for the top 5 teams
Memphis - 76%
North Carolina - 79%
Pittsburgh - 63%
Connecticut - 64%
Duke - 67%
Carolina is the most apt to perform worse after performing well. Etc.
how often does a team do better after doing worse, and vice versa
yes, I have made a stat for this (% of sinusoidality)
Upset Watch!
"Upset" watch for today.
(Games that Kenpom predicts to be the opposite of what the polls say---just click any team to see the predicted score of their future games).
-Unranked Davidson over #22 Butler, 68-66
-Unranked Georgetown over #11, 73-68
-Unranked USC over #19 Washington, 72-71
-Unranked Texas over #2 Oklahoma, 73-72 (hehe!!)
-Unranked St. Marys over #23 Utah St, 67-65
Let's go Texas!
Memphis, #1?
I have an immense amount of trust in kenpom.com and the LRMC's ranking systems. In head-to-head, I will almost always choose what kenpom.com says.
But is Memphis better than North Carolina? Kenpom.com would like you to believe so. Personally, I'm pretty sure their stats are this way because they've done better than predicted four games in a row (they've won by an average of 12 points/100 possessions better than predicted in these last four games). Unfortunately, I'm afraid Kenpom.com adjusts too much for recency.
But on Scout.com, there is talk of them being ranked higher because of their exceedingly good performance against cupcakes, and "beating no real competition so far." This sounds like what many people felt about Kansas and Davidson about this time last year.....hm......
Anyways, I made a graph showing HOW cupcake-eatey Memphis is. This predicts how well they will do against specific competition (for reference, ~.85 and up is NCAA-atlarge tourney level competition, and .95 and up is top 10 teams).
(clik to enlarge)
(clik to enlarge)
I have added their predicted performance againinst Louisville and North Carolina for reference. According to their performances in the past, Kenpom only overpredicts Memphis by about 3.6 points per 100 possessions (which in a game against North Carolina, would translate to about 2.75 points).
And since Memphis is predicted to beat North Carolina by less than 1 point on a Neutral Court(in Kenpom's prediction formula), yes, Memphis is overrated.
But BARELY.
Predicted NCAA performance
Here are the top 10 kenpom.com teams, ranked by my predicted bad day, average day, and good day performance (in predicted point margin against an average team)
click the underlined headers to sort: http://lazydrumhead.googlepages.com/days.html
The REAL Winehouse Factor
Kenpom.com and basketball prospectus have a statistic called the "winehouse factor." Unfortunately, it is based on statistics that do not have all that much to do with actual consistency.
It simply measures a teams' inconsistency in scoring throughout their schedule. However, if a team plays a BAD team, they should expect to do very well, and vice versa. In order to account for this, the better measure is the standard deviation in Actual-Expected efficiencies. And using only conference play is stupid--it gives us a sample size that is very small.
There are actually two more factors we must consider to get an accurate number of a teams' consistency, and only one that I can factor in (at the moment). The first is that offense and defense BY THEMSELVES are relatively unimportant. Examples being: Wright State has the 10th best adjusted Defense, but is 83rd overall in kenpom.....also: Notre Dame has the 6th best adjusted Offense and is 36th overall in kenpom. As we all know, it is the COMBINATION of these two that makes a team good. (You can have a 'bad' defensive day and win by 30). So instead of using offensive and defensive efficiencies, we should just use Efficiency Margins. (That is, one team's points per possession minus their opponents').
The second is this: if one team has a high standard deviation, but plays teams with high standard deviation, then their actual consisency is better than the numbers suggest. Unfortunately at this time I don't have the time to calculate this adjustment.
So let's look at consistency:
I call it the Rice-Factor. Tyrese Rice (and Boston College) is the reason for some of the most inconsistent team play in recent college basketball history (Beating Duke and Carolina, losing to St. Louis and Harvard).Rice Factor: Standard Deviation of Actual Minus Expected Offense or Defense
Here's a random list of teams I think are inconsistent, and ones that John Gasaway calculated to be inconsistent.
Georgetown: .203 PPP
West Virginia: .201 PPP
Notre Dame: .174 PPP
Kentucky: .171 PPP*
Boston College: .171 PPP
Duke: .167 PPP
Tennessee: .140 PPP*
Arizona: .127 PPP*
Arizona is inSANELY consistent. John Gasaway getting mad at their offensive consistency is silly.
And poor Georgetown.
Well
Good job Heels.....you knocked Duke off their #1 spot on kenpom.com and took it for yourself.
Now for a fun look at player efficiency vs. point margin with Carolina's starters.
You will see two trends:
as the 2-point men are more efficient, the scoring margin goes up (makes sense).
as the 3-point men are more efficient, the scoring margin goes down. This is pretty easily solved, as we all know that 3-point shots are used most often in comebacks.
One thing is pretty clear: the heels need the Hansbrough (R^2=.39 !!!!!)
I finalized a prediction method today, using Kenpom's stats and how teams perform under those given stats. Maybe tomorrow I'll get the Carolina-Duke prediction finalized.
For now, here are my 3 favorite game-predictors, ranked by how accurate I feel the scores are in general.
2. LRMC: Carolina wins.
My normal distribution formula gives:
Duke-59%, Carolina-41%
Sad teams
Just some guessin:
#23-Georgetown - (and, against #1, #2, #4,#5, and #7 they have been outscored by an average of only 4.4 points)
#33-UAB
#32-Southern Cal
#37-New Mexico
#38-Wisconsin
#41-Tennessee
#44-Temple
#46-Oklahoma St.
#49-Houston
#52-Tulsa
#54-Washington St.
#55-Stanford
#56-Notre Dame
#59-Nebraska
#60-Rhode Island
#61-Cleveland St.
#62-Virginia Tech
#64-Northwestern
Now, for the top bubblers:
#7-West Virginia
#12-Purdue
#19-Washington
Now, for some overratedness:
| RPI rk | Kenpom rk | Overratedness | Overratedness % | |
| Dayton | 21 | 75 | 54 | 15.70% |
| Florida St. | 18 | 51 | 33 | 9.59% |
| Minnesota | 25 | 48 | 23 | 6.69% |
| Ohio St. | 19 | 40 | 21 | 6.10% |
| Tennessee | 20 | 41 | 21 | 6.10% |
| Butler | 10 | 28 | 18 | 5.23% |
| Utah St. | 33 | 50 | 17 | 4.94% |
| Xavier | 6 | 20 | 14 | 4.07% |
UNC vs. Duke: game plan
Duke is predicted to win by five on Wednesday. I hope not. But it sure is looking that way.
But lets look at the game-makers.
For UNC:
#1: Floor defense- This is Carolina's biggest factor that makes them better or worse than expected. In order to make up for those five points, Carolina would need to hold Duke to around 41.5 eFG% (Field Goal percentage with an extra 50% weight for 3-pointers). Unfortunately, I predict Duke to shoot ~50% eFG (which, I would predict on its own to Duke an extra six points).
#2: Turning the ball over- This is UNC's 2nd biggest x-factor. In order to make up for the five points they are predicted to lose by, they would need to turn the ball over on only 14.5% of the time. Unfortunately again, they are predicted to turn the ball over on ~22.5% of their possessions. This alone would give Duke an extra nine points.
For Duke:
#1: Floor defense: - Duke's defense correlates to changing the score almost as much as Carolina's. With the predicted 48% eFG that Carolina will likely shoot, Duke would be predicted to lose 3 points of their predicted 5 point margin.
#2: Offensive rebounds: - This is their 2nd- highest x-factor. By this alone, they are predicted to lose 2.3 points of their predicted 5 point margin.
Honestly, unless Roy makes some defense and ball-handling changes for Carolina, this game looks favored in Duke's direction. But as we all know, rivalries change the way things work.
Really? I mean, really?
1) Wake Forest....really? 27 points? To Miami??? Really?
2) Duke.......really? 27 points? To Clemson??? Really?
3) Marquette....really? South Florida?
Selection Sunday Predictions....
I know this is a long ways off, but I enjoy using kenpom.com's expected record (often vs. realtimerpi.com's rankings) to figure out who might win, and what they'll be seeded. I feel like this year, the Big East will suffer from ACC's illness last year: having too tough of a conference schedule.
Bubblers: -these will be up in the air-
Kansas St.
Ohio St.
Utah
Georgetown*
Syracuse
Southern Cal
Lucky At-Larges:
Virginia Military
Utah St.
Siena
Snubs:
Notre Dame
Underseeded/Sleepers:
Gonzaga
Georgetown*
West Virginia
Overseeded:
Connecticut (barely)
Oklahoma (barely)
Xavier (barely)
Clemson
Michigan St.
This is very preliminary, and humorously far-removed from March, but I would like to know how close to this the brackets tourn (pun accidental) out.
*=couldn't decide
WFU v. Duke, Part II: Weaknesses
Duke's Weaknesses:
#1: Shot defense
This may be counter-intuitive, as Duke has one of the best AVERAGES for defensive efficiency in the nation (81.2 points / 100 possesions, #2). However, examine this:
Duke's performance on guarding the shot (or equally, how well their opponents shoot from the floor) has an extremely high correlation (as far as basketball stats go) to their final score. So, the BEST way to overcome Duke's efficiency is through good 3-point and 2-point shooting. Similarly, but not as highly correlated (R^2=.20), is an opponents' defensive rebounding %, or:
#2: Duke's Offensive Rebounding
Also, if Kyle Singler gets in foul trouble, Duke loses their leader in points, assists, rebounds, and steals.
Nothing Wake Forest does correlates as well as Duke's 'weaknesses,' but a couple of things stand out:
1) Beating them to the foul line is a plus.
The Virginia Tech game, Wake's only loss, was the most lopsided Free-Throw Rate game I have seen with any good team. While Wake had a miserable 18.5, VT had an insane 74! Both of these were by far the worst on Wake's schedule. In general, this is their highest-correlated 'changer of expected point margin'.
That's about it.
Duke v. WFU, UPDATED AGAIN
COMPLETELY UPDATED: WOOPS. I thought the game was at Duke. And I had entered the pace incorrectly.
Accuscore: Wake, 78-76, (60%-40%)
Kenpom: Duke, 74-71, (63%-47%)
Nathan: Wake, 75-74, (49%-51%) -- I know, it's backwards, just deal with it.
16-8-4-1
Based on the predicted NCAA tournament bracket by Joe Lunardi (http://sports.espn.go.com/ncb/bracketology)
Using Kenpom.com's data. Let's just say Pittsburgh has it rough.
16: Wake Forest-Georgetown, Memphis-Oklahoma, Pitt-Gonzaga, Xavier-UNC, Duke-West Virginia, Purdue-Louisville, Connecticut-Arizona St, Kansas-Michigan State
8: Wake Forest-Memphis, Pittsburgh-UNC, Duke-Louisville, Arizona State-Kansas
4: Duke-ASU, UNC-Memphis
2: Duke-UNC
1: Duke
That would be a crazy tournament. There were several games that had teams ranked one spot over the other in these matchups.
One problem with kenpom.com
Is that teams play towards a specific point margin, not a specific efficiency margin, and especially not a specific point margin. A good example of this is the Notre Dame - UNC game
UNC's expected defensive performance against Notre Dame should have yielded 100.2 points / 100 possessions. Instead, they allowed 119.8 points / 100 possessions. This is a factor for why UNC's defense is as low (12th) as it is. The problem was, UNC didn't play defense simply because they didn't have to! If your offense has a very good day, and you are up by 20 or 30, there isn't a great need for playing your best defense. While this isn't smart when you're up 5 or 10, when you're up 20 or 30, teams can play more relaxed (reducing injuries) and can allow bench-players to have more time (giving them experience).
As it just so happens, UNC was having an insane offensive day: 140.5 points per 100 possessions! UNC began dominating Notre Dame by being up about 18 or so. At one point, the margin got down to 11 or so, but UNC answered with an 'offense-for-defense' and boosted their margin more. What does this tell us:
In perhaps a small amount, UNC's offense is as overrated as their defense is underrated. A better representation of a team's performance would be their efficiency margin. Carolina was probably predicted to win by about 10 points per 100 possessions, when in actuality they won by around 20 points per 100 possessions.
As long as the teams' efficiency margins are adjusted for strength-of-schedule, I think this would be a better rating system than kenpom's, as it involves what goes through the players' and coaches' heads. Scoring margin, as we have discussed, is a false indicator of team-matchup winners, and therefore would work less.
nathan's sweet 16
if the brackets were to be split up where these teams were the appropriate 1-4 seeds, this would be my sweet 16, elite 8, and final four
| 1 | Duke |
| 2 | UNC |
| 3 | Gonzaga |
| 4 | Pittsburgh |
| 5 | Georgetown |
| 6 | Missouri |
| 7 | Connecticut |
| 8 | Arizona St. |
| 9 | Memphis |
| 10 | UCLA |
| 11 | Louisville |
| 12 | Wake Forest |
| 13 | West Virginia |
| 14 | Oklahoma |
| 15 | Purdue |
| 16 | Xavier |
Clemson @ UNC
iHistorically, Clemson has a 0% chance of winning. They have never won in the Dean Dome.
But here are the computer-predicted scores/% chances of winning.
Accuscore: UNC-90 (87% win), Clemson-75
Kenpom.com: UNC-88 (84% win), Clemson-77
Nathan's win%: UNC: 70% Clemson: 30%
% chance of final margin being within __ points-
1: 4%
3: 12%
6: 24%
9: 35%
12: 47%
13: 51%
15: 58%
Also, I have a new predictor for 3 of kenpom's four factors:
| eFG % | TO% | OR% | DR% (1-Opp OR%) | PPP | |
| North Carolina | 52.4 | 19.6 | 44.1 | 68.0 | 1.100 |
| Clemson | 48.8 | 21.0 | 32.0 | 56.0 | 0.957 |
We'll see how it goes! UNC just has to take advantage of their rebounding advantage.
Right now....
Right now, Duke is the best team in the nation. So who has the best chance of beating them?
There you have it. UNC has a higher % chance of win because the normal distribution assumes that Boston College was an off-day, and Michigan State was generally a pretty good day.
| Predicted Point Margin | % chance of neutral win | ||
| 1 | North Carolina | -2.15 | 48.50% |
| 2 | Gonzaga | -1.75 | 48.13% |
| 3 | Pittsburgh | -3.6 | 43.70% |
| 4 | Georgetown | -3.33 | 43.60% |
| 5 | Arizona St. | -4.25 | 41.95% |
| 6 | Wake Forest | -4.21 | 41.84% |
| 7 | Missouri | -4.44 | 41.30% |
| 8 | Connecticut | -4.25 | 40.42% |
| 9 | West Virginia | -5.45 | 40.24% |
| 10 | Memphis | -5.25 | 38.76% |
| 11 | Louisville | -5.71 | 38.47% |
| 12 | UCLA | -5.43 | 37.99% |
There you have it. UNC has a higher % chance of win because the normal distribution assumes that Boston College was an off-day, and Michigan State was generally a pretty good day.
Subscribe to:
Posts (Atom)