Showing posts with label four factors. Show all posts
Showing posts with label four factors. Show all posts

Friday, January 14, 2011

Friday thoughts

Not really sure where I'm going with today's post, but just sitting here riffing on some stats.
--------------------------------------------------------------

One of the great conceits of being a sports fan is that we all think we know more than ______, and that if the coaching staff only listened to us, our team would be that much better.  Some of us even start blogs to make sure we can get the word out.

But, after a few years of blogging, it starts to become clear that maybe we don't know as much as we'd like to think.

Here's an example:  I can make a simple model using the four factors to estimate Georgetown's offensive efficiency for each game played so far this year, and from this I can give an estimate of how important each factor is to Georgetown's offense.
Factor      Weight
eFG%         0.64
OReb%        0.18
TO%          0.13
FTM/FGA      0.01
Total        0.96
That's to say, about 64% of the variability game to game in Georgetown's offensive efficiency can be explained by how well they shoot (eFG%). Once we account for that, about 18% of the remaining variability is due to how well they offensive rebound, etc.

I can even make a nifty graph to show how well the model works:


 I can also repeat the exercise for the defense (no nifty graph this time), for a slightly surprising result [this table has been revised]:
Factor      Weight
eFG%         0.36
TO%          0.46
DReb%        0.11
FTA/FGA      0.02
Total        0.95

But here's the thing - what can we do with this information?

I can demonstrate that Georgetown's offense is most dependent upon how well the team shoots from the floor, but I can't tell you how the Hoyas can improve their shooting accuracy from this exercise.

More importantly, just because their defense this year is most dependent upon how well they turn over the opponent doesn't mean the team should be running a full-court press all the time.  The law of unintended consequences reminds us that the result could just as easily wind up that teams would end up shooting more layups (and therefore improve their shooting accuracy) as a result, and Georgetown's defense wouldn't improve at all.

But it sure seems tempting.

Tuesday, December 2, 2008

The Offense: Five Games In

With the graduation of Roy Hibbert and Jon Wallace, and the assumed ascendancy of several players more suited to a different style of play, the big question concerning the offense this off-season was whether the style would change. (A better question was whether the offense would be as effective, but style before substance, I suppose.)

No one expected the team to abandon the Princeton principles, but with a faster, more gifted team that lacked an experienced low post threat, most fans expected an offense that utilized the fast break and the dribble drive more than in previous years.

Let's see what has happened so far, using the context of the "four factors" plus pace.


Overall Offensive Efficiency (adjusted for competition)

Last Year: 117.2 (18th)
This Year: 119.9 (13th)

Surprised? What's the cause? Let's look at each of the factors in turn.


Pace


Last year: 62.2 possessions per 40 minutes (316th)
This year: 68.1 possessions per 40 minutes (170th)

Well, pace isn't really a factor, but it is important to most people anyway.

Is this increase in pace real? Given the slower pace of the Maryland game, you could make an argument that Georgetown's pace is inflated by playing Tennessee and some cupcakes.

Maybe it is, but it doesn't seem likely to fall back to the crawl of prior years. Georgetown didn't break 64 possessions until Game 7 last year; this year it has four games at 64 or above and the low is 63. Last year the team routinely fell below 60 possessions -- ten times, in fact -- plus seven more games under this year's low of 63.

From watching the games, is there any doubt the team is looking to fast break more? Wright, Freeman and Summers have all gotten a significant number of their points on the break this year.


Effective FG%


Last year: 56.8 (4th)
This year: 54.3 (41st)

With Roy Hibbert graduating, no one should be surprised that our effective FG% is down.

Here's what is surprising: it's because of the three point shooting, not the two point shooting.

Our two point FG% has actually increased so far this year, from 56.3% to 62.9%. Now the latter is inflated because of the relative level of competition, but in 2007-08 and now in 2008-09, we rank 2nd in 2-point FG%. Considering our schedule has at least been average, we haven't yet seen a drop-off in 2-point FG%. (Note that we shot 61% from two versus Tennessee.)

My guess? Our fast break game and Monroe's better than advertised post game (67%) have helped to offset the loss of Hibbert.

What is somewhat befuddling is the decline in three point shooting. Last year, the Hoyas shot 38.4%. This year, the Hoyas have dropped to 28%, 295th in the nation.

I knew Jon Wallace would be missed, but that's ridiculous.

Is it all Jon Wallace? Well, no, not really. There are a number of issues. Check out the composition of our threes, year over year:

Position2008 3PT%2008 % of Team's 3's2009 3PT%2009 % of Team's 3's
PG (Wallace/Wright)47%25%29%17%
Sapp41%22%40%15%
Freeman40%15%25%16%
Summers34%22%37%19%
Backups & Other28%16%18%33%

Really boils it down, doesn't it? There are three main issues:
  1. Freeman and Wright are not shooting at last year's proficiency. In Wright's case, he's neither duplicating Wallace or his own efforts.
  2. The backup guards are taking a much larger percentage of the threes than the starters. Rivers may have been awful from three, but he knew enough not to shoot much. The same can't be said of Clark and Wattad so far, who are shooting at almost quadruple Rivers' share of the threes.
  3. The backup big men aren't making shots. Ewing was hardly a marksman, but with Vaughn and Sims at an "0-fer" and the same % of shots, it's a large gap.
A couple of these factors will likely right themselves.

Freeman is too good of a shooter for this to continue.

Vaughn and Sims may not shoot as well as Ewing, but they will either shoot better than 0%, or stop shooting.

The other two may or may not.

Clark and Wattad likely will cut back on their threes a bit, whether just through reduced playing time or simply better shot selection in close games. But it seems unlikely they will shoot less than Rivers, who truly understood it was a bad shot. Will Omar and Jason shoot well enough to justify pulling some shots from Sapp and Freeman? Perhaps.

Perhaps more important is whether Chris Wright will recover. Watching him shoot, I like his stroke. He shot well last year, and as long as his shot selection is strong, there doesn't seem to be an obvious reason for him not to shoot as well this year.

However, a better historical predictor of this year's 3 point percentage for a player is last year's free throw percentage. In other words, Chris' awful 52% from the line last year may be more relevant than his 48% from 3-point land in terms of predicting this year. Players who don't shoot well from the line don't tend to be good shooters.

Wright is shooting 63% from the line this year, which is an improvement. But his free throw struggles might signal that his 3-point % is not going to recover. And either way, he was never likely to hit 47% again.

The 3-point shooting will recover some. It's not going to be as good as last year because no one will replace Jon Wallace. But overall, our shooting % looks to be strong because of Monroe's advanced play and fast break points.


Turnover %

Last year: 21.2% (192nd)
This year: 21.9% (192nd)

That's not a typo. It's almost eerie in its consistency. We still have a mediocre offense in terms of turning it over.

I'm not sure if Georgetown will ever be a great ballhandling team. We pass and handle more than most teams. All of our players handle. I suspect the lowest turnover teams have a great point guard that handles the ball most of the time. We simply have the ball for more seconds than most teams.

However, there's hope for the turnover rate:



Players% of Poss ending in TOAverage % of Poss Used
Four Good Starters16%22%
Sapp34%18%
Backups32%17%

Despite the Tennessee game, Freeman, Wright and Monroe have committed very few turnovers. Summers has been average. But Sapp has been atrocious so far. He's certainly not that bad; but the question is are the four "good" starters that good? I expect that to average out.

Where the real possibility for improvement comes merely in the distribution of possessions. As the backups get less playing time the overall number of turnovers will decrease. Perhaps some of them will get more comfortable as time goes on as well. Clark, Wattad, and Vaughn (who was not good at holding onto the ball at FSU as well) are the main culprits. Even a minor improvement could reap large dividends.


Offensive Rebounding %

Last year: 33.8% (132nd)
This year: 29.4% (260th)

I've discussed this in detail already.

However, one thing to note is that last year, our offensive rebounding was essentially Roy Hibbert and nothing else. No one else broke 10%, and only Macklin was even close.

This year, Greg Monroe is at 7%.

That's not the whole issue, but it is about 3% of a 4% difference.


Free Throw Rate (FTM/FGA)

Last Year: 22.9 (259th)
This Year: 40.7 (7th)

So, if the rebounding and three-point shooting are worse, and turnovers are the same, why is the offense as good or better so far?

The answer is that the Hoyas are getting to the line.

It's been long considered a huge weakness of the Princeton offense that it doesn't get foul shots. The best rank a JTIII-led Hoya team has had was 136th, when Jeff and Roy were taking the shots down low.

However, this year, that has all changed. And it isn't just end of game fouling. While Jacksonville and Wichita State did have very high free throw rates, the Maryland and Drexel were over 40 and the loss at Tennessee was over 50. And in the Wichita State game, the Hoyas registered a very high FT Rate in the first half.

Editor's note:
For the stats I post after each game, I use FT Rate = (FTA/FGA). KenPom also uses this formula for his Game Plan pages, but he reports FT Rate = (FTM/FGA) for the offense only on his Scouting Report pages. To convert between the two, just multiply by the team FT%; I post both types on the team splits stats page.

Visually, there's no doubt that the team is driving to the hoop more. That, combined with more fast break points, has generated more fouls. And Monroe has drawn a ton of fouls down low. Almost every player has an extremely high FT Rate -- the lowest starter is Chris Wright at a 46.

Can the team continue to effectively drive and draw fouls in the Big East? If it does, this offense has a chance to be better than last year's.

One big wrench into this could be the ratio of man defenses to zone. I'm going to try to track which teams play primarily what defense to see what the actual difference is in effectiveness. But we certainly seemed to have been helped by playing Tennessee and Maryland, two teams committed to man to man.

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

Addendum by CO_Hoya:

Individual and team stats pages are now and running for this season. I'm hoping to make some changes/improvements to these pages, but this will be mostly behind-the-scenes stuff.

More splits will be become available as the season progresses.

Friday, June 20, 2008

Team Performance Update

I've finally started to get around to posting final stats for the 2007-8 season. I've run into a bit of a snag, in that I don't have 1st half / 2nd half split box scores for the NCAA games (Tom Gower, help!!), so I've only posted the final performance charts.

A few points:
  • Before I get into the details of Georgetown's perfomance this season, I thought I'd direct my reader's attention to a post on VBTN (yes, a month ago), reviewing Big East teams' offensive and defensive ratings. greyCat may have the axes mislabelled, but he does a nice job explaining the relative strengths of the 16 BE teams. In a nutshell, Louisville and G'town played defense at a much higher level than the other 14 teams, while G'town's offense was only slightly better than league average. I believe that greyCat is using the entire season for his analysis, and I'd like to see his analysis reworked for only conference games, as the non-conf. SOS varied widely.
  • The winner(?) for worst game of the season goes to the BET final against Pitt, where the Hoyas were more than 13 points worse than expected. This number is likely skewed by the fact that Pitt was playing much better at that point due to the return of Levance Fields - indeed, this is a great example of a shortcoming of using Pomeroy's stats to predict performance, as his system isn't "aware" of injuries.

Finally, I thought I'd break down the team's performance into their four factors, aggregated by performance. I've separated the season into 3 chunks, good performances (> ~3 pts. above expected), poor performances (< ~6 pts. below expected) and middling (those games in between). This is somewhat similar to KenPom's correlations, found at the bottom of his game plan page, reprinted here for reference.
                          Correlations
to OE to DE
Pace: +0.13 -0.08

eFG%: +0.71* -0.24
OR%: +0.32 -0.07
TO%: -0.44* -0.07
FTR: +0.09 -0.16

Opp eFG%: -0.03 +0.65*
Opp OR%: -0.26 +0.26
Opp TO%: +0.14 -0.55*
Opp FTR: -0.18 +0.33

Bold values are significant with a 95% confidence
Bold* values are significant with a 99% confidence


First, general info:
          Good  Middling  Poor
Record 11-0 10-2 7-4
Pom Rank 100.0 92.6 91.0
Pace 64.8 62.9 61.9
Even when Georgetown played poorly they usually won; since performance accounts for opponent quality, a close win against a big underdog would be considered a poor performance. The Hoyas best games came against the weaker teams, but this is a small effect. Faster pace seems to track performance (the "good" pace appears significantly faster).


Next, offense:
          Good  Middling  Poor
OEff 121.1 114.4 104.1
eFG% 60.9 58.2 53.9
OR% 33.3 32.9 30.8
TO% 19.4 20.8 22.8
FT Rate 25.2 22.3 24.1
It's no surprise that Offensive Efficiency tracks performance; after all, it represents half of performance. As Pomeroy suggests, eFG% shows a good correlation with OEff, the other factors less so. What can be seen here is that the Hoyas shot well even in average games, but poor shooting meant poor performance. Georgetown shot 34% (76/223) on 3FG in their poor performances, versus 40% (193/478) in other games; the Hoyas were 54th of 341 teams in dependence on 3FGs, and this shows here. I was expecting a bigger difference in TO% (turnovers committed), but the effect seems small here.


Finally, defense:
          Good  Middling  Poor
DEff 85.9 90.1 98.9
eFG% 37.0 42.0 42.8
OR% 34.1 32.1 31.0
TO% 21.8 22.0 15.5
FT Rate 39.1 35.6 39.0
The difference between middling and poor defensive games wasn't opponents' shooting or defensive rebounding, but turnovers forced. I didn't expect that, although the Davidson game would be a clear example of this. When Georgetown played very well, their opponents had a miserable time putting the ball in the basket.