http://www.armchairgm.com/NBA_Point...Most_Power_Stat NBA Point Differential - The Most Power Stat In yesterday's Predicting the NBA article, it was determined the quickest, most accurate measurement of NBA success to compare how many points a team scores per 100 possessions against how many points they gave up per 100 possessions; however, we only ran a test on it for the Eastern Conference teams from the last NBA season. That is not a very accurate means of determining reliability in predicting the future. In order to cure this problem I took the difference between the two values for every single team over the last 10 years (The Bobcats only have existed for 3 years and I adjusted the 98-99 season to 82 games because of the lockout) and mapped out their point differential against their wins in the season. The spreadsheet is much to massive to really post in any form for viewing on the webpage, but if you would like to see it/play around with the numbers it can be downloaded here. If you just want to see the end results though, here is what good old Microsoft Excel shoots out after being handed all the information. What the resulting graph's best fit line states is that Wins=2.4168*PointDifference+40.959. Now what exactly does this mean. Well, as expected if your team manages to simply score and give up an equal number of points per 100 possessions you will win about 41 games (exactly half of a given team's 82 for the year). From that point it states that for every one point more you can score rather than give up your team will usually increase their win total by 2.4 games. Likewise, if you surrender one more point than you score you can expect to have your win total decreased by 2.4 games. Using this method of predicting NBA final standing came up with an R^2 value (see here for details) of .9402 which meant that our data fits very closely to the resulting equation that was just discussed. Besides being able to help predict what a teams final win total will be for an NBA season, we can also use to it find some interesting, if in the end not too meaningful numbers on some NBA extremes. According to our graph, if a team was to lose every single game it played it would need a points per 100 differential of about negative 17 points games meaning they would basically have to be so bad they that were hardly ever in any games at all during the course of the season. On the other hand, if a team wanted to win all 82 games in a season, they would need to be approximately plus 17 points per 100 possessions for the season, basically meaning they be completely dominant in every game. Now, even if teams were able to post numbers like those during the course of a season it is unlikely they would post those records logically (even statistically it is very questionable since we are extrapolating into areas where we have no data), but it is fun to imagine. Another interesting thing which can be observed from the points per 100 possessions data is that since Steve Nash joined the Suns and the hand checking/physical contact rules have been more strictly enforced in the last 3 years, the NBA has seen scoring per 100 increase each of the last 3 years (106.1, 106.2, 106.4) but all three of those years have been the record over the last 10 years when they were set. So, the NBA is becoming a more offensive minded league which makes sense in the light of so many teams wishing to now become like the Suns who have posted amazing per 100 numbers since Steve Nash joined them in the 04-05 season (114.7, 111.6, 113.9). That is what makes the Suns so deadly and is often missed when talking about their uptempo style. They play extremely fast, but they also play extremely efficiently even when on a possession by possession basis likely due to the fact that their quick pace leads to many easy shoots for them throughout a game. What is also a little surprising is the who won a given year's championship based on the point differentials. The 97-98 Bulls, 98-99 Spurs, 99-00 Lakers, 04-05 Spurs, and 06-07 Spurs all won the NBA title and the point differential crown in the same season. Therefore, 50% of the time the NBA champion could be predicted by simply looking at who won the point differential in the regular season. There are times when other circumstances (teams have injured players come back and they improve immensely or lose players in injury) may make it seem unwise to predict the champion based solely on that number, but to be able to guess the NBA champion 50% of the time based solely on one simply number is amazing. Only 4 times did the league leader in wins during the regular season win the NBA title (97-98 Bulls (tied), 98-99 Spurs (tied), 99-00 Lakers, 02-03 Spurs (tied)) and 3 of those were tied so if you took the good statistical approach and flipped a coin to break the tie on each of those years wins during the regular season would only have a 25% of giving you the correct NBA champion. So the point differential is twice as effective using wins to predict the NBA champion. Another funny note that can be tested using this numbers is the theory that "Defense wins championships". The best offensive per 100 team has never won the title over the last 10 years. Meanwhile, the 98-99 Spurs and 04-05 Spurs won the titles while playing the best regular season points per 100 possessions defense. So, defense did win infinite times more championships than offense over the last 10 years, it was not as if playing the best regular season defense in the league guaranteed anything more than a 20% chance of winning it all in a given season. In the future I would like to test the adage that "rebounding wins championships", but I'll have to gather the data for that study at a later date. One final note among the many more than could be pulled from the data, who has been the best overall organization on the basketball court on the last 10 years based on point differential. In a complete blowout, the San Antonio Spurs with an average point differential over the last 10 years of 7.86 (the next closest was the Detroit Pistons with a +3.71). This is not surprising considering that they have won 40% of the titles during the past decade. The Spurs also posted an average defensive value of 98.1 over the last 10 years which also stomps the next closest competitor (Philadelphia and Indiana at 102.6). They only area they did not manage to dominant was in offense over the last 10 years where they were only able to be a top 10 team at 105.96 (Dallas was the best at 108.56). This is a great testament once again to Tim Duncan, the only constant for the team as a player over the last 10 years. Combined with his individual statistics, these teams stats along with his titles should be a sure path for him to be called the greatest power forward in NBA history. For hilarity's sake, the worst NBA offense over the last decade was the Chicago Bulls even with their last championship leading of the decades as they were the only sup-100 team at 99.6. The worst defensive team of the last decade was the LA Clippers at 107.4 who posted some truly terrible numbers in both offense and defense over the last decade overall and in certain years (the Bobcat would be last at 107.466, but they have only played 3 years so it would not be fair to laugh them out of the building yet). In conclusion, these point differential numbers are extremely useful, I am willing to bet that using these alone with enough comparison work (looking at who won vs. who in the NBA playoff series in the past based on differential and how many games the series went) it would be possible to out predict all the NBA "experts" who give 30 second blurbs of information concerning who they believe will win the title and certain series without watching a single minute of game action. This is something I look forward to testing further in the future during the coming years to see how well it holds up in predicting power. If anyone else plays with the numbers and comes up with another interesting observation (there are so many possibilities still remaining I haven't even tried looking at yet) don't hesitate to contact me about it. I would love to share everything the data has to offer to the readers of this site. For now though, just ponder how true it is when someone jokingly replies to the question, "How are we gonna win?", with the answer, "Score more points than the other team". So who are the leaders in point differential this year? Here are the top 5 for the season. Of course the Rockets started off the year pretty badly so it's lucky we are even in the top 5 in point differential for the season. 1. Celtics +10.8 2. Pistons +7.63 3. Lakers +7.5 4. Jazz +6.26 5. Rockets +5.3 6. Spurs +5.25 7. Hornets +5.0 It is interesting to note that the Rockets during the last 10 games are averaging an eyepopping +16.6 point differential best in the NBA. The Celtics are second averaging 13.2 and the Lakers third averaging +11.9. This bodes well for the Rockets going into the playoff season. The defending Champion Spurs during the last 10 games are averaging a pedestrian +4.5 point differential.
Very nice... I live and love stats and was mesmerized by your post. Too bad many people don't appreciate stats. I believe that everything can be quantified and explained by numbers. So, thanks for taking the time to perform the analysis and post it. Since you have the data already, could you post a plot of the mean of the last 10 years (point diff) for all teams with one standard deviation (or standard error)? I wonder how much overlap exist between the teams over a 10 year period since the overall roster structure of a team does not remain the same during that time frame. Once again thanks for your work, jsmee2000
Nice post, but this year's imbalance between the conferences has distorted the Celtics'/Pistons' pt differentials. In fact even after the entire regular season is played the sample size might still be too small to be 50% confident that the regular season pt differential king (undoubtedly the Celtics) will win the ship.
Point differential is probably the best stat but as with any stat there are two major flaws with it that make it difficult to judge. 1. Teams like Boston and Detroit may have an inflated differential based on an easier schedule - so it's hard to determine across conference. 3. As point out - the rockets are a team in flux - so looking at the 10 games makes sense, but it's a small sample size and doesn't match against the schedule of other teams (not apples to apples). Therefore - it's hard to really say much about our point diff relative to other teams this year. I think you would have to take the teams we've played during the second half of the season and compare them against the same group that the other teams have played - and weighting out the impact of playing a team multiple times versus other teams that played it once. Even so - the variablitity would be pretty high. So it's still very uncertain when it comes to measuring the Rockets potential. Are we a +5 team, or a +15 team - I don't think it's possible to determine that accurately through statistical means.
Anyway, we are west best 3 in this stat. Our defense is proved, the problem is whether we can keep our offense in the playoffs.
WOW, another crappy "advance stats analysis"? There is no problem with the linear correlation between Point Difference and Wins (Winning Percentage). But, what does "linear correlation" mean? Why Point Difference is so much more important than Winning Percentage? Why Winning Percentage is so meaningless, and Point Difference is so meaningful? Point Difference really must be something.
I honestly think with Adelman we can keep the offense going. He somehow makes even mediocre players look great.
just put him on ignore, he likes to spout off nonsensical arguments (offense more important than defense without any proof other than "because i say so") and in general antagonize other posters.
how about weight it like this, lets say Nets' pt diff is -5 for the past 10 games, we beat them by 15pts, so adjusted pt diff we won for this game is 15*(100-5)%, if they are +5 for the past 10 games, it'd be 15*(100+5)%
FYI those stats you posted are the differential in points per game, not points per 100 possesions like that guy used. I'm still mulling over whether that makes any difference or not.
Interesting thread. I hear all the time that coaches tend to like point differential to evaluate the strength of a team as an alternative to plain old win %. Hubie Brown, in particular, seems to refer to it a lot. He knows a thing or two about the game. There was a nice discussion on point differential as an estimator for win % and as a descriptive measure of how good a team is in this thread at the APBRmetrics board.
Ahh, if ever there was a post for Durvasa to chime in on, this is it. lol To be fair, the sample size of 10 games is small but shows the Rockets and how they are currently playing. I would definately throw out the 2007 statistics because they are not indicative of what is going on now. However, if you want a larger sample size, I would say we should just use all the 2008 games played thus far and see what happens to the statistics. I dare say that the Rockets are at the top. In fact, I'll do that right now for the top 3 teams in the NBA. Rockets/Celtics/Lakers ok I'll throw in the Spurs too simply because they are the defending champions. Rockets pt differential 2008 Celtics pt differential 2008 Lakers pt differential 2008 Spurs pt differential 2008 What this data says is that Garnett doesn't matter that much to the Celtics. The Lakers are our most dangerous opponent but the Celtics are not far behind. The Spurs are very far behind and likely not to repeat as NBA Champions.
But we do know that we are somewhere between +5 to +16. Anyway you slice it, we are an elite team. Sweet! Great stats, again!
I see a dogfight to the championship between the Lakers Celtics and Rockets. If we get past the Lakers, I think we take out the Celtics in 6.
Just for kicks, I thought I'd put up the pt differential for the last 10 games for all the NBA teams to see how everyone is playing lately and how we stack up.
