What's a good early season worth?

If you've been reading the Hoya interwebs since the Memphis game, you may have come across a note that Georgetown has wrapped up the pre-Big East part of the schedule with a 10-1 record, and this is the fifth consecutive season that the Hoyas have started the season 10-1.

Of course, not all 10-1 records are the same.  As I write, Georgetown sits 12th in both polls, and 14th in Ken Pomeroy's ratings.  Meanwhile, the Seton Hall Pirates have run off an 11-1 record so far, but can get not a whiff from either poll and Ken rates them 55th overall.  While the teams have played comparably difficult schedules so far [G'town = 252nd, SHU = 274th], the Hoyas rate high by beating the bad teams by a lot, and for beating (or losing to) some better teams - Georgetown's only loss was to Kansas, while the Hall fell to Northwestern.

But that's not why I'm writing.

What I am actually curious about, is whether early season performance is a useful predictor of what comes later in the year.

The simple answer is "Yes."

I want to be careful to spell out a few definitions before I proceed.

• The data set shown in the figure above is for all Big East teams for the seasons 2006 to 2011 (the six seasons played since the expansion with Conference USA teams)
• Early season refers to games in November and December - I'm not discriminating for whether a game is conference or non-conference here.  Early season is represented on the x-axis
• Later games means just that:  all games played after Dec. 31st. More pointedly, I'm not limiting the discussion to merely how a team performs in a single-elimination tournament (conf. or NCAA).  Late season is represented on the y-axis.
• I'm not going to be using win-loss record but rather Ken Pomeroy's stats [adjusted efficiencies] to evaluate how a team performs.  The stat of interest here I'll call "net adjusted efficiency" which is simply adj. offensive efficiency [aOE] minus adj. defensive efficiency [aDE].  You can find aOE and aDE on Ken's ratings page (free to all).  I'm happy to acknowledge that his "Pythagorean" rating is probably a bit more accurate, but it's also a lot more complicated.

The scatter plot is fairly impressive - using Ken's stats, there is a clear linear trend between how well a team plays early and how well they will play later.

Lots of fancy-pants analysis, what it means for Georgetown, and predicted Big East wins for all teams after the jump.

Tuesday Truthiness

It's normally about this time of year that I make my annual blog post criticizing John Gasaway and his Tuesday Truths column (see here in 2009, and here in 2010).  And thanks to some recent articles posted at VBTN (link) and the HSAC (link), I reckoned it was time I get off my lazy duff and get to it.

But first, some background.

If you don't live in your mother's basement, you may not be aware that there is a very smart gentleman named John Gasaway.  Mr Gasaway himself lives in a basement, but it is the basement of the sports think tank called Basketball Prospectus.  From this secure location, he produces a series of articles called Tuesday Truths, which seek to bring knowledge to the ignorant masses.

To quote extensively from the master:

Over the next eight to nine weeks these 126 teams will play over a thousand possessions each. Half of those possessions will take place at home, and half of them will occur on the road. All of that basketball will be played against opponents that by conference affiliation have been designated as nominal equals in terms of programmatic resources. (Though, granted, a league like the A-10 certainly exhibits some notable diversity in terms of member heft.) And, not least, all of that basketball will take place in increasingly close temporal proximity to the NCAA tournament.

In other words, with all due allowance for injuries and funky scoring distributions, I look at these thousand-odd possessions very closely. And in leagues featuring true round-robin scheduling (Missouri Valley, Mountain West, Pac-10, and WCC, among others), per-possession performance in conference play tells me exactly how surprised I should be in mid-March when the league’s best team in tempo-free terms loses in first round of the NCAA tournament.

This per-possession performance is often termed efficiency difference [Off. PPP - Def. PPP] or net efficiency [Off. Eff. - Def. Eff.], and I'll use the terms interchangeably; the nuance is that eff. difference is just net efficiency divided by 100.  Also, to be clear, per-possession performance is not the same as my own Performance stat, which I continue to champion, albeit alone.


But all is not well between myself and Mr. Gasaway.

Conference efficiency margins and strength of schedule

A couple of my favorite basketball bloggers are big proponents of conference efficiency margin (what I normally call net efficiency).  Efficiency margin is simply the difference between points per possession scored and points per possession allowed.

Basketball Prospectus author John Gasaway, who may have pioneered this stat, offered a pithy explanation of it's utility last month:

Why track per-possession performance in league play?

Over the next eight to nine weeks these 126 teams will play over a thousand possessions each. Half of those possessions will take place at home, and half of them will occur on the road. All of that basketball will be played against opponents that by conference affiliation have been designated as nominal equals in terms of programmatic resources. (Though, granted, a league like the A-10 certainly exhibits some notable diversity in terms of member heft.) And, not least, all of that basketball will take place in increasingly close temporal proximity to the NCAA tournament.

And greyCat over at Villanova by the Numbers has been tracking these stats for the Big East biweekly, and in great detail.  Mr. Cat (I actually do know his name, but I'm not telling) also provided an important caveat before jumping into the numbers with his latest posting

I understand the argument against looking only at (for example) conference games -- the sample is too small, the schedule is unbalanced, the time-frame is too restricted, etc. etc. etc. -- all of which are valid points.

I actually took a look at this very point last season around this time, and it turned out to be more more informative than I realized at the time.  I'll come back to why that is at the end, but first I thought I'd re-run that analysis tonight.

I've compiled the strength of schedule for all Big East teams for conference games played so far (through Monday, Feb 1st), and plotted eff. margin against strength of schedule. Here, I'm using KenPom's Pythagorean rating (0-1), rather than his ranking (1 to 347) for the strength of schedule calculation.  The strength of schedule is weighted for home vs. road games.  The efficiency margin stats come from StatSheet.com.

Here we go (click to enlarge):

Just as we saw last year, there is a strong correlation between conference strength-of-schedule played to date and the efficiency margin posted.  So teams such as West Virginia and Villanova, which currently lead the league in efficiency margin are definitely getting some help from a soft early-season schedule, while Seton Hall is a bit below water thanks in no-small-part to a brutal start from the schedule-makers.

Now a bit of that effect is circular, in the sense that good teams don't have to play themselves, and therefore their schedules are slightly easier on average, while bad teams schedules are harder because they're missing an easy opponent (themselves).

However, in any league with an unbalanced schedule like the Big East, these effects are only one part of the overall scheme that won't entirely even out at the end.  For example, our own Georgetown Hoyas have been rewarded this year with home-and-away series with the current AP-ranked #2 and #3 teams.

Here's a table summarizing the conference season so far, and what's left for each team (rankings are from hardest to easiest):

.                      Current           Remaining          Exp. Final
Team                  SOS    Rank        SOS    Rank        SOS    Rank
Cincinnati           0.825    15        0.902    3         0.868    14
Connecticut          0.878    6         0.862    11        0.870    11
DePaul               0.913    2         0.823    14        0.875    6
Georgetown           0.867    10        0.876    8         0.872    10
Louisville           0.882    5         0.872    9         0.877    4
Marquette            0.890    4         0.819    16        0.858    16
Notre Dame           0.856    11        0.892    5         0.875    7
Pittsburgh           0.873    9         0.878    7         0.876    5 
Providence           0.812    16        0.936    1         0.888    1
Rutgers              0.910    3         0.838    13        0.879    3 
Seton Hall           0.919    1         0.819    15        0.872    9
South Florida        0.873    8         0.860    12        0.867    15
St. John's           0.877    7         0.864    10        0.870    12
Syracuse             0.851    12        0.885    6         0.869    13
Villanova            0.845    13        0.911    2         0.886    2
West Virginia        0.839    14        0.897    4         0.874    8 

Seton Hall has had the pleasure of playing the toughest Big East schedule so far, and are showing the effects with a 3-5 record.  They're path will get much easier, so expect them to make a push towards legitimacy.  DePaul and Rutgers have also had tough early-season slates, but I'd think the best they can hope for is just a step up from the absolute dregs of conference efficiency margin (anything worse than -20 is really bad).

Providence has a brutal second-half schedule, harder than what any team has played so far.  Many had pegged the Friars to be one of the worst teams in the league this year due to the loss of so many players, and that result may still be coming.  Meanwhile the Big East's Big Three (Villanova, Syracuse and West Virginia) have all benefited from relatively easy first-halves and should all see things get a bit tougher the rest of the way.  In fact, Syracuse may be in the best position of the three to win the regular-season conference title because of this.

It's also worth noting that the spread between easiest expected final schedule (Marquette) and hardest (Providence) is about the equivalent of the difference between playing U of Miami or UConn on a neutral floor, respectively (i.e. not really that much).


Finally, returning to that bit about what was so informative about last year's plot (reproduced below).  It turns out that teams that met two conditions went on to the NCAA tournament:

1. They sat above the fitted line, i.e. were better than average, adjusted by schedule strength.
2. They already had positive efficiency margins midway through the season.

So, right now, it looks like there are seven teams in good position for March (Villanova, West Virginia, Syracuse, Georgetown, Louisville, Marquette (!), and Pitt).  There are three other teams that meet only one of those criteria: Cinci, UConn and Seton Hall.

Here's last year's plot (from 24-Jan-09):

Looking back in anger

As the Big East regular season gets under way, it comes time to wonder what we've learned about the Hoyas so far this year.

About this time last season I also wondered what November and December had demonstrated - heading into a home game against Pittsburgh, Georgetown was ranked #1 overall by Ken Pomeroy, with a net adjusted efficiency (adj. off. eff. - adj. def. eff) of + 42. That is, we expected that the Hoyas would outscore the median Div-I team (e.g. Holy Cross or Hofstra in 2009) by 42 points in 100 possessions.

I was so excited that I made one of those fancy-pants graphs to demonstrate visually just how good Georgetown was (click any figure to enlarge):

2008-09 Big East Aerial (31-Dec-08)

You'll need to read this post to understand everything in this figure, but simply:  upper-right = good; lower-left = bad.

The Hoyas went 6-14 the rest of the way.  Obviously, that figure - and the Hoyas early season performance - didn't tell us what was to come, only what had happened so far.

Now I could simply knock out this year's version and let it rest, or I could stare at a bunch of numbers on a spreadsheet and try to understand why the early season performance by the Hoyas was such a poor predictor of the rest of the season.

Let's do both.

First, here's this year's aerial, through games played Monday:

2009-10 Big East Aerial (28-Dec-09)

As conference play gets underway, Syracuse and West Virginia appear to be the class of the Big East, both with fairly balanced teams (offense vs. defense efficiencies).

Georgetown, surprisingly, is a solid third by KenPom's ratings, although the team is highly dependent upon its defense (#1 in the conference) as the offense is merely average.

Following is an enormous cluster of teams, headed by Marquette and Villanova and followed by St. John's, Pitt, UConn, Louisville, Seton Hall, Cincinnati and South Florida.  That's nine teams all grouped together.

Notre Dame - which almost fell off the chart with its extreme of the league's best offense and worst defense - and Providence are slightly behind the peloton, and Rutgers and DePaul hold the final two spots.

A few points to make:

• Note that the scaling on this year's aerial is slightly different than last year's.  The offensive scale doesn't go quite so high and the defensive scale doesn't go quite so low.  For instance, Notre Dame had better adj. off. and def. eff. stats last year, while this plot may lead you to think that their offense is actually better this year.
• The elite teams this year aren't quite as elite as those from last season headed into January.  Since four Big East teams made it to the Elite 8 last year, a drop-off this time isn't unexpected.
• No team last season made it to the NCAA tournament from behind the isopleth that crosses the solid diagonal at OE=107, DE=93.  This year, S. Florida sits just behind that line.  Three teams that were ahead of that line last year (Georgetown, Notre Dame and Cinci) failed to make the tourney, although they were bubblicious until the end.  If that's any sort of bellwether, it would mean 12 Big East teams are capable of making the NCAAs this year (although obviously some will not).
  
• A few teams may look out of place:
• Georgetown's pedestrian offense may seem unusual, but did you know that the Hoyas have finished higher than 8th in off. efficiency in the Big East only two times in JTIII's first five seasons (2006 & 2007)?
• Louisville's offense is better than their defense?  The Cardinals have finished 4th, 2nd, 1st and 1st in def. efficiency in conference their first four seasons.  If the offense can tread water, expect Louisville to improve as its defense does.
  
• Pitt was one of the most lopsided teams last year - they had a great offense and mediocre defense.  It's the opposite so far this year.
• To my eye, the most dramatic year-over-year improvement is S. Florida, but I'm not sure how they'll do without Gus Gilchrist for the foreseeable future.  St. John's is also far ahead of last year.
  

----------------------------------------------------------------------------------

Okay, that was fun, but what about the real issue - how well can we predict the rest of the season from how well a team plays in November and December?

Here's Georgetown's won-loss record for the past 6 seasons in the early part (all games in Nov. and Dec.) and late part (Jan. - Apr. games) of the schedule:

.                Early                  Late
Season       W   L     Net         W   L     Net 
2003-04      9   0   + 28.8        4  15   - 11.8
2004-05      8   3   + 13.7       11  10   +  1.1
2005-06      8   2   + 21.2       15   8   +  7.7
2006-07     10   3   + 23.1       20   4   + 15.7
2007-08     10   1   + 31.9       18   5   + 12.1
2008-09     10   1   + 29.6        6  14   -  3.6

That third column for each segment ("Net") refers to the difference between off. and def. efficiency during those games.  It turns out that the best two W/L records, and the 2nd and 3rd best net efficiencies from early season preceded two epic collapses - yes, this dataset goes back to the final Esherick season, although this is just a coincidence since this is as far back as KenPom's database goes back in time.

Now hopefully, if you're my regular reader or you've bothered to make it this far down the article, you'll want to dig a bit deeper into those numbers above.  For instance, what was the quality of competition like during each early season?  If we adjust the underlying off. and def. efficiencies for that competition, what would the net eff. numbers look like?

Let me tackle those one at a time.  First, I'll re-post those net. eff. numbers with the opponents' average KenPom rating:

.             Early                    Late
Season      Net   Rating             Net  Rating
2003-04   + 28.8   258             - 11.8   63
2004-05   + 13.7   175             +  1.1   55
2005-06   + 21.2   190             +  7.7   49
2006-07   + 23.1   150             + 15.7   48
2007-08   + 31.9   160             + 12.1   59
2008-09   + 29.6   119             -  3.6   61

Clearly, JTIII does not subscribe to Esherick's (and by proxy, his father's) scheduling philosophy.  The 03-04 team was simply getting fat on a bunch of low-major teams early season, and never had its mettle tested until conference play got rolling.  The 08-09 bunch, however, played the toughest early-season schedule so far - remember that the UConn game last year was played before New Year's, so it is included.  The huge drop in net efficiency last year came after the team looked to be able to handle just about anyone early on.

Let's now take a look at each of the individual efficiencies, now adjusted for competition during the segment.  This is analogous to looking at KenPom's Adj. Efficiencies, rather than raw efficiencies.  As I mentioned at the top, when we discuss adj. efficiencies, what we mean is how would we expect the team to perform against an average Div-I team.

Here we go, looking at offense first:

.                Off. Eff.
Season     Early  Late   Diff.
2003-04    103.9   96.2  -7.7
2004-05    106.9  113.2   6.3
2005-06    109.9  117.7   7.7
2006-07    113.8  125.9  12.1
2007-08    119.1  114.3  -4.8
2008-09    117.4  108.3  -9.1

The last Esherick team was the worst offensive team in the group, both early and late season, once you adjust the stats for the level of competition.

One of the great mythical concepts of the Georgetown/Princeton offense is that it is massively complex and takes considerable practice time and game experience to master.  The first three seasons under JTIII certainly give evidence to that, as the team improved from early season to late season.  This wasn't the case in 07-08, where the team was stocked with upperclassmen and operated at a high level even with the step back in conference.  But last year's team showed the biggest overall drop intra-season, and the roster had only 1 junior and 1 senior - surely an improvement should have come.

It's also worth noting that, in spite of the collapse last year, the team was still scoring more than 12 points per 100 possessions more than the 03-04 club.

.                Def. Eff.
Season     Early  Late   Diff.
2003-04     91.6   92.8   -1.2
2004-05     97.3   94.4    2.9
2005-06     92.9   92.6    0.3
2006-07     89.1   88.7    0.3
2007-08     87.2   85.3    1.9
2008-09     82.6   95.0  -12.4

Excluding last season, this table just makes a lot more sense to me.  I can understand offenses going to slumps, for example due to a prolonged stretch of poor outside shooting, but I'd expect that a team's defensive ability will be fairly well established in the first two months of the season, once the level of competition is accounted for.

An important trend, and one that I don't think most fans or analysts are picking up on, is that each Hoya team in the JTIII era has been better defensively than the previous year's team, and had improved from early-season to late-season.  Heading into Jan. 2009, Coach Thompson was beginning to look like some sort of defensive genius.

Then the wheels came off.

If we use the Esherick team as an archetype for one that had lost its confidence, or perhaps better a coach that had lost his team, we still see only a small drop in the quality of defensive play.  Last season, there were no significant injuries yet the defense on the court in conference play not only didn't resemble the early season team, it broke with a general trend over a five-year period.

.                Net Eff.
Season     Early  Late   Diff.
2003-04     12.3    3.4   -8.9
2004-05      9.6   18.8    9.2
2005-06     17.1   25.1    8.0
2006-07     24.8   37.2   12.4
2007-08     31.9   29.0   -2.8
2008-09     34.8   13.3  -21.4

Finally (for tonight), I'll summarize the previous tables with the net adj. efficiencies for the past six seasons.  I think its worth noting that, as much as we like to decry the late season swoon last year, it's worth recognizing that the early season performance was about as good as the Hoyas have played since JTIII arrived, just a touch below the 06-07 team during its run to the Final Four.  The shock wasn't so much how poorly they were playing by the end of the season, but rather how far they had fallen in such a short period of time.

While the Esherick team's collapse wasn't nearly as large, that team simply wasn't good enough to start with to be able to collapse very far.

Tomorrow, I'll dig a bit deeper into the underlying stats, and take a look at what we've learned so far this year.

NIT-log5 and Adjusted Efficiency Predictions

Following up on yesterday's teaser post, it's time to post preview odds for the 2009 NIT.

Contrary to my hopes, I haven't been able to expand my bracket to account for championship chances, so you're just getting finals chances for every team. If the Hoyas win Wednesday and after the first round is over, I'll post complete odds for the rest of the NIT.

Because the NIT is played at campus sites, team's chances of making it to NYC for the semifinals are greatly affected by seeding. First, here's what team's chances would look like if this were the NCAA tournament and home court advantage was not taken into account, via the log5 method:

Seed  Team           2nd Round Quarters New York  Finals
1     Auburn           79.93%   43.19%   19.60%    8.89%
8     UT-Martin        20.07%    4.58%    0.82%    0.14%
4     Tulsa            53.29%   28.60%   12.30%    5.28%
5     Northwestern     46.71%   23.63%    9.43%    3.75%
3     Baylor           37.57%   22.07%   12.20%    5.76%
6     Georgetown       62.43%   43.78%   29.38%   17.48%
2     Virginia Tech    63.87%   24.63%   12.68%    5.50%
7     Duquesne         36.13%    9.52%    3.59%    1.10%

1     San Diego St     81.33%   47.34%   27.67%   15.94%
8     Weber St         18.67%    4.57%    1.14%    0.28%
4     Kansas St        59.62%   30.84%   16.71%    8.92%
5     Illinois St      40.38%   17.25%    7.71%    3.39%
3     South Carolina   52.41%   21.14%    8.65%    3.76%
6     Davidson         47.59%   18.12%    7.01%    2.88%
2     St Mary's        36.30%   18.92%    7.92%    3.51%
7     Washington St    63.70%   41.81%   23.19%   13.42%

Seed  Team           2nd Round Quarters New York  Finals
1     Creighton        78.07%   35.31%   14.01%    6.46%
8     Bowling Green    21.93%    4.16%    0.65%    0.13%
4     Kentucky         58.72%   37.48%   17.95%    9.75%
5     UNLV             41.28%   23.05%    9.06%    4.15%
3     New Mexico       63.69%   32.50%   19.72%   11.58%
6     Nebraska         36.31%   13.54%    6.38%    2.88%
2     Notre Dame       52.37%   28.83%   17.52%   10.30%
7     UAB              47.63%   25.13%   14.71%    8.31%

1     Florida          81.12%   43.34%   27.70%   14.55%
8     Jacksonville     18.88%    4.10%    1.20%    0.25%
4     Miami            66.70%   38.62%   25.45%   13.92%
5     Providence       33.30%   13.94%    6.86%    2.60%
3     Niagara          52.42%   29.91%   12.59%    5.29%
6     Rhode Island     47.58%   26.02%   10.36%    4.12%
2     Penn State       58.05%   27.52%   10.62%    4.09%
7     George Mason     41.95%   16.55%    5.21%    1.63%

Alas and alack, playing at on-campus sites significantly decreases the odds for teams that will likely end up playing a lot on the road.

Here's what it looks like with home field advantage taken into account, including Miami playing at Providence and assuming the higher (better) seed hosts in all second and third round games:

Seed  Team           2nd Round Quarters New York  Finals
1     Auburn           88.76%   61.58%   38.76%   17.28%
8     UT-Martin        11.24%    1.46%    0.16%    0.03%
4     Tulsa            68.66%   25.89%   10.07%    4.21%
5     Northwestern     31.34%   11.08%    4.00%    1.55%
3     Baylor           52.95%   25.35%   12.50%    5.77%
6     Georgetown       47.05%   27.83%   13.97%    8.23%
2     Virginia Tech    76.85%   42.87%   19.63%    8.29%
7     Duquesne         23.15%    3.95%    0.91%    0.27%

1     San Diego St     89.70%   64.40%   48.36%   28.63%
8     Weber St         10.30%    1.15%    0.18%    0.04%
4     Kansas St        74.11%   27.00%   14.09%    7.73%
5     Illinois St      25.89%    7.45%    3.20%    1.44%
3     South Carolina   68.19%   28.72%   10.08%    4.48%
6     Davidson         31.81%   12.68%    2.88%    1.21%
2     Saint Mary's     52.28%   35.16%   12.60%    5.72%
7     Washington St    47.72%   23.44%    8.61%    5.11%

Seed  Team           2nd Round Quarters New York  Finals
1     Creighton        87.58%   52.47%   28.24%   12.67%
8     Bowling Green    12.42%    1.23%    0.10%    0.02%
4     Kentucky         73.56%   35.55%   13.21%    7.03%
5     UNLV             26.44%   10.75%    3.23%    1.44%
3     New Mexico       77.71%   35.44%   21.01%   12.18%
6     Nebraska         22.29%    7.35%    2.28%    1.00%
2     Notre Dame       67.69%   45.92%   27.17%   15.71%
7     UAB              32.31%   11.29%    4.76%    2.64%

1     Florida          89.22%   62.56%   47.87%   25.07%
8     Jacksonville     10.78%    1.39%    0.24%    0.05%
4     Miami            51.97%   22.24%   12.48%    6.82%
5     Providence       48.03%   13.81%    5.51%    2.06%
3     Niagara          67.38%   32.57%   12.53%    5.22%
6     Rhode Island     32.62%   15.16%    3.98%    1.57%
2     Penn State       72.78%   45.52%   16.10%    6.13%
7     George Mason     27.22%    6.75%    1.30%    0.40%

Methodological note: This latter calculation is made using the Pythagorean expectation formula, with an exponent of 11.739. That seemed to produce results very similar to log5, so it's more of an apples-to-apples comparison.


The great advantages of a home team are pretty clear here, as all of the higher seeds look much better. Creighton's chances of making it to New York essentially double, and they're not alone.

In fact, game location probably matters more in the NIT than it does in the NCAA, simply because the teams, except for the automatic qualifying dreck at the bottom of the bracket, are fairly evenly matched in quality.


Finally - not that you need a reminder - efficiency ratings are not destiny.

More on luck

I noticed a discussion of luck and consistency on HoyaTalk tonight, and that thread pointed to an article by John Gasaway on Basketball Prospectus.

Gasaway provides a list of top lucky and unlucky teams, based on winning percentage compared to net efficiency. This is something that I set up to track a while ago, but just have been too busy to get around to posting. Now that I've been scooped, I thought I'd just dump what I've got.

First, to define terms:

• winning percentage is for conference play only, as are all following stats
  
• net efficiency is simply the difference between offensive efficiency (points scored / 100 possessions) and defensive efficiency (points allowed / 100 possessions)
• luck is the difference between actual winning percentage and expected winning percentage, based upon net efficiency

So how do we determine expected winning percentage? Well, there is a fairly strong correlation between net efficiency and actual winning percentage. Here are the actual winning percentage versus net efficiency numbers for the past five seasons (2005-2009) for the Big East:




I've colored the markers for this season (through 2/26) in blue, all previous seasons in red. The horizontal and vertical black lines represent the break-even points, either as net efficiency is zero or winning percentage is 50%.

The diagonal black line is the best linear fit to the data and the dashed blue lines are the 99% prediction bands - there is only a 1 in 100 chance that a team will lie outside of the prediction bands.

Teams that lay above and to the left of the fitted line are "lucky" in the sense that they have a better winning percentage than a team would expect to have based on net efficiency. And teams below and to the right of the fitted lines are unlucky.

The infamous 2006 Notre Dame team has been the unluckiest team in the five years I've run - they ended the year 6-11 (including a loss in the BET), but would have been expected to end up 10-7. Actually their expected won/loss record in the Big East was 9.7-7.3; I'm not a big fan of decimal wins and losses, but I'll use the extra significant digit to distinguish between teams (below).

That same year, the Syracuse Orangemen ran off four improbable wins in a row in the BET, which makes them the luckiest team in the last 5 years.

We can also glean a bit more from the graph:

• While the conference is normally thought of as the Big 5 (UConn, Pitt, Louiville, Marquette and Villanova), the net efficiency rankings indicate that 'Nova may be just a bit behind those other four, and at about the same level as West Virginia.
• It may not be apparant by winning percentage, but the efficiency stats indicate that there is a big jump between the top-12 teams in the conference and the bottom four (Rutgers, USF, St. John's, DePaul).
• While there are some lucky and unlucky teams, this year there are not gross outliers either way - at least not yet.
• It's always important to keep in mind that this analysis is for the conference season as a whole, and doesn't account well for injuries (D. James, J. Dyson, etc.).
  

Note: The actual method used by Ken Pomeroy (and probably John Gasaway) to calculate luck is based on Pythagorean winning percentage. The math is a bit more complicated, but the results are essentially the same as the graph presented above - the only big differences are for extremely good or bad teams, such as DePaul. For the table below, I'll use this more rigorous method.


Here's a table summarizing this year's teams, ranked by luck.

.                 Wins  Losses   Exp Win  Exp Loss   Luck
Cincinnati          8     7        5.6      9.4       2.4
Providence          9     7        7.1      8.9       1.9
St. John's          4    11        2.3     12.7       1.7
Villanova          11     4       10.4      4.6       0.6
South Florida       3    12        2.4     12.6       0.6
Louisville         13     2       12.7      2.3       0.3
Marquette          12     3       11.7      3.3       0.3
Seton Hall          6     9        5.7      9.3       0.3
Connecticut        14     2       14.1      1.9      -0.1
Pittsburgh         12     3       12.6      2.4      -0.6
Notre Dame          7     8        7.7      7.3      -0.7
DePaul              0    15        0.9     14.1      -0.9
Syracuse            8     7        9.1      5.9      -1.1
Georgetown          5    10        6.4      8.6      -1.4
Rutgers             1    14        2.5     12.5      -1.5
West Virginia       8     7       10.4      4.6      -2.4

As Gasaway says, Cincinnati, Providence and St. John's are the luckiest teams so far this season in the Big East, and Georgetown and West Virginia are amongst the unluckiest (for some reason, he left out Rutgers).

Gasaway on the Hoyas

John Gasaway of Basketball Prospectus had a short post on Georgetown Friday, "Georgetown's Internal Bleeding."

Some highlights, annotated:

The Hoyas have now played a third of their conference slate. There’s plenty of season left to be played, of course, but in order to make something of that season GU will have to improve dramatically on defense.

I'd love to have a snarky rebuttal, but this is true. I've updated the season's Performance Charts through the West Virginia game, and the Hoyas have been consistently underwhelming on defense since the UConn game. That's six games in a row where the Hoyas have allowed more points than expected, based on venue and KenPom's season stats to date.

A month ago I remarked somewhat raffishly that Georgetown appeared to be inventing a new category: “outstanding defense without rebounds.” Well guess what. Turns out you need rebounds after all. In a conference with Seton Hall, the Hoyas can at least take solace in the fact that they will always be spared the indignity of being “last in defensive rebounding,” but the truth is their defensive rebounding is terrible.

For the season as a whole, Georgetown's defensive rebounding has been miserable at 61.7%, ranked 318/344 - worst amongst the BCS conference schools. Within conference play, Georgetown (56.9%) is just a fuzz better than Seton Hall (56.4%) which is worst. The comparison to SHU is particularly appropriate since the teams have 5 common opponents so far this season (UConn, NDU, Cuse, Provy and WVU). The only difference is that the Pirates have played Villanova (OR = 36.3%; 69/344) while G'town played Pitt (43.6%; 2/344).

There’s a sense at large that Thompson will right this talented ship and, who knows, that sense may be proven correct. But as of this moment Georgetown has been merely the tenth-best team in Big East play on a per-possession basis. Unthinkable on December 29, but true.

Here's where Gasaway is being a bit disingenuous - Georgetown may be tenth in efficiency margin [= points scored per 100 possessions - points allowed per 100 possessions], but they have played the hardest conference schedule to date, based on KenPom's ratings.

To look at this further, I've compiled the strength of schedule for all Big East teams for conference games played so far (through Saturday, Jan 24th), and plotted eff. margin against strength of schedule. Here, I'm using KenPom's Pythagorean rating (0-1), rather than his ranking (1 to 344) for the strength of schedule calc.

The blue line is a linear fit to the data. There is clearly a trend here - the harder a team's conference schedule so far, the worse their efficiency margin. If the fitted line is to be trusted - and I don't have any way of knowing if it should, beyond the fact that it is statistically significant at 95% - it implies that there are 6 top-tier teams in the Big East, once you account for the quality of opposition. Georgetown is one of them, along with Louisville, Pitt, Marquette, UConn and West Virginia.

One thing I'm not adjusting for here is home vs. road in evaluating strength of schedule, mostly because I'm not clear what the best weighting factor would be. It should be noted that Georgetown has played 6 conference games so far this season, with only 2 on the road.

A few other comments from looking at this chart:

• Notre Dame is surprisingly (to me) below the line, implying that they are not a top-half team in the Big East.

• Providence and South Florida are quietly hanging around in the second tier of Big East teams. The Friars aren't a big shock, as they were a darling pick to make the leap this year due to their experience; the Bulls may be reaping the rewards of Gus Gilchrist's eligibility.

• The four worst teams are Rutgers, St. John's, DePaul and Cincinnati, in descending order. A few weeks ago, I had figured that Cincinnati was probably the 10th best team in the Big East, heading into conference play.

• Seton Hall, today's opponent for the Hoyas, has played the 2nd hardest conference schedule to date - i.e. they may not be as bad as their current conference record indicates.