to win. The only other teams with a positive ROI in this scenario are Purdue, WVa and Wisconsin. The most negative ROI's belong Kentucky, Syracuse, Villanova and Michigan State.

Some interesting food for thought in the numbers vs perception debate on how good Duke (and other top teams) are this year.

[superdave]

Right on topic with this thread, Vegas Watch has written a really good piece with a simulated 2010 tourney (h/t Rush the Court), based on their modified version of the Pomeroy rankings which factor in recent point spreads and performance relative to the point spreads. The simulation results with Duke as the second most likely to win at 16.8% aren't that different from Pomeroy. They do bump up Purdue and Kentucky a couple of notches and take Wisconsin and Syracuse down a couple.

This poll last year had Duke with better than 7% chance of winning. Nova was not even on the list, and we all know how poorly that turned out for us.

[Troublemaker]

As I've written before, it would be nice if the computer guys (Pom, Sagarin) would do home / road splits for their ratings. I think Duke would rank a little bit lower in "away from home" ratings, but I do like this Duke team a lot relative to the competition. The field is just weak.

Anyway, I wanted to post the current national championship odds from a well-regarded online book. Just to add to the discussion. Sorry, the formatting won't be great and you can ignore the first column. For those unfamiliar with odds, "Duke +1615" means you bet $100 to win $1615 on Duke.

2451 Kansas +185
2452 Kentucky +525
2453 Villanova +1215
2454 Texas +1215
2455 Syracuse +815
2456 Ohio State +1615
2457 Duke +1615
2458 Purdue +1615
2459 Michigan State +2550
2460 West Virginia +3050
2461 Georgetown +3550
2462 Kansas State +3550
2463 Connecticut +4050
2464 Tennessee +6050
2465 Pittsburgh +6050
2466 Washington +6050
2467 Xavier +6050
2468 Illinois +6050
2469 Wisconsin +6050
2470 Baylor +6550
2471 California +6550
2472 Louisville +8050
2473 Notre Dame +8050
2474 Gonzaga +6050
2475 North Carolina +8550
2476 Mississippi State +8550
2477 Memphis +8550
2478 Georgia Tech +8550
2479 Clemson +8550
2480 Oklahoma State +8550

[loldevilz]

I think one of the issues is that Duke fans don't put a lot of stock in home wins. If kentucky or kansas played us at cameron they would have a very difficult time pulling out a W. Unfortunately tournament games are played at neutral sites and not at home.

[jipops]

As I've written before, it would be nice if the computer guys (Pom, Sagarin) would do home / road splits for their ratings. I think Duke would rank a little bit lower in "away from home" ratings, but I do like this Duke team a lot relative to the competition. The field is just weak.

Anyway, I wanted to post the current national championship odds from a well-regarded online book. Just to add to the discussion. Sorry, the formatting won't be great and you can ignore the first column. For those unfamiliar with odds, "Duke +1615" means you bet $100 to win $1615 on Duke.

2451 Kansas +185
2452 Kentucky +525
2453 Villanova +1215
2454 Texas +1215
2455 Syracuse +815
2456 Ohio State +1615
2457 Duke +1615
2458 Purdue +1615
2459 Michigan State +2550
2460 West Virginia +3050
2461 Georgetown +3550
2462 Kansas State +3550
2463 Connecticut +4050
2464 Tennessee +6050
2465 Pittsburgh +6050
2466 Washington +6050
2467 Xavier +6050
2468 Illinois +6050
2469 Wisconsin +6050
2470 Baylor +6550
2471 California +6550
2472 Louisville +8050
2473 Notre Dame +8050
2474 Gonzaga +6050
2475 North Carolina +8550
2476 Mississippi State +8550
2477 Memphis +8550
2478 Georgia Tech +8550
2479 Clemson +8550
2480 Oklahoma State +8550

How is North Carolina still in that list?

[MADevil30]

This may be a little tangential, but one of the things that I think drops us down in the human polls compared to the computer polls is the thinking that "Duke plays poorly on the road and wins a lot at home, but none of the NCAA tournament games will be played in cameron." A bunch of ESPN guys and others have written this and it makes so sense to me. No NCAA tournament games will be played in any other team's home gym either! This argument completely ignores the fact that we are 4-0 on neutral courts (you know, the type all the NCAA games WILL be played on), including a win over a very legit Gonzaga team. I can't imagine that our neutral site record wont look even better after the ACC tournament. So count me amongst those who don't buy that our "road woes" (aka only two losses) will hamper our chances at the end of the year.

[hc5duke]

How is North Carolina still in that list?

same reason people could buy gm stock the day before it filed for bankruptcy?

[RoyalBlue08]

I have two points:

1.) That Kansas is a sucker bet...wow! better than a 2-1 favorite to win the thing, really?

2.)I think Duke (and the ACC as a whole for that matter) do a very good job with pre conference scheduling by avoiding those schools at the very bottom of DI that will drag all your numbers down. But do I think that if Duke wins a bunch of games against mediocre schools early it makes them better than teams that win games against really terrible teams. Not really. I think it is much more informative to look at how you have done in your own conference and against other Big 6 and mid major schools. In other words, I understand while the computers have us at #2, but I don't honestly think we are the second best team in the country. Just my opinion obviously.

[GODUKEGO]

Sagarin ratings
http://www.usatoday.com/sports/sagarin/bkt0910.htm

Latest RPI
http://www.realtimerpi.com/rpi_Men.html

Latest AP
http://nmsn.foxsports.com/cbk/polls

[Olympic Fan]

As far as why Duke doesn't get the same respect from humans as from computers, you have to remember that according to the computers we've been similarly dominant almost every season. For example, according to Sagarin, we were #4 in 2005, #1 in 2006, #9 in 2008, and #6 in 2009.

Actually, your example indicates that Duke gets almost exactly the same respect from humans as from the computers.

In 2005, when Duke finished #4 in Sagarin, Duke finished #3 in the final AP poll

In 2006, when Duke finished #1 in Sagarin, Duke finished #1 in the final AP poll

In 2008, when Duke finished #9 in Sagarin, Duke finished #9 in the final AP poll

In 2006, when Duke finished #6 in Sagarin, Duke finished #6 in the final AP poll

So it seems that before THIS season, Sagarin (and RPI and Pomeroy) have pretty much agreed with the human voters. I agree that you can't use the "dork" polls to predict one-game results -- such as in the NCAA Tournament. But over the long term, the computers pretty much agree with the human voters -- at least as far as Duke is concerned.

[Dr. Rosenrosen]

I see the computer rankings as more reflective of a team's net potential. That is, given past performance and a variety of statistics, computer rankings say this year we have great potential to go all the way.

Whereas the traditional polls are more refelctive of writers' and coaches' perceptions of each team at the moment. Why else would a team move up and down from week to week? If you really believe Nova is #3 in the country and nothing drastic has changed about the team, why all of a sudden should they be bumped down 3 or 4 spots in just one week, only to earn their way back up as teams above them lose. It's a very short term focus. Which may in fact be more predictive of how the teams might play "in the moment" during tourney time.

Which is really to say, I have a hard time reconciling the computer rankings. They're nice to look at but... And honestly I have a hard time with the regular polls, too.

In the end, isn't it awesome that it all gets settled on the court!

[Troublemaker]

How is North Carolina still in that list?

Believe it or not, I think people are still betting them (and hoping UNC can sneak into the tournament). It makes sense for books to offer odds on as many teams as possible as long as the odds are in their favor.

I have two points:

1.) That Kansas is a sucker bet...wow! better than a 2-1 favorite to win the thing, really?

They're all bad bets with negative expected value, imo. The books wouldn't be doing their jobs correctly otherwise. But you're right, the Kansas odds seem especially egregious, because the Jayhawks are probably the most heavily bet on team this season and those +185 odds really punish the bettor. I mean, I think Kansas is probably the best team, but they'd have to win the championship 35% of the time just for you to break even at +185. Kansas isn't that good, imo.

[Kedsy]

In 2008, when Duke finished #9 in Sagarin, Duke finished #9 in the final AP poll

Actually, I committed a typo. We were #7 in Sagarin in 2008, not #9. But that shouldn't detract from your point.

[91_92_01_10_15]

I've been thinking all season that it would be interesting to see how Pomeroy's predictions for individual games compare with Vegas odds. From my own observations, probably 25 games or so, I'd estimate that Pomeroy has been right against the spread 75% of the time.

Don't get me wrong. I'm not really a bettor or anything like that. I'm more interested in using any data on this as a measure of the accuracy of his predictions, and the implication that may have as to the accuracy of his ranking system in general.

It seems obvious to me that someone must have done some research on this because of the obvious potential financial implications; however, maybe they just aren't sharing it.

[Kedsy]

I've been thinking all season that it would be interesting to see how Pomeroy's predictions for individual games compare with Vegas odds. From my own observations, probably 25 games or so, I'd estimate that Pomeroy has been right against the spread 75% of the time.

Don't get me wrong. I'm not really a bettor or anything like that. I'm more interested in using any data on this as a measure of the accuracy of his predictions, and the implication that may have as to the accuracy of his ranking system in general.

It seems obvious to me that someone must have done some research on this because of the obvious potential financial implications; however, maybe they just aren't sharing it.

I haven't studied it for Pomeroy, but years ago I had my own computer system that I used to place bets with real bookies (only a few bets in Vegas, mostly just my corner bookie). The answer is you can make a little money if you play smart, but it's nowhere close to 75% because the real spread is almost never what you see in the newspapers. In most of the games where there are large disparities between Pomeroy and the spread (let's say 2 points or more -- you don't have much of an advantage with a one point difference), either the spread moves quickly away from the early spread (which was probably close to what you see in the paper) or the disparity is caused by some piece of information (e.g., an injury) that the bookies factor in but the computer generally doesn't.

Put another way, if Pomeroy says team A will win by 5 but the line you see is only 2, when you call your bookie to place a bet the most likely results are either (a) he won't let you lay 2 but instead will insist on 4 (give or take a half point); or (b) will let you lay 2 but only because team A's sixth man sprained his ankle in practice the day before.

[Trooper]

I agree that the fact that our home court advantage is greater than most any school, so our numbers are skewed vs. some other teams. This gives us a bit of a bump in the computers, but I don't think that's the major reason we are always high in these ratings. We just happen to be in a year where our home/road splits are unusually large (we're normally great on the road too), so naturally our minds are floating to that issue.

Another factor that undoubtedly plays a role is Coach K's intelligent scheduling, which eliminates what I'd call the "computer bottom feeders". People have correctly brought that up, but I still don't think this is the primary difference.

I think the biggest issue we see is due to how amazing of a motivator Coach K is. Most years, Duke is one of the 10-15 best teams in the nation (often much better). And every year I see nearly all of those other teams have 6-8 games where they simply don't come to play. They don't necessarily lose those games, but they may beat a far inferior team by 2-3 points instead of 15, and at times they will get upset. For Duke, that happens 2-3 times a year (NC State & G'Town this season). Simply put, Coach K gets his guys to play their hardest in nearly every game...even against teams that the players probably expect to beat. That's an amazing ability he has as a coach. It has the net effect of increasing our adjusted scoring margin relative to others, which is a huge driver of every computer ranking (especially pomeroy and sagarin predictor where we always fair the best).

The only downside of this is that we lose this advantage come tourney time, because at that point every team is playing as if it's their last game, because, well, it could be.

If we were to play Kansas on a neutral court tomorrow, I'd say they'd be favored in that game by maybe 4.5pts or so (at least in my mind). However, if you told me both teams were playing the 42nd ranked team in the country this week on neutral courts, I'd probably wager that Duke would win by more points. Not to knock Kansas or Bill Self in particular, but you get my point...

[Jderf]

Not perfectly relevant, but I always wonder what these rankings would look like if you cut out the last 1:30 or so of every game, where the play usually breaks down (whether a tight game or a blowout) and the game/stats become a crapshoot. Take our G'town game this year. We were much further down than the final score would suggest, only because we cranked up our score when G'town wasn't really trying anymore. The computer rankings weigh these points equally, which I feel they shouldn't.

Although I guess you could argue that those endgame situations are just as much a part of the sport as the rest and need to be calculated equally. Still though, I'm not convinced.