Dave Sez

Its interesting, I'm having a free ACC Tournament Bracket contest on my site, and in the analysis of the picks thus far, the numbers play out pretty close to the log5 predictions...

http://www.scacchoops.com/forms/tt_acc_tourney_analysis.asp

If you want to participate in the contest, go to SCACCHoops.com and click on the brackets.

scacchoops.com wrote:

Its interesting, I'm having a free ACC Tournament Bracket contest on my site, and in the analysis of the picks thus far, the numbers play out pretty close to the log5 predictions... http://www.scacchoops.com/forms/tt_acc_tourney_analysis.asp If you want to participate in the contest, go to SCACCHoops.com and click on the brackets.

I filled out my brackets!

Dave wrote:

25% indeed.

C'mon, Dave, you're not starting to doubt the ability of these stats to perfectly predict future events are you? The numbers know everything....

My picks are in. I'll see all you losers from the Winner's Circle on Sunday. Go Hoos!

Dave, I think you're missing the point. Virginia's gotten this far by exceeding their ability - the luck factor (in which you might include 'heart' or something else intangible) is extremely high. Virginia is not, by any means, a very good team, and although they've beaten NC State and Duke all three times they've played them, a sample of 3 games is worthless. Pomeroy's ratings are not, and neither are the Log5 predictions. So while it's easily possible that Virginia takes care of business once more, the odds of them doing so based on their past performance is not as high as their record makes you think. Virginia hadthe lowest SOS within the ACC. Duke had the highest. The true numbers don't lie.

spser wrote:

Dave, I think you're missing the point. Virginia's gotten this far by exceeding their ability - the luck factor (in which you might include 'heart' or something else intangible) is extremely high. Virginia is not, by any means, a very good team, and although they've beaten NC State and Duke all three times they've played them, a sample of 3 games is worthless. Pomeroy's ratings are not, and neither are the Log5 predictions. So while it's easily possible that Virginia takes care of business once more, the odds of them doing so based on their past performance is not as high as their record makes you think. Virginia hadthe lowest SOS within the ACC. Duke had the highest. The true numbers don't lie.

Oh, but "true" numbers lie all the time. Look at Pomeroy's ratings. Virginia is, and has been, the ninth best ACC team according to his ratings. Does that really seem accurate? Sure, it's possible, but I think any impartial observer of the ACC over the past couple of months would say that Virginia is at least better than Clemson, Florida State and Georgia Tech (and they are indeed 3-0 against those teams). They might be better than BC as well.

The problem is that UVA had a bad stretch back in December when they lost two ugly games in Puerto Rico. Those games killed Virginia's power rating and it's never really recovered, despite all the good wins they've accrued.

Any rating system has its flaws. I think Pomeroy's is particularly good, but this year it has done a lousy job rating the Cavaliers. It happens.

IMO, log5 principles are effective in analyzing past performance, but can't be truely relied on to forecast future outcomes, especially for something as luck and emotion filled as college basketball.
But, that doesn't mean it ain't fun to look at!

Re: the "Virginia had the lowest(-rated) SOS within the ACC" argument -- there is, in my mind, no good way to rate the relative strength of unbalanced conference schedules. The most common method -- adding up the conference records of your opponents -- is terribly flawed because teams that finish higher automatically have "weaker" schedules by this perception (because they handed a loss to most of the teams they played). It's okay to simply compile opponent records when calculating the relative SOS for the whole country because we have 300+ Division I teams playing close to 30 games a year and the overall SOS-decreasing effect of a win shouldn't affect your rating that much (not that they calculate SOS that way -- the RPI considers opponents' opponents' records to try to get a more objective look at the relative schedule strength of a team). In a closed system of 12 teams who play only 16 games against each other, every win negatively impacts the so-called strength of schedule because it makes your victims look worse. If you're going to use the composite record of your opponents, then you have to compensate the top teams for being so good -- and whatever number you use to pad their stats will likely be fairly arbitrary. There are other methods that use power ratings, RPIs, and things like that, but all of them have their flaws as well.

Of course if we still had the double round-robin/balanced schedule, no one would be arguing strength of conference schedule. This discussion is caused solely by the fact that some teams play team A once and team B twice where others may play team A twice and team B once (or both once. Or both twice. Etc.). Moral of the story: conference expansion is bad.

A last word: in this era of unbalanced schedules, you just can't say conclusively which teams had an easier road through the conference slate compared to others. The only thing teams can do is play everybody their schedule tells them to play and try to win more games than everybody else in the conference. By that measure, Virginia did its job better than 10 of the other 11 teams in the conference (and tied the other). That achievement doesn't lie.

Jerry -- you say eloquently what I have been thinking with respect to all those who are whining about UVA's in-conference SOS. UVA has no control over who was on their schedule -- they had control only over what happened on the court -- and they took care of business in that regard.

If you might allow me to paraphrase your post: scoreboard.

Jerry does make some good points. If you are simply going to use in-conference record as a way to measure strength of schedule, at the very least you need to subtract the team's record. So for example, if UVA's opponents were 40-50 (I don't know what it really is), you'd subtract 11-5 from that so that Virginia's impact is eliminated. That way, the good teams aren't punished.

But even that's not as accurate as it could be. It should also take into account where the games were played. It may not matter if you play the best or worst teams at home or on the road - you'll likely lose or win against those teams in either location. It certainly does matter with the teams closer to your team's ability.

The bottom line is that I think that Virginia-had-it-easy meme is grossy overstated and not nearly so true as most think.

The bottomer line is that everyone plays by the same rules and Virginia came out tied for the best record. You still have to play the games, and hell, UVA even dropped two games against "bad" teams and still came out on top.

Disclosure: I'm a grad of both UVa and Duke. Take that for what you will.

My take is that Virginia has been both good and lucky, but the problem is that they've coasted by on what appears to be so much luck that it's difficult to split the two. Some things to note:

Virginia is 6-3 in games decided by 5 points or less.

Virginia is 3-5 on the road in calendar year 2007. The 3 wins were
- Clemson by 1 immediately after the Tigers' heartbreaker at Duke
- Maryland by 4 in the last game before the Terps went on their tear
- NC State in the first game they got Atsur back

Yes, UVa is 3-0 against Duke and NCSU. But remember that NCSU didn't have Atsur in one of those games and had just gotten him back in the other. UVa required some last-second heroics against Duke to send it into overtime (along with a 3:1 FTA advantage in regulation), and then a prayer by Singletary to pull it out in OT.

UVa is solid. Coach Leitao brought that UConn toughness to C'ville. But the Cavs have slid through a lot of tough spots to end up where they are. We'll find out soon enough if that was a product of "mental toughness" or just dumb luck soon enough.

I'm a Duke grad (BSE) and UVa grad (MS) as well.

Featherston compiled the opponents W-L records and then subtracted off the team's own record, just as you suggested, Dave. Here's the article with a bit of commentary, scroll to the bottom.
http://www.dukebasketballreport.com/articles/?p=22084

I'll just post the results here for discussion. It does not take into account road/home, or injuries, etc.
1. N.C. State 127-113
2. Duke 126-114
3. Miami 123-117
3. Clemson 123-117
5. Virginia Tech 121-119
6. Florida State 119-121
6. Wake 119-121
8. Georgia Tech 118-122
8. Maryland 118-122
10. Boston College 116-124
11. North Carolina 116-124
12. Virginia 114-126

Matt reran the numbers after Thursday's games. You can see the results here - http://www.davesez.com/ACClog5_070309.pdf

With Maryland and Duke out, the Heels are even bigger favorites. One nice thing for me is that UVA's chances went from less than 2% to just under 7%.

Of course, keep in mind that all four games went against the odds yesterday.