Thanks to bmd and his links I may be able to start working on what I call the Offensive Contribution Points (OCP). The basic premise is simple - the game is really a team game but the normal stats only credit the scorer of the point when in reality it took a few passes (and involvement of other team members to get the chance to score). OCP will try to work out who have contributed to a scoring play. A scoring play is a sequence from the point when the team is in control of the ball e.g. a defensive rebound by Player A who passes to Player B to dribble up the court who then passes to Player C, who may fake a pump and then spotted that Player B had run up for an Alley Ooop and lobs to him for a score - Player B may then be fouled in the process for an extra FT and converts. So essentially a 3-points play but seriously the points should be shared by Players A, B and C. For simplisticity sake - I will take an offensive rebound of a missed shot as a continuation of the play - i.e. play ends when attacking team loses control of the ball to its opponent. Take for example our last game against the Raptors, Rockets in total scores 110 points out of 55 plays (one of them was a tech) Our 9 players would have accumulated the following OCPs AB0 - 0.2 PB2 - 7.5 OA3 - 2.9 TJ6 - 6.4 JL7 - 26.57... DH12 - 11.8 JH13 - 31.18.... OC18 - 8.27... CP25 - 15.17... To make OCP more meaningful to compare across players - we should divide it by the minutes played (rounding up for players not playing the full minute) - so the OCP/m for our players would be AB0 - 0.2 PB2 - 0.3 OA3 - 0.15 TJ6 - 0.19... JL7 - 0.58... DH12 - 0.26... JH13 - 0.65... OC18 - 0.31... CP25 - 0.30... Which is a surprise to me since I would have expected DH12 OCP/m to be higher, as it is, for the last game our leading offensive players were JH13 and JL7. Will try and analyze the other games to see how (I am in the midst of working other analytics so for those interested - my apologies if my updates are slow). have fun! Go Rockets! (PS - yeah, I am also thinking of Offensive Demerit Points, Defensive Demerit Points as well as Defensive Contribution Points but those will be further down the pipeline).
I may have interpreted this wrongly, but wouldn't your score be calculated in a way skewed towards the PG or ball carrier? Seeing as they always get a pass in every possession, often not as crucial (and therefore should not be as equally weighted) as the assist pass
Absolutely the PG will get the highest count in this system..... Seems whack to me. PG gets credit for getting the ball over the halfcourt line, and giving it to someone else who sets up somebody else for a score. Fail.
Not sure why you're shocked, considering that Daryl Morey has stated in public interviews that JLin is the highest impact player on the offense besides James Harden, and in his view is the #2 offensive option, as well as the player that is expected to step up when Harden struggles. For all DH12's strengths, generating team offense is not one of them. Like Chandler Parsons, he benefits heavily from play makers like Harden and Lin. The other factor is Dwight's free throws. Dwight's poor free throw shooting impacts the team in numerous ways. First it reduces his effectiveness per possession. It's hard to score when the opposing team would rather foul you than let you score. Second it's hard for Dwight to earn the confidence of team mates in terms of passing to him late in games when they're aware that he's going to be intentionally fouled. Third it's hard for Dwight to maintain his own confidence in terms of demanding the ball in the post when he's aware that no matter how hard he works on his post game, there's a good chance that many opposing teams will automatically foul him before he even gets the pass, let alone gathers the ball or takes a shot. Now if he could get his FT's to 60% or better, his offensive impact would be significantly larger, not just in terms of players passing to him more aggressively in late game situations, but most likely in terms of his own willingness to aggressively seek for scoring opportunities. D12 is huge defensively, not so much offensively... at least not yet.
After the two guards that initiate almost every play, the third on the list is Omri Casspi (Chandler Parsons basically the same). That's interesting because Casspi seems to be involved a lot more than he actually scores.
DrN if you look at bev (who's supposed to be doing pg duties) he isn't really scoring very high compared to his other team mates so no this is not skewed to pg only in fact it is a method that start to count contributions from the other players
It's not obtained from another place I had to look through the 55 plays that our rockets obtained points from and see who had been involved in them. Say for example a play: Bev brings the ball up passes to harden who then passes to parsons who then drives in but see that his path is blocked and passes back out to harden who then takes a 3 pointer and scores with a and 1 and then converts the ft. So this play was worth 4 points bev gets 1 point, harden gets 2 points and parsons get 1 point We are basically trying to get ocp awarded to players involved in the lead up to points obtained.
Maybe Morey is using something similar to track, in my mind though I am tryi g to think how I would be able to make use of ocp/min - the main one I can think of is as a gauge to check if it is worth adding playing time for players. I.e. if we shorten harden's playing time and see an improvement in his ocp/min we may want consider it, or more importantly extending a player's minutes leading to a drop of ocp/min would be a clear indicator that's too much time given to him. Coud be worth considering
So if Bev or Lin get the first pass after a rebound, immediately gives it to Harden who comes up court and scores on a PnR with Howard... Bev/Lin gets a point in this system? That makes no sense.
No, the points gets divided, so for your example the rebounder lets say it's Asik, passes to bev who passes to harden who passes to Dwight and 2 points was scored, then Asik, bev, harden and Dwight gets 0.5 points each. However if it's a stop start situation, we don't count the inbounder.
On that hypo, what credit does the pick setter get in the PNR if the guard scores instead of dumping off? The pick setter is obviously a critical part of the PNR but would not get credit for the basket since he didn't touch it.
To piggy back on this point, this doesn't account for spacing that allows for the basket. Also there is a gap in difficulty between each of a players actions in a given play (grabbing rebound, making pass, hitting shot). It seems like your point distribution needs to be skewed more towards the person hitting the shot. This is interesting in that its a summarized map of the route of the ball. I'd actually like to see this on the level of detail of how many touches per point, and the most common paths. While it difficult to use this to compare PG with forwards and centers, if you did this for the whole league, you could figure out benchmark values to see above average or below average. I get that Beverley scores low on this because he doesn't initiate the offense as often as Lin. I don't know that this is especially surprising, but it'd be cool to validate whether those plays start with Harden. Also, if you did the same thing for misses, we could see the least efficient paths as well. Thanks for putting in the work, some stuff to think about at the least.
Your points are actually accurate but your conclusion is wrong, as the pg IS typically the most important offensive creator in the game. It's an accurate reflection of how our offense works. Of course, it doesn't come close to truly reflecting offensive contributions as you would have to account for all the confounding variables as well as nearly every other variable that you can think of. As in offensive rebounds, fouls created, passes that led to free throws, the timing of the passes (was the basket scored within 3 seconds of the pass?), etc. But it's not like assists aren't under the same issues that you mentioned. As in, a simple pass to Dwight after which Dwight plans his move then jukes and scores or a simple perimeter pass that results in a score should be counted "less" of an assist if that makes sense than one where a player breaks down a defense and basically creates a dunk, layup, or open 3 pt shot. Anyways, I look forward to seeing how your model evolves, but if you can please note the site where you reference the hockey assists etc. as selection bias can be easily introduced even when things may seem straightforward.
Here's a link with similar stats, but a bit more subjective, in which the "analyst" awards good credit and demerits based on every little variable such as deflections, passes that lead to an open shot make or miss, etc for both offense and defense! http://www.red94.net/stats-say-jere...in-playing-like-might-worth-15-million-salary red94.net
I see many people here are saying that the PGs are getting freebies because they are bound to dribble the ball first. For example Lin or Beverley dribbles it over half court and Harden gets the first pass. That means Lin or Beverley and Harden are favored. In the OP's defense the title for this analysis is "Offensive Contribution Points" so it's bound to be biased towards the attackers which are typically PG and SF. To resolve the PG and SG bias just allocate less points for the above mentioned touches.
You are correct it does not cover all the angles, but I was looking for something simple to track and player touching the ball is a clear indicator of involvement in the play. If we start to count every player for each play we'd end up something similar to the general +/- stats Look on the bright side at least with this we can start recognizing some other players besides the final passer (assists) and the scorer (points) I personally don't think we should use ocp/min to compare between players but maybe use it as a gauge to check on whether a player is being used correctly, I.e. we should compare ocp/min of a player game to game this might tell us if we are playing him too many minutes ( likely to show up as a drop in this fig) or if there is something else wrong (e.g, playing him in the wrong role etc.) Just some thoughts...
Managed to look through the game of HOU vs PHI (13 Nov 2013) - Rockets managed to get 117 points out of 57 successful plays (one difference this time round is I am counting offensive board as a reset to a play - this is to req
