Saturday, March 3, 2007

Analysis: Free Vernon Macklin?

With several days until the BET starts, we can look back on the season's statistics, and see what kind of trouble we can get into.

First off, I just looked at the offensive statistics for each player. Since I don't have the formula for offensive rating (per Dean Oliver) handy, I just made up a statistic on the fly and ran it for the cumulative individual stats for each player. If someone else already is using this stat, I'd be happy to credit them.

First, I looked at possessions used, although not as described by Ken Pomeroy. I simply add up all field goal attempts, free throw attempts x 0.475 (this is from KenPom) and turnovers, then subtract back any offensive rebounds. This gives a measure of possessions used, with credit for saving a possession on the offensive glass.

Poss. Used = FGA + 0.475*FTA + TO - OR

An obvious problem with this formula is that can end up with zero or a negative number of possessions used if a player gets some ORs but doesn't do much else; Tay Spann ends up with a -1 against Fairfield and Patrick Ewing Jr. has a -1 against Notre Dame (in 12 minutes!). This can lead to some curious math if you plan to use poss. used as a divisor on a game-by-game basis.
That's why I'm keeping it simple, and just looking at the end-of regular season numbers.

Another problem, if you bother to work out the math, is that when you add up the individual players' poss. used and compare to the box score possessions, as per KenPom, you inevitably end up 2 or 3 higher than Pomeroy. This is due to not accounting for team (deadball) offensive rebounds, which aren't assigned to any individual player.

There's one fairly significant event during the season that had an effect on individual stats, namely the departure of Marc Egerson after the 1st semester. Since this is a natural break in the schedule (end of out-of-conference, start of Big East schedule), I've removed Mr. Egerson from the dataset, and only looked at BE-season stats (16 games).

Okay, time for some charts (you know you want them).

First, I just take total possessions used for the season, then divide by minutes played. This gives possessions used per minute of playing time. The players are sorted by total minutes played this season (as always, click any image to enlarge).



Other then the two seniors at the end of the bench, the general trend is about as you'd expect. Jeff Green is the most dominant player on offense, with the other 4 starts about equal (although there are a few posters on Hoyatalk who would probably like to see Roy Hibbert's possessions used / minute number increase). The bench players are more deferential, especially Macklin and Spann. Sead Dizdarevic seems to play at a whole 'nother level when he's in the game.

Now this is all well and good, but if I were to compare Georgetown to players from another team, I'd have to take pace (total possessions available) into account. So, I reworked the stat into poss. used % (and this may be close to Ken Pomeroy's Poss. %, but I'm not clear), where I just divide a player's possessions used by total possessions used by the team. Since each player used some percentage of all available possessions, this also allows me to make my first pie chart. Woo-hoo!



Here things look a bit more rational, as Sead is no longer quite the offensive force. Most interesting to me is that Roy Hibbert is the lowest of the 5 starters, and PE Jr. is as low as he is. Ewing in particular may show a trend over the course of the BE season, but that is a post for another day.

Edit
I've thought about this a bit, and I don't think the pie chart is useful if I'm trying to look at tempo-indepedent stats. All the pie chart is showing is what percentage of total team possessions a player used, while what I'm after is "What percentage of available possessions did a player use?" To get this, I take poss. used / minute, divide by possessions / game (team average = 58.6) then multiply by 40 minutes / game.

Poss. Used./Minute * Game / Total Poss. * Minute/Game

What you end up with is the ratio of possessions used per player over possessions played per player. This gives a truly pace independent stat, using the assumption that the game pace is the same for all players. Here's the graph:



You'll notice that it looks the same as the first chart (Poss. used /min.). The only difference would be that you could compare these percentages to another teams' players directly. I suspect this is quite similar to whatever Pomeroy uses for Poss %.

End edit


Since we've addressed how aggressively each player is using his time on the floor, the next logical step would be to look at how successful each player is with his playing time.

A simple analysis is just points per minute, which I'll show first. For this and the remaining charts, I've decided to drop Messrs. Spann, Dizdarevic and Izzo to look at those players likely to see minutes while the game's outcome is not yet decided (with my apologies).



This is the sort of thing you're likely to find when you look at the Big East media guide (well, probably not as a nifty graph, but at least the underlying numbers). One thing that jumped out at me right away is how PE Jr. suddenly looks like the 6th starter that he's now become; also, Vernon Macklin appears to be playing quite well on offense.

However, I don't want to get too carried away looking at this statistic, since we've seen that, e.g., Roy Hibbert doesn't use nearly as many possessions as Jeff Green. So to find how productive each player is, relative to how aggressively he's using his playing time, I'll look at points per possession used.

Oh, and another aside is that Jessie Sapp and Jeremiah Rivers may be getting a short shrift here, as their job as the nominal point guards may discourage scoring as their primary goal. So, I'll also present (points + assists) per possession used.



I like this last graph for a few reasons:
  • My attempt to aid Sapp and Rivers helped a bit, but it really shows a key element of the Princeton/Georgetown offense, namely that most players show a significant (>0.3) increase due to assists per possession used - the team does a good job sharing the ball.
  • Hibbert is easily the most efficient scorer among the six core players, although Ewing is close behind if you're willing to credit assists.
  • Rivers strengths as a ball-handler and defender are not much of an asset in an analysis such as this; I'm only looking at offense, not the whole player.
  • Conversely, Vernon Macklin may really be as good as his recruiting rankings indicate. I mean, he's nearly literally off the charts! Again, I'm not measuring his defense (and don't really know how), but this young man certainly seems effective when given the chance.

Okay, that's it, too tired to check for typos until tomorrow.

No comments:

Post a Comment