« Baracky | Main | Brewing Instability » Margin of Victory19 Apr 2008 09:05 am
Here's the margin of victory differentials for the Western Conference. As you can see, Utah and Phoenix both underperformed in terms of wins and losses relative to the margin of victory metric, whereas the Hornets and Spurs both overperformed. Margin of victory is, however, the better predictor of future performance. Comments (30)
standard deviation dude 
"Margin of victory is, however, the better predictor of future performance." Tell that to the Toronto Raptors.
Specifically, past margin of victory is a better predictor of future margin of victory
The average victory margin for Suns is suppressed even further by the the 116-86 rickrolling the Pistons delivered in Phoenix on 2/24/08 just after the Shaq trade when Nash & D'Antoni hadn't yet figured out how to play with Shaq. That was an aberration. Note to roundball punters: Detroit has the 2nd best point differential (+7.5) in the league and better than every team in the West. Boston was great this year (+10.3) and should be the favorite but it's not going to be a shock if Detroit comes out of the East. Detroit had a better differential than Boston against teams with above .500 winning percentages (+5.8 vs. +4.7) it's just that Boston was a crush machine against the dogs of the NBA. Why don't the Pistons blow out bad teams like the Celtics? The Pistons didn't need to develop a winning culture this season, they already have one. Detroit needed to develop a bench and make sure their older starters were healthy and rested come playoff time.
The Hornets and Spurs did NOT overperform. If the Suns and Jazz underperformed, the next two teams in line move up. Notice that if you remove the Suns and Jazz from your bar chart, the other teams line up precisely.
Bah. Sorry Matthew, but this is just number-crunching for sports nerds and fantasy league players. I bet the Dallas Mavericks last season had a pretty intimidating margin of victory with their 67 wins, and they were making their Memorial Day vacation plans in time to get a really cheap 21-day advance fare. At the end of the day it's about match-ups because we are talking about the playoffs, not the regular season. It's about how well one team matches up against the team they are playing in the 7 game series, it's about how individual players match up against their counterparts. I'm glad the Suns drew the Spurs in the first round, I'm far more confident about bouncing the defending champs than I am about us playing LA.
...besides, with the amazingly tight Western Conference this season, there really is not a lot that separates the Lakers from the rest of the pack. You could have a completely different seeding if a couple of games had a different outcome. The only team here that doesn't measure up favorably to the rest is Denver. LA is going to make quick work of them.
whoa whoa whoa -- joejoejoe, if the the Suns' loss is factored into their "margin of victory" these numbers are even more ridiculous & meaningless. a statistic measuring a team's "margin of victory" should include only, you know, victories. If this is just the differential b/n average points scored & average points allowed, that's not really a team's "margin of victory." It's a team's average point differential. also, sa lyons, what is your comment supposed to mean? I'm aware of basic terms in statistics, but just looking at the chart one cannot know whether the differences there are statistically significant. you know, just like how polls have different margins of error depending on their sample size, etc Just seeing 48-43 doesn't tell me much if, in fact, the small print says the poll margin of error is +/- 5.5%.
"Bah. Sorry Matthew, but this is just number-crunching for sports nerds and fantasy league players. I bet the Dallas Mavericks last season had a pretty intimidating margin of victory with their 67 wins" The Spurs last year had a better margin of victory than the Mavericks despite winning 9 fewer games. I wonder how that all turned out? ----- Past performance never predicts future performance perfectly, (which is why they bother to play the games), but some indicators are better predictors than others. And margin of victory has an oddly good record of prediction. And also, if you have a problem with "sports nerds", why on Earth are you commenting on a blog post about sports?
Ok, the relevant stat here does not sound like it could possibly be "margin of victory," so let's stop saying that. It's "point differential." A MoV stat would be really worthless b/c different teams might have different philosophies about closing games out -- blooding youth or not, and scores may or may not be "run up." Much like in college football, teams that run up gaudy scores on weak opponents do not always fare that well when matched against real competition. Point differential has the potential for much more meaning, but again, late game decisions seem like they would distort this quite a bit and that has to account for some of the variation in the chart. Also, let's remember that Western Conf teams above were all grouped together absurdly tightly in the standings, so it's not like there's much b/n them in the W-L category that produced the left-to-right ordering above.
"Point differential has the potential for much more meaning, but again, late game decisions seem like they would distort this quite a bit and that has to account for some of the variation in the chart." Margin of victory is useful as a predictor not because intuition tells us that it's useful. In fact, it's usefulness is rather counterintuitive. Margin of victory is useful as a predictor because history tells us that it's useful.
Point differential. Not margin of victory. (And I know I'm repeating a point made above, but it bears repeating.)
I agree with everything MY said in this post. Hollinger does this type of analysis all the time, and he was the only one I remember last year saying the Spurs were the team to beat going into the play-offs last year. At the same time, this doesn't take into account teams who underwent serious changes to their personnel - say Phoenix, Dallas, or LA - or teams who had significant injuries - like San Antonio and LA. The Spurs, for example, had 25 games in which one of the big three didn't play - including one game against the Clippers where Manu played 2 minutes without scoring or registering an assist or rebound. In those 25 games, the average margin was about 3.5, while in games where all three played, the average margin was a bit over 5. Standard deviations tend to increase based on this because of the smaller sample size, so it is harder to get statistical significance. Playing with the data further, I came across this interesting nugget for the Spurs. Only having two of the big three didn't affect performance much against marginal opponents but did affect a great deal against the best opponents (defined arbitrarily as a big game by me if it was against a top nine team out West or Boston or Detroit.) The average margin of victory was within a half of a point against the marginal teams, but over five points different against the best teams depending on whether all three of the stars were playing. Just playing with spreadsheets on a Saturday morning stuck at work ....
Really? - You are correct. 'Point differential' is a better term. I used 'margin of victory' to be consistent with Matt's post. The lesson? When in doubt, blame Yglesias.
This is wrong because it fails to take into account standard deviation. An example: Team A averages a margin of victory of 5 points with a standard deviation of 1 point. Team B averages a margin of victory of 10 points with a standard deviation of 10 points. Using this chart (along with some simple math) we can predict that Team A will lose less than .01% of its games while Team B will lose approximately 17% of it's games. If the Hornets play a low risk form of basketball that doesn't result in big point swings they can afford a lower margin of victory than the wild Suns.
All fair points, Curtis. And if the Spurs were definitely healthy going into the playoffs, you'd have to bump up their chances a bit. But as I understand it, Manu is banged up, and Tony Parker might still not be 100%. Plus their whole lineup is older. I expect the Suns to win that series, but it could go either way, it seems to me. And it'll probably be, at worst, the third-most interesting series in the whole playoffs.
This is wrong because it fails to take into account standard deviation. An example: Team A averages a margin of victory of 5 points with a standard deviation of 1 point. Team B averages a margin of victory of 10 points with a standard deviation of 10 points. Using a z chart (along with some simple math) we can predict that Team A will lose less than .01% of its games while Team B will lose approximately 17% of it's games. If the Hornets play a low risk form of basketball that doesn't result in big point swings they can afford a lower margin of victory than the wild Suns.
"This is wrong because it fails to take into account standard deviation. An example: Team A averages a margin of victory of 5 points with a standard deviation of 1 point. Team B averages a margin of victory of 10 points with a standard deviation of 10 points. Using a z chart (along with some simple math) we can predict that..." This is specious reasoning. It's a roundabout way of saying that number of victories is a better predictor than margin of victory, which is historically untrue.
Math heads: you can run your own pythagorean test with the various points for/points against numbers of NBA teams and translate Matt's entirely reasonable point into something simpler for non-stat heads. More formula info and stats below... Expected Winning %=(Pts scored)^(16.5)/[Pts scored^(16.5) + Pts allowed^(16.5)] http://www.rawbw.com/~deano/helpscrn/pyth.html
"It's a roundabout way of saying that number of victories is a better predictor than margin of victory, which is historically untrue" It could be true for a particular team, but not a common enough occurrence to effect the overall set of data. However, without seeing some concrete numbers of particular teams, or historical examples it does seem like an end around the facts.
OK. I was planning on skipping all four first round series in the East, but I'll watch WAS/CLE. DeShawn has sucked me in.
Speaking of DeShawn Stevenson, I'd like to start a rumor that he is secretly a Muslim and the bearding grow contest with Drew Gooden is just a cover story.
Manu is fine. He missed three games last week mainly as a precaution. He would have played if they would have been meaningful. And he lobbied Pop throughout to get into the games. I told my wife last night that I thought Phoenix had a 55/45 shot at winning that series. In truth, I think it is probably the other way around, but that is my way of getting a little chip on my shoulder for the next two weeks. The key to the series is the same as always when it comes to the Spurs - can Duncan be effective enough on offense to force the other team into doubling him? If yes, then it becomes bombs away time. If no, then the defense will have to hold the Suns below 90. And if Shaq refuses to defend the pick and roll, then Manu will break down the Suns defense repeatedly. Which is to say I like our chances. I don't love them. But I like them. As for the broader point of the thread, I think you could do similar things for basically every team. What is Phoenix's +/- in the first twenty games after the trade compared to the other 62 games? Same for Dallas, and so on. There should be nine pretty awesome series this year - all seven in the west, the ecf, and the finals - so let's get it on!
Basketball-Reference has the actual records and pythagorean record (expected win% based on points scored and points allowed) of every NBA team for each season. Boston is best in both actual and expected pythagorean win % this year. Here are the actual and expected pythagorean best teams going back several years. 08 actual Boston 66-16, pythagorean Boston 67-15, champion ____ http://www.basketball-reference.com/leagues/NBA_2008.html Health and personnel obviously play a big role in determining how good a team is in the NBA. The year the Pistons won the championship they got Rasheed Wallace in a midseason trade. Shaq only played 59 games the year Miami won the title. But all other things being equal, pythagorean record is a pretty good indicator.
"Speaking of DeShawn Stevenson, I'd like to start a rumor that he is secretly a Muslim and the bearding grow contest with Drew Gooden is just a cover story." The interesting already existing rumor is that Shaquille is a secret Muslim. There is supposed confirmation from the Turkish Hedo Turkoglu.
Edward Tufte would cry after seeing this bar chart. First off, the information about margin of victory is more easily absorbed in a table of numbers. Second, you should have arranged the teams by margin of victory, either highest to lowest. (I recently went to one of Tufte's conferences, so this sort of thing is on my brain lately)
Margin of victory may predict future success slightly better than wins and losses do. Hollinger certainly claims it does. Nevertheless it's hard to dispute that margin of victory involves a lot of meaningless statistical noise (how many minutes the scrubs play, for instance) that Ws and Ls doesn't, for the simple fact that the players are actually playing for wins and losses, and couldn't care less about average point differential. (sort of like delegates v. popular vote)
ChrisWWW - Tufte's books are awesome. I applaud Matt's attempts to add graphics to his blog but the primative "FIRE BAD!" quality of said graphics could definitely be improved upon. A Tufte seminar is $380 bucks and includes 4 kickass books. I recommend all you blogger types spend less money on Wii controllers and new games and more on improving your data presentin' skillz.
But wins and losses have a lot of meaningless noise in it, as well, when it comes to projecting a winner of a play-off series. We only have 2-4 games with basically no preparation time ahead of them for the regular season to use head to head. But matchups play such a key part of any play-off series, and the amount of preparation time increases dramatically. Why should we think basically meaningless games in Memphis in November should predict the outcomes of playoff games in April? That seems as noisy as anything else.
Comments closed May 03, 2008. |
Search
Subscribe to this blog by Email Books by Matthew Yglesias
Recent Comments
Recent Posts
Weekly Archives
Categories
|
Copyright © 2007 by The Atlantic Monthly Group. All rights reserved.
| 
Heads in the Sand
"Margin of victory is, however, the better predictor of future performance."
That sounds like a ridiculous statement. Where's your proof for that?
Just intuitively, great teams might blow out other teams throughout the first 40 or 44 minutes of games and let their chumps play out the clock, and the margin of victory goes down.
Also, look at your chart. The MoVs here range from less than 7.5 to what looks like 3.5 or so. That's really some statistically significant variation??
I mean, to say that the Spurs & Hornets "overperformed" is ludicrous, when you're comparing the Spurs (apparent MoV of 4.9) to the Suns (apparent MoV of 5). Are the MoVs for the Hornets, Spurs, and Suns even significantly different? I doubt it.
Posted by Really? | April 19, 2008 9:51 AM