Tuesday, December 14, 2004
MATH EXAM
(This post has a lot of numbers and is kind of long. If you don't like numbers, or my prose, you should watch out. It also owes a lot to 82games.com, the Journal of Basketball Studies, and other stathead/basketball nerd sites. You can also look at an interactive version of these stats here)
Not a lot in the way of basketball action this week-- everyone is busy studying for their finals. Until play resumes, I've got some time to plug away at the stat sheets. The sample size is too small and the difference in schedule difficulty is too large at this point to draw any conclusions from 2004's action. However, we can take some meaningful looks at what happened last year during league play, when everyone played the same opponents the same number of times. Blogger is formatting this weird-- so scroll down please.
The above is a lot of stats about what happened during league play offensively for the 8 PL teams. Don't fall asleep yet. You should be familiar with the top line, since those are all typical stats you see in a basketball box score, and I talked about Possessions in my last post.
Floor % is how often a team scores at least one point during their offensive possessions, whether it be by jump shot, layup, three pointer, or resulting foul shot. As you can see, Army struggled mightily to score any points last season, and it shows up here. They scored at least one point on about 1/3 of their possessions, meaning they probably hit a field goal every six possessions. Not good. Holy Cross, on the other hand, did better than every other team in the league, scoring on 55% of their possessions. Notice that Lafayette's floor percentage is the lowest among the non-academy teams. More on that later.
Ignore points per possession for now.
Possessions per game is a measure of tempo on the court-- a team that prefers to run and gun will have more possessions (as will their opponent), while a team that prefers the half-court offense will typically have fewer. An example might be helpful: UNLV in the early 90s had a lot of possessions, as did their opponents, because they were an athletic team that liked to push the ball upcourt. A team that runs the Princeton offense, on the other hand, prefers to use the clock to their advantage and work a half court game, resulting in fewer possessions.
There are not a lot of outliers on possessions per game, meaning pace was about the same for most PL games. There are two exceptions, with Lafayette having about 5 more possessions than the league average in their games, while Army was typically 2 or 3 short. This indicates what we all probably already knew-- Lafayette played a full court game, trying to outrun their opponents, while Army tried to limit their (and their opponents) possessions. This makes sense for both of them, because favorites will typically beat underdogs if they have more opportunities to do so. That is, the odds are better for the underdog hitting their shots better than the favorite in the first 10 possessions than they are for the underdog hitting their shots better in the first 100.
Last season, going into league play, Lafayette looked great. They were the favorite for sure, winning 8 of their first 9 games, and going undefeated (with 3 OT victories no less) in the first half of league play. It was then good strategy for them to try to increase the number of possessions in a game, since their guys were the favorites. On the other hand, Army looked lousy the whole season, and it was better for them to try to be lucky rather than good, and slow the game down.
You might be asking yourself "Matt, so what if they had 3 extra possessions? That's nothing. A statistical anomaly. Who gave you a keyboard?" If so, you should probably stop asking yourself questions that are directed to me. However, faithful skeptic, I will explain why a difference in two or three possessions is significant.
Teams can really only control tempo during about 50% of the game (this part is not scientific, just conjecture). They are probably responsible for 40% of the tempo of the game when they have the ball, and 10% when they don't. Sure, teams can press, implement a zone defense, or anything else, but if their opponent wants to start throwing up jumpers, they're going to start throwing them up. That means that these stats are dulled down by the fact that the other team (i.e. whoever LC and Army were playing) also plays a role in the number of possessions in a game. So the numbers do mean something.
Offensive rating is a measure of how good a team is on the offensive end. It's just points per possession*100, giving us a number that is easier to read than a bunch of decimal points. You'll see that LC was by far the best at scoring points. Everyone else is relatively equal, with Lehigh above most of the pack and the academies expectedly sucking.
I threw in fouls here with the offensive stats, but that's a little confusing. The fouls number here is those committed by the team listed, not however many fouls they drew. The only notable thing here is that there isn't anything notable-- these numbers show that teams typically were called for fouls about the same amount. I'm not sure that makes sense, however, given that some teams are inherently more physical than others, and teams like Navy and Army typically need to foul more often because teams are getting better looks at the basket than them. It may reinforce the idea that referees often like to call a game "both ways", despite the fact that one team may be more at fault than another.
Above is measurements of the school's defense. Again, you should be familiar with most of the top part of stats. They actually have a few interesting things to show us when you put them together for the whole league.
Look at offensive rebounds-- Lafayette got killed! (remember, this is how many offensive rebounds their opponents had) That's another indication that the Pards were more interested in running the ball than crashing the boards. Their opponents also scored a lot more points than they did against anyone else-- not to keep shoving this down your throat, but they were clearly looking for an uptempo game, and they got it. (Or is it that opponents wanted an uptempo game? Perhaps coaches thought the Pards could be beaten in quick game, but I doubt it.)
The second row has a lot of fun stuff. Bucknell is rated as the stingiest defense (see "offense rating", which really means defense rating, but you get the idea), and for good reason. They not only allowed the fewest points per contest, but they also allowed the fewest per possession. Note that their opponents had the second most possessions per game, but scored the least. Navy, Lafayette, and Colgate are rated the worst, with Holy Cross playing their typical excellent defense. So while Army played good defense, but were unable to overcome their terrible offensive woes.
I don't remember the Bucknell defense being that great, to tell the truth. But this is the kind of thing that shows maybe I'm an idiot that needs a computer to figure out what's going on. It will be interesting to see if they can keep it up this year.
Okay, now time to delve into some stats that are limited by the small sample size we have here (14 games are much tougher to analyze than the 82 in the NBA). They are also a little more complicated. But let's take a look, just for fun.
Points per possession divided by opponent points per possession is a way of balancing offensive and defensive numbers. If a team has a number over one, they scored more often than did their opponent. You'll see that while LC's defense was bad, their offense made up for it, and they come out with an overall rating of 1.05. But that was only good enough for fifth in the league. In the end, they couldn't keep up the tempo on the defensive end like they could on the offense, so they ran into trouble down the stretch (something I actually predicted and got write-- write down the date).
The other numbers are a little weird-- Bucknell's great defense makes them the best all-around team according to the stats. Holy Cross is also ranked above the eventual champions Lehigh. Why would this be?
The problem with these numbers is that they're combined to give a big picture of all 560 minutes of a season, rather than 14 segments of 40 minutes. So inconsistency doesn't mean anything in these numbers. If a team played very well one game and terrible another, it balances out to be mediocre here. If there were more games to analyze, then inconsistency would matter less.
However from these numbers we can imply that Holy Cross' inconsistency is what killed them. If they put up a consistent effort with the numbers above in every single game, then they would have been a contender in the league. Instead, they finished with a 7-7 record and a first round exit in the playoffs.
The pythagorean formula and luck numbers are just something I was fooling around with. If you care more about them, you can read about how it works here. The numbers really aren't that significant given they only go over a 14 game period.
Phew. That's a long post. Congratulations if you read through the whole thing. If anyone has any thoughts on this stuff, please let me know. I'm still trying to figure out how to apply this stuff to pre-conference play, using RPI and such, but who knows what that will result in. I'll have some time just before the new year to play around with the numbers more. Hopefully some of you will still be around to read it. @ 12:25 PM
(This post has a lot of numbers and is kind of long. If you don't like numbers, or my prose, you should watch out. It also owes a lot to 82games.com, the Journal of Basketball Studies, and other stathead/basketball nerd sites. You can also look at an interactive version of these stats here)
Not a lot in the way of basketball action this week-- everyone is busy studying for their finals. Until play resumes, I've got some time to plug away at the stat sheets. The sample size is too small and the difference in schedule difficulty is too large at this point to draw any conclusions from 2004's action. However, we can take some meaningful looks at what happened last year during league play, when everyone played the same opponents the same number of times. Blogger is formatting this weird-- so scroll down please.
The above is a lot of stats about what happened during league play offensively for the 8 PL teams. Don't fall asleep yet. You should be familiar with the top line, since those are all typical stats you see in a basketball box score, and I talked about Possessions in my last post.
Floor % is how often a team scores at least one point during their offensive possessions, whether it be by jump shot, layup, three pointer, or resulting foul shot. As you can see, Army struggled mightily to score any points last season, and it shows up here. They scored at least one point on about 1/3 of their possessions, meaning they probably hit a field goal every six possessions. Not good. Holy Cross, on the other hand, did better than every other team in the league, scoring on 55% of their possessions. Notice that Lafayette's floor percentage is the lowest among the non-academy teams. More on that later.
Ignore points per possession for now.
Possessions per game is a measure of tempo on the court-- a team that prefers to run and gun will have more possessions (as will their opponent), while a team that prefers the half-court offense will typically have fewer. An example might be helpful: UNLV in the early 90s had a lot of possessions, as did their opponents, because they were an athletic team that liked to push the ball upcourt. A team that runs the Princeton offense, on the other hand, prefers to use the clock to their advantage and work a half court game, resulting in fewer possessions.
There are not a lot of outliers on possessions per game, meaning pace was about the same for most PL games. There are two exceptions, with Lafayette having about 5 more possessions than the league average in their games, while Army was typically 2 or 3 short. This indicates what we all probably already knew-- Lafayette played a full court game, trying to outrun their opponents, while Army tried to limit their (and their opponents) possessions. This makes sense for both of them, because favorites will typically beat underdogs if they have more opportunities to do so. That is, the odds are better for the underdog hitting their shots better than the favorite in the first 10 possessions than they are for the underdog hitting their shots better in the first 100.
Last season, going into league play, Lafayette looked great. They were the favorite for sure, winning 8 of their first 9 games, and going undefeated (with 3 OT victories no less) in the first half of league play. It was then good strategy for them to try to increase the number of possessions in a game, since their guys were the favorites. On the other hand, Army looked lousy the whole season, and it was better for them to try to be lucky rather than good, and slow the game down.
You might be asking yourself "Matt, so what if they had 3 extra possessions? That's nothing. A statistical anomaly. Who gave you a keyboard?" If so, you should probably stop asking yourself questions that are directed to me. However, faithful skeptic, I will explain why a difference in two or three possessions is significant.
Teams can really only control tempo during about 50% of the game (this part is not scientific, just conjecture). They are probably responsible for 40% of the tempo of the game when they have the ball, and 10% when they don't. Sure, teams can press, implement a zone defense, or anything else, but if their opponent wants to start throwing up jumpers, they're going to start throwing them up. That means that these stats are dulled down by the fact that the other team (i.e. whoever LC and Army were playing) also plays a role in the number of possessions in a game. So the numbers do mean something.
Offensive rating is a measure of how good a team is on the offensive end. It's just points per possession*100, giving us a number that is easier to read than a bunch of decimal points. You'll see that LC was by far the best at scoring points. Everyone else is relatively equal, with Lehigh above most of the pack and the academies expectedly sucking.
I threw in fouls here with the offensive stats, but that's a little confusing. The fouls number here is those committed by the team listed, not however many fouls they drew. The only notable thing here is that there isn't anything notable-- these numbers show that teams typically were called for fouls about the same amount. I'm not sure that makes sense, however, given that some teams are inherently more physical than others, and teams like Navy and Army typically need to foul more often because teams are getting better looks at the basket than them. It may reinforce the idea that referees often like to call a game "both ways", despite the fact that one team may be more at fault than another.
Above is measurements of the school's defense. Again, you should be familiar with most of the top part of stats. They actually have a few interesting things to show us when you put them together for the whole league.
Look at offensive rebounds-- Lafayette got killed! (remember, this is how many offensive rebounds their opponents had) That's another indication that the Pards were more interested in running the ball than crashing the boards. Their opponents also scored a lot more points than they did against anyone else-- not to keep shoving this down your throat, but they were clearly looking for an uptempo game, and they got it. (Or is it that opponents wanted an uptempo game? Perhaps coaches thought the Pards could be beaten in quick game, but I doubt it.)
The second row has a lot of fun stuff. Bucknell is rated as the stingiest defense (see "offense rating", which really means defense rating, but you get the idea), and for good reason. They not only allowed the fewest points per contest, but they also allowed the fewest per possession. Note that their opponents had the second most possessions per game, but scored the least. Navy, Lafayette, and Colgate are rated the worst, with Holy Cross playing their typical excellent defense. So while Army played good defense, but were unable to overcome their terrible offensive woes.
I don't remember the Bucknell defense being that great, to tell the truth. But this is the kind of thing that shows maybe I'm an idiot that needs a computer to figure out what's going on. It will be interesting to see if they can keep it up this year.
Okay, now time to delve into some stats that are limited by the small sample size we have here (14 games are much tougher to analyze than the 82 in the NBA). They are also a little more complicated. But let's take a look, just for fun.
Points per possession divided by opponent points per possession is a way of balancing offensive and defensive numbers. If a team has a number over one, they scored more often than did their opponent. You'll see that while LC's defense was bad, their offense made up for it, and they come out with an overall rating of 1.05. But that was only good enough for fifth in the league. In the end, they couldn't keep up the tempo on the defensive end like they could on the offense, so they ran into trouble down the stretch (something I actually predicted and got write-- write down the date).
The other numbers are a little weird-- Bucknell's great defense makes them the best all-around team according to the stats. Holy Cross is also ranked above the eventual champions Lehigh. Why would this be?
The problem with these numbers is that they're combined to give a big picture of all 560 minutes of a season, rather than 14 segments of 40 minutes. So inconsistency doesn't mean anything in these numbers. If a team played very well one game and terrible another, it balances out to be mediocre here. If there were more games to analyze, then inconsistency would matter less.
However from these numbers we can imply that Holy Cross' inconsistency is what killed them. If they put up a consistent effort with the numbers above in every single game, then they would have been a contender in the league. Instead, they finished with a 7-7 record and a first round exit in the playoffs.
The pythagorean formula and luck numbers are just something I was fooling around with. If you care more about them, you can read about how it works here. The numbers really aren't that significant given they only go over a 14 game period.
Phew. That's a long post. Congratulations if you read through the whole thing. If anyone has any thoughts on this stuff, please let me know. I'm still trying to figure out how to apply this stuff to pre-conference play, using RPI and such, but who knows what that will result in. I'll have some time just before the new year to play around with the numbers more. Hopefully some of you will still be around to read it. @ 12:25 PM
