Sunday, February 28, 2010

Good Wins, Bad Losses

The current motto of Hoya Prospectus is "thinking rationally about Georgetown Hoyas basketball so you don't have to."  What we set out to do here is discuss basketball in an objective and statistically defensible manner.

So when I saw the shrill reaction to yesterday's loss, I got to wondering about where the loss ranked for Georgetown and within the Big East generally.

There are two ways to think about the results of basketball games:  wins and losses, and the final score.  Stat-heads tend to use the latter - probably to a fault - since it has proven the best means of predicting future results.  But at the end of the day, wins and losses are what get you paid.


Let's take a crack at each.


Margin of Victory

If you're just wondering where a Georgetown game ranks in terms of the difference between final score and expected final score, you can always find that here at the Performance Charts page (link).  The math behind the performance charts (original discussion here and here) is not perfect, since it applies teams' current stats from KenPom.com retroactively, but I think it does a good job providing context to a game.

I'll reproduce the Georgetown's current overall performance chart below (note that I also break out offensive and defensive performance on that page):


By this metric, yesterday's loss was Georgetown's worse of the year.

As of this morning, if Georgetown and Notre Dame played again at the Verizon Center, we'd expect Georgetown to win by 7 points; instead they lost by 14.  Of course, our system is blind to certain vagaries, such as Austin Freeman trying to play with a viral infection or Notre Dame enjoying some new-found ability without their purported leader (more on that in my next post).

Yesterday's game noses out the South Florida loss as the worst performance of the year by the Hoyas.

This got me to wondering how this game fits into the Big East picture as a whole.  Was this the worst performance by a Big East team all year?  In conference?

Here are the worst perfomances for Big East teams so far this year, using today's stats.  Note that I'm being more tempo-free rigorous by sorting by efficiency difference [actual eff. margin - expected eff. margin] rather than scoring difference [actual score - expected score].  Also, I'm using efficiency as points per 100 possessions to avoid those pesky decimals.

Team           Opponent         Site    Result  Score   Poss.  Eff. Diff
Louisville     Charlotte        Home       L    87  65   67      -50
Georgetown     Notre Dame       Home       L    78  64   58      -37
Louisville     Western Carolina Home       L    91  83   83      -36
Pittsburgh     Indiana          Neutral    L    74  64   74      -34
South Florida  St. John's       Home       L    74  58   59      -32
Louisville     St. John's       Away       L    74  55   64      -32
Connecticut    Cincinnati       Home       L    60  48   63      -30
Notre Dame     Loyola Marymount Home       L    87  85   71      -29
Georgetown     South Florida    Home       L    72  64   65      -29
DePaul         Mississippi St.  Neutral    L    76  45   68      -28
St. John's     Providence       Home       L    74  59   73      -28
Pittsburgh     Notre Dame       Away       L    68  53   51      -28
Notre Dame     Idaho St.        Home       W    80  70   65      -28
Syracuse       Pittsburgh       Home       L    82  72   75      -28

I've arbitrarily picked the 14 worst games in order to make the point that a team can win a game but still play terribly doing so, if they're playing a massively inferior opponent.

A few other thoughts:
  • This is a glass half-empty, glass half-full sort of analysis.  You can also say that Notre Dame's win over Georgetown was the Irish's best performance of the year.
  • Four teams show up with multiple slots on the list (Louisville, Georgetown, Pittsburgh and Notre Dame).  Both of Louisville's and Notre Dame's games and one of Pitt's games were in the early going.  Both of the Hoyas' bad performances were in February.
  • Generally, it's harder to have a bad performance on the road, since teams are expected to play worse on the road. Louisville and Pitt are the exceptions to the rule here.
  • Speaking of the Cardinals, one game is unlike any other on that list.  Hopefully, that 22-point blowout at home has some extenuating circumstances that I'm not privy to.
  • It's hard to believe that losing by 10 points to Pitt would be considered a terrible performance, but that should provide some context as to just how well Syracuse is playing this year.

Allow me to produce a second list, looking only at conference games:
Team           Opponent         Site    Result  Score   Poss.  Eff. Diff
Georgetown     Notre Dame       Home       L    78  64   58      -37
South Florida  St. John's       Home       L    74  58   59      -32
Louisville     St. John's       Away       L    74  55   64      -32
Connecticut    Cincinnati       Home       L    60  48   63      -30
Georgetown     South Florida    Home       L    72  64   65      -29
St. John's     Providence       Home       L    74  59   73      -28
Pittsburgh     Notre Dame       Away       L    68  53   51      -28
Syracuse       Pittsburgh       Home       L    82  72   75      -28
Providence     Louisville       Home       L    92  70   71      -27
Syracuse       Louisville       Home       L    66  60   58      -26
Providence     Marquette        Away       L    93  63   67      -26
Seton Hall     Pittsburgh       Away       L    83  58   66      -25
Villanova      Connecticut      Home       L    84  75   70      -23
Connecticut    Providence       Away       L    81  66   78      -23
Rutgers        West Virginia    Away       L    86  52   64      -22

Georgetown's loss yesterday was the worst performance in conference so far this season, by a substantial margin.  Interestingly, the loss at Rutgers doesn't look as bad now as it did then (it's way down the list here:  -19 eff. difference), as the Scarlet Knights are 5-3 in their last 8 conference games.

Syracuse has lost only two games all year, and they both rank as some of the lousiest performances of the year in the Big East. Translation: a team will have to play brilliantly to beat them.

There are some schizophrenic teams in the league:
  • Providence has two Top-14 performances (at St. John's and vs. UConn) and two of the Bottom-10 (vs. Louisville and at Marquette).
  • St. John's, Pitt and Louisville have two great games and one terrible one on the list; UConn has one great performance and two terrible ones.

Wins and Losses

I suspect that my reader is blinking his eyes a bit by now, wondering why I'm so obsessed with scoring margins when the bragging rights and suffering are determined only by whether the Hoyas can win the game.

So how do we rank wins and losses?  Well, I can calculate the win probabilities (as seen on KenPom's scouting report pages) for each game played, and look for the biggest outright upsets, here as losses when expected to win.  I'm limiting this to just conference games:
Losing Team    Opponent         Site    Score      Odds
Georgetown     South Florida    Home   72   64      86%
Syracuse       Louisville       Home   66   60      85%
Marquette      DePaul           Road   51   50      85%
Syracuse       Pittsburgh       Home   82   72      83%
Georgetown     Rutgers          Road   71   68      82%
Connecticut    Cincinnati       Home   60   48      78%
Georgetown     Notre Dame       Home   78   64      78%
Villanova      Connecticut      Home   84   75      76%
Notre Dame     St. John's       Home   69   68      76%
West Virginia  Villanova        Home   82   75      72%
Notre Dame     Rutgers          Road   74   73      72%
St. John's     Providence       Home   74   59      70%
Marquette      Pittsburgh       Home   58   51      68%

This should all be straight-forward except the odds column, which is the odds that the losing team would have of winning the game if it were to be replayed today (same venue).  The list is limited to games with the odds greater than 68%, i.e. games where the favorite is more than one standard deviation above even odds.

About half of the games show up on both lists: MU at DePaul is the worst outright loss that didn't make the worst performance list, and South Florida hosting St. John's is the worst performance that didn't make the worst outright loss list (the Bulls would be 64% favorites in that game today).

It should be noted that good teams are far more likely to be here than bad teams - for example, DePaul would only be a favorite to win one game this season (at home against Rutgers [61%], a game they lost).
  • Georgetown once again has the worst loss, now as an outright loss rather than as performance.  The Hoyas are also the only team with three bad outright losses on the list.  This is making me more and more depressed as I write this up.
  • Both of Syracuse's losses were games they should have won easily.
  • The worst road loss so far is Marquette's 1-point loss to DePaul, a game that didn't show up on the performance lists (since Marquette only lost by one point, it wasn't as bad a performance as their home loss to NC State [77-73].
  • Showing up on the "Opponent" column multiple times is a mixed blessing - Rutgers will be a big underdog even in some of their home games, so many of their wins are a big deal, while Pitt would probably prefer to have not been such an underdog at Marquette in the first place.
  • Four of the upsets were 1-possession games:  Marquette/DePaul, Georgetown/Rutgers, Notre Dame/St. John's and Notre Dame/Rutgers. It's tempting to invert the results of each of those games, making Marquette 10-5, Georgetown 10-6 and Notre Dame 10-6 heading into Sunday's games.

So there it is.  Georgetown holds the distinction of having the worst performance and the worst loss in conference play so far this year, and they're not even the same game.

      No comments:

      Post a Comment