So our fast pace is only good for home games? And on the road we need a better D even if it damages our O?
I've got it! The answer to our problem! Let's all put bra's on our heads and hook up our TRS-80 pc's to a doll and create a defensive superstar
Very nice job! Just need to note Parsons picks up the tougher assignment among SG/SFs and sometimes even PGs.
The following data is interesting too. Copied from another thread. To be verified. Harden TO 3.7avg 6TO 2W7L, 22% winning% 4TO 13W20L 39% 3TO 18W7L 2% 2TO 11W3L 79% 1TOor less 8W 100% Parsons 2.0avg 4TO or more 2W6L 25% 2TO or less 23W16L 59% LIN TO 2.9Avg 1TO 3W6L 33% 3TO 18W22L 45% 4TO 13W6L 68% 5TO 6W2L 75% 6TO 4W 100% from http://bbs.clutchfans.net/showpost.php?p=7689272&postcount=1373
Great stuff! But then I love NBA analytics, which not everyone's cup of tea. But since you have all the data in a spreadsheet I would love to see: 1: Team Ortg and Drtg on a per game basis color coded for W/L. Or at least the 5 players you focus on. 2: Lineup Ortg and Drtg for the most used lineups like at 82games. I do realize that the trade deadline screws this up, but it would be interesting none the less. But I only had time to skim over the graphs. I will come back to this thread later.
so in other words, don't try to restrain our players from playing the way they know how, because sometimes the eye test for turnovers can be deceiving.
I'm pretty new to this, thus I have only get these 5 player's game log data, and I don't have the team Ortg and Drtg though. I'm pretty sure it should be somewhere, just i don't know. If somebody could point me to places that I could down load those info in large batch.......
you cant decide games by turnover numbers because you don't know how the turnover was committed. For example someone makes a good pass to Asik but Asik can't catch it, it would be a turnover for the one that passed the ball. I would blame it on Asik if he can't catch a good pass, but the stats would say that the passer committed the TO. And I feel that bad shots should also be committed as TO, because if you just shoot the ball and miss, while a player is open and a better shooter, the ball will usually go to the opponent.
First of all, thanks for doing this analysis. Very interesting stuff. And here's some rep going at you for all of this. A few things to note about the analysis/interpretation of the data. On the box plot, what would make it a bit more informative is if there were some hypothesis testing so we can see whether there is a statistically significant association between a player's ORtg / DRtg with the team's wins/losses. Doing this in a simple OLS regression, you could then easily add in other controls, like opposition strength (Ave Pts for and Ave Pts against), days rest, home/away, etc. I really liked the scatter plot, because it really highlights what aspects of each player's game is correlated with wins and losses. I agree that Harden's scatter plot is the most interesting, but I will disagree a bit with your interpretation that if he does better on defense then we will win. First, if you look at the wins/losses to the left of the origin (i.e. when Harden has a bad offensive night), it is clearly the case that we generally lose. We only have 3 wins when he has a bad offensive night. So, it's fair to say that it is a necessary condition for us to win that Harden has a decent game offensively. Now if we restrict our view to games where Harden has good offense (i.e. to the right of the origin), that's where Harden's defense becomes important. Conditional on a decent offensive game, we always win if Harden's DRtg is below 104 or so. I 'm not sure the vertical line you drew in Lin's graph is actually significant. Generally speaking when Lin's offense is above 100, we tend to win. But, the fact that there isn't a very clear line that can be drawn between wins and losses for Lin's plot suggests that Lin's performance is not central to whether we win or lose. Not like Harden's offensive performance--which if his ORtg is low, we lose. Parsons is interesting too because, as you point out, we tend to win when his DRtg is below a certain point. And this line appears much more pronounced to me than Lin's vertical line. Since Parsons is usually assigned to match up against the opposing team's best wing player, this means his effectiveness against those players is a real key. This is first and foremost Parsons' responsibility, but it's also about how well other guys help out on his guy. Similarly, when Asik is effective on defense, we win. So it seems to me the main takeaways here are: 1) Harden has to have a decent offensive game for us not to lose 2) We have to play good defense, especially Parson, Asik, and Harden.
Because Jeremy is not merely a Rockets player -- he's a SUPER player! He's the most amazing player in the universe!!
very interesting. So I generated this view to look at TO distribution histogram (sort of) to compare Lin and Harden. The line curve is a rough curve fit for distribution, and bar indicate for each TO number, how many games were win and how many were lost. Usually one would expect the win curve (red) should peak at the left side of the lose curve (blue), that is, collectively, one should have more game win when they have less TO, and more game lost when they have more TO. But Jeremy have very unique curve fitting, whereas red curve peak at the right side of the blue curve. Then you compare Lin and Harden, for the blue curve, Harden peak at 4 TO, and Lin peak at 3 TO, basically that says, with less TO, Lin tend to lose the game. And if you add the number together, for TO less or equal to 3 per game, Lin lost 13 and win only 5. And with TO > 3, Lin win 7 and lost 5. For Harden, with TO less or equal to 3, Harden have 6 wins and 5 lose, and with TO > 3, 6 win and 13 lose. I think Lin's TO really means different from other people's TO, and we could not just take the number as is.
yeah, agree with you, Jeremy's line is really iffy. I think your two takeaway message is more accurate. I might trying to read too much there. In terms of correlation analysis or calculate statistically significance, I think with this limited sample size, with this small changes, it would hard to find statistical significance trend, that's why I'd like to see the real distribution of the data point to get direct impression, rather than numbers alone. I still think human eye-brain coordinate does much better calculation than computers. Plus, a lot of times, the tightness of the data would have more impact on significance than actual changes from two groups. With a young team, we have plenty of variation and scatter of player performance from game to game.....
Great work! What is the sample size? How many games did you look at? The first graph seems to correlate game losses with Harden's TOs. This confirms my opinion that coach McHale should focus more on Harden's TO's and not Lin's TO's. In fact other people like you have noted that we are more likely to win when Lin has higher TO's, probably due to more aggressive, risky passes.
Believe it or not, he is one of the best known NBA players (currently active), even though he is just an NBA starting PG, average at the best for the moment.
Excellent first thread! Keep it up! Basically the Rockets need to improve their defense, and Jeremy should be a little more aggressive in his play. Both of these points pass the eye test for me, so it's nice to see numbers backing it up. Well done