Wednesday, March 11, 2009

SOS and Tourney Preview

Let's start with strength of schedule. Like last season, I made a full home/road adjustment, so Duke only gets credit for playing Clemson as the Tigers played at home, and only gets credit for playing Virginia as the Cavaliers played on the road. To start, here's each team's home and road efficiency numbers:

Home
Duke
18.79
North Carolina 18.12
Clemson 14.17
Wake Forest 10.78
Florida State 8.11
Miami
6.38
Maryland 0.84
Virginia Tech -1.19
Boston College -2.28
NC State -3.59
Virginia -6.11
Georgia Tech -7.40

Road
North Carolina 8.50
Wake Forest -0.05
Florida State -0.31
Clemson -0.45
Duke
-1.02
Boston College -2.01
Miami
-6.32
Virginia Tech -7.28
NC State -9.14
Georgia Tech -12.12
Virginia -14.03
Maryland -15.26

Remarkably this year, only one team (Carolina) was on the plus side on the road. Duke had the biggest cap between home and road performance (helped by routs of Maryland, Virginia Tech, and Virginia, hurt by the blowout at Clemson). The top 5 teams in the league were top 5 both at home and on the road, while the bottom was a little bit more spread out. Boston College was the most road immune team - actually better on the road than at home, although by an extremely small margin.

So, next up, here's what everyone's schedule looked like.
Offense (ranked by best opponent offense to worst):
Georgia Tech - 105.43
Virginia - 105.41
Florida State - 104.89
Duke - 104.83
Miami - 104.82
Wake Forest - 104.57
NC State - 103.81
Virginia Tech - 103.77
North Carolina - 103.31
Maryland - 103.30
Boston College - 102.75
Clemson - 102.07

Defense (ranked by best opponent defense to worst):
Virginia Tech - 102.86
North Carolina - 103.09
Maryland - 103.23
NC State - 103.46
Duke - 103.52
Clemson - 103.76
Wake Forest - 104.16
Boston College - 104.27
Miami - 104.59
Virginia - 104.91
Florida State - 105.02
Georgia Tech - 106.97

The Jackets' offensive futility is remarkable considering their opponents' general defensive futility. Credit goes to Carolina considering they faced one of the stronger defensive tests in the league, and still were dominant offensively (their offensive numbers were actually suppressed slightly by their schedule).

Finally, here's the schedule, by best opponent efficiency margin (so toughest opponents) to worst:
Overall
Duke - 1.31
Virginia Tech - 0.92
Virginia - 0.49
Wake Forest - 0.40
NC State - 0.36
Miami - 0.23
North Carolina - 0.21
Maryland - 0.07
Florida State - -0.12
Boston College - -1.52
Georgia Tech - -1.54
Clemson - -1.69

The schedule adjustments moved some teams around, but was not too disruptive. By raw margin, here's how the team stacked up:

UNC
Duke
Clemson
Wake Forest
Florida State
Miami
Boston College
Virginia Tech
NC State
Maryland
Georgia Tech
Virginia

When the strength of schedule adjustments are made, Wake and Clemson swap places, Virginia Tech and BC swap places, and Virginia and Georgia Tech swap places.

On to the tournament. As you can see from the list above, UNC actually got a tough draw out of the 8/9 game - adjusted for schedule, Miami and Virginia Tech were the 6th and 7th best teams in the ACC. Wake's draw is much more favorable - neither NC State or Maryland was one of the 8 best teams.

As you can see from the odds report below, the top 5 teams are comfortably above the rest. FSU and Clemson have lower odds of winning because they have the prospect of facing each other on Friday, but all 5 are above 47% to reach the semis and the other 7 are below 10%. So, without further ado, here's the odds (derived from a log5 analysis using pythagorean winning percentage based on schedule-adjusted offensive and defensive efficiency - yike, that was a mouthful):




Quarters Semis Finals Win
1 North Carolina 100.00% 86.53% 65.53% 41.86%
3 Duke
100.00% 88.98% 61.52% 32.24%
2 Wake Forest 100.00% 84.17% 33.56% 12.87%
5 Clemson 91.08% 47.90% 14.94% 5.79%
4 Florida State 100.00% 51.22% 14.96% 5.46%
9 Miami
60.83% 9.36% 3.41% 0.89%
6 Boston College 70.68% 9.29% 2.60% 0.42%
8 Virginia Tech 39.17% 4.11% 1.11% 0.21%
10 NC State 54.41% 8.03% 1.12% 0.14%
7 Maryland 45.59% 7.81% 0.95% 0.10%
11 Virginia
29.32% 1.73% 0.26% 0.02%
12 Georgia Tech 8.92% 0.88% 0.05% 0.00%

2 comments:

Matt said...

Is your log5 table from ACC play only, or from all games? If so, I find the comparison to my own table very interesting (I did not adjust for schedule). Duke comes out with better chances, at the expense of Wake and Carolina. I do think it's important to consider the unbalanced schedule, but we run into a serious sample size issue. This can be compounded by the single-game anomalies such as the Duke-Clemson and Duke-Maryland games.

What do you think of the observation that the average efficiency in ACC play was 104.5? I think it means that the ACC has generally an offensive heavy environment, and that teams with above average defenses (Duke, FSU) are really quite impressive in this context.

PS: Should Maryland really have a 95% chance of making the finals? Georgia Tech 88% chance of making the semis?

Paul Rugani said...

First, thanks for catching the Maryland and Georgia Tech errors - those were obviously supposed to be 0.95% and 0.88%.

Second, it is ACC only - the schedule adjustment wouldn't work across the raw numbers. The sample size issue is certainly there, but it is whether you use 16 games, and don't balance for schedule, or use two sets of 8 games, with an attempt to balance for schedule. I tried to take that noise out by zeroing out a team's own games (which also may be too much of an adjustment) - so Duke, for example, doesn't get a strength of schedule boost from Clemson's 27-point win (or a SOS hit from Maryland's 41-point loss).

Nonetheless, it is important to adjust for schedule. The difference between Duke's and Clemson's was worth a full win (and the difference between each and average was ~ 1/2 win). I think it's reasonable to make Duke's numbers better by that 1/2 win (and Clemson's worst by that 1/2 win) to try to balance. It's really asking, if they all played the same schedule, what numbers would we see? I'm aware that my method is far from perfect (I am very decidedly not a mathematician), but I think it gets us closer to the actual relative performances of the teams over the year.

The ACC's offensive rating has been consistently high over at least the past three years. 104.5 may be slightly higher than in the past. Usually there are only a couple teams whose offensive ratings (or defensive) are below 100. And yes, it does make FSU and Duke stronger on defense in context.