Proposed method to track performance of our starters

Hooray for my predictions being wrong - glad that we could beat the Bulls! Yes some of our players weren't up to par but hey it's great to break the losing streak!

Unfortunately, I am still working on my new method to compare whether our players are up to standard - so please bear with me as I continue to slog on it. But here's a gist of what I will be working on:

(1) Again it is baselined against all players playing in the NBA - but I am not targeting specific persons but comparing PGs to PGs, SGs to SGs, SFs to SFs, PFs to PFs, Cs to Cs as well as Bench to Bench.
(2) We have 30 teams in the NBA - so my criteria will be the top 10% or 3 highest score is to be considered Elite - we have many member who like to keep claiming someone is elite or someone really sucks etc. but there is so far no statistical basis aside from their personal eyeball test. Essentially, I will be breaking the scores of the 30 in each category (i.e. PG, SG, SF, PF, C and Bench) into 10 blocks of 3 and I will be determining their PPM (points per minute played), RPG (rebounds per game), APG (assists per game), SPG (Steals per game), TPG (turnovers per game), BPG (Blocks per game).
(3) The bench calculations is a tad complicated but I will be attempting to normalize them as a group (as sometimes the bench has different number of players - so the normalizing will roughly estimate them into a single anonymous player. Essentially, I will be working on the scores from the first 8 games of all the teams - and normalizing that with the averaged number of bench players played. PPM is not an issue as it would be the aggregate score divided by the total time played by either player(position) or bench.\
(4) Hopefully, at the end of it - I would have a method to gauge if a particular is performing at 10-segmented scale within a particualr facet analysed - i.e. Elite being 10 points down to Scrub being 1 points and yes with the method I am doing - it may even be possible for some players to score way above Elite or even get below 1 (or maybe negative score) - I am also hoping we might be able to look at the composite scores of the 6 areas might also help us to have a basis to discuss if certain players are indeed suitable for consideration to the Elite class or close to it.
(5) Am considering the following 10 sub-divisions: Elite, Superstar, Star, Good, Decent, Average, Underperformer, DLeaque, Spacefiller, Sucks.

In the meantime, here's the PER score and rankings comparison between Houston and Knicks (not that it is a useful tool at all for predictions):

Houston:
JLin 11.98 (45)
JHar 20.24 (7)
CPar 13.61 (36)
PPat 14.03 (49)
OAsi 11.07 (49)
Totals 70.93 (186)

NYK
RFel 18.46 (20)
JKid 22.10 (5)
RBre 15.44 (24)
CAnt 22.06 (9)
TCha 20.44 (15)
Totals 98.5 (73)

It doesn't look good for Houston from the past statistics - but hey, hopefully the rule of averages kicks in and bring NYK performance down enough for our rockets to win it - Go Rockets!

 

Okay folks, I am back with a new method to track our team performance - Great win over the Knicks!

Here's a lowdown - please note that this is not a prediction tool (so for those looking for a method to guess what our next result would be - sorry).

The basis is again comparison of our players against their peers. Basically I will be making use of the basic NBA stat data set (available from NBA.com). I will be comparing mainly the following positions (note I am not zooming to individual players) - i.e. SF, PF, C, SG, PG and the Bench (as a whole).

The current baseline I am working on is only up to 8 games deep per team - my next release of comparison is likely when all teams have played out at least 16 games (i.e. I am roughly doing it around 10% of the season).

I will also be only comparing the following 6 measures PPM (point per minute played), RPG (Rebounds per game), APG (assists per game), SPG (steals per game), TPG (turnovers per game) and BPG (blocks per game).

I developed a 10-point scoring scale by rationalizing the 30 players into 10 groups of 3 - the top 3 getting 10 points, the next 3 9 poins until the bottom 3 gets only 1 points. So with the 6 measures - we can get a composite score for the positions.

I have also roughly categorized the 10 levels as: Elite, Superstar, Star, Good, Decent, Average, UnP (Underperforming), NMD (needs more development), PTW (painful to watch) and BOP (bottom of pile).

So for our first installment of 8 games - here is how the Houston Rockets did in the various positions:

For SF - we got a composite score of 29 - this is roughly ranked around 22 out of 30 (my rough level puts this at NMD! For comparison's sake teams with Elite players in this position are: POR, MIA, OKC.

For PF - we got a composite score of 21 - this is BOP (yep rock bottom). We really need help in this department. Teams with Elite players are: SAS, UTA, LAC and PHX.

For C - we got a composite score of 32 - this is at Average level. Teams with Elite players are: LAL, DET, CHI, MEM and UTA.

For SG - we got a composite score of 46 - yes this is at Elite level. Other teams with Elite players are: PHI and MIA.

For PG - we got a composite score of 41 - this is at Superstar level. That's a surprise (but since scoring is only 1/5 of the composite that may make sense). Teams with Elite players are: CHA, MIL and TOR.

For Bench - we got a composite score of 40 - this is at Star level - which is another surprise (with the amount of complains going round but statistically our bench is holding out just fine). Teams with elite bench are: DEN, MIL and ATL.

Once again - would like to appeal for feedback and comments - and whether this method at least has some correlation to those of you with your "eyeball" tests.

Have fun and yeah looking forward to the next game against the Raptors - as a side note comparison our guys scores versus their guys look like this:

Houston
SF -29
PF- 21
C - 32
SG - 46
PG - 41
Bench - 40
Total = 209

Toronto
SF - 18
PF -30
C - 28
SG - 36
PG - 45
Bench - 23
Total = 180

It looks like a Rockets win - Go Rockets!

 

Not for this no. My main peeve with FG% is the difficulty to distinguish between 1/1 which is 100% from 5/5 which is also 100% but obviously 5/5 shooting is way better than 1/1 shooting.

I am considering PPM (point per minute played) - so I put in for 20 minutes how many points have you gotten me? Mind you - this is not comparing you to the time played but in comparison with the other 29 players playing the same position you do - so if they can score more than you - you obviously need to improve.

One mistake I made in my post was I mention shooting was 1/5 of composite score - this was inaccurate - it would only account for about 1/6 - since I am considering PPM, RPG, APG, SPG, TPG and BPG. It is obvious that it is unlikely any player may get super good in all 6 areas - in general the Elites would score well above 40 points out of a total of 60 points.

 

Okay - it was a very nice win - especially with 3 of our starters getting double-doubles!

Anyway here's how it looks for the game against OKC.

HOU
SF -29
PF- 21
C - 32
SG - 46
PG - 41
Bench - 40
Total = 209

OKC
SF - 47
PF - 34
C - 25
SG - 33
PG - 41
Bench - 23
Total = 203

It's gonna be a close one but I think our bench may hopefully pull us through this - go Rockets! and yeah Go Harden too - you bearded one - go and show them thunder for thinking you are not good enough to start.

 

Well the game against OKC was a downer - sigh CP being injured didn't help - but I think our bench did not play well too. On to the next game - with regards to my tracking - am still waiting for a couple of more teams to chalk up their 16th games before issuing another comparisons.

But for those of you interested in my new method, here's what it says of our team performance (comparing 1st 8 games to all teams):

SF position - we are ranked 22nd out of 30 teams (but I think this will improve when we look at this after 16 games)
PF position - we are ranked 29th out of 30 (bottom) - we really need to work on this but it should probably improve after 16 games).
C position - we are ranked 16th out of 30 (pretty decent)
SG position - we are ranked 2nd out of 30 (may drop a bit after 16 games)
PG position - we are ranked 5th out of 30 (I expect this to drop a bit too)
Bench - 9th out of 30 (I think this may drop as well).

anyways - comparing Utah with us

Houston 209 vs Utah 224 - it looks like a loss once again for us - but am still rooting for a rockets win - Go Rockets!

 

Well the game against OKC was a downer - sigh CP being injured didn't help - but I think our bench did not play well too. On to the next game - with regards to my tracking - am still waiting for a couple of more teams to chalk up their 16th games before issuing another comparisons.

But for those of you interested in my new method, here's what it says of our team performance (comparing 1st 8 games to all teams):

SF position - we are ranked 22nd out of 30 teams (but I think this will improve when we look at this after 16 games)
PF position - we are ranked 29th out of 30 (bottom) - we really need to work on this but it should probably improve after 16 games).
C position - we are ranked 16th out of 30 (pretty decent)
SG position - we are ranked 2nd out of 30 (may drop a bit after 16 games)
PG position - we are ranked 5th out of 30 (I expect this to drop a bit too)
Bench - 9th out of 30 (I think this may drop as well).

anyways - comparing Utah with us

Houston 209 vs Utah 224 - it looks like a loss once again for us - but am still rooting for a rockets win - Go Rockets!

 

Yeah I am using spreadsheets - not so sure about combining the guards/forwards direction - we would end up with just 2 stats - if we go the way NBA is doing for All-star - only 2 guards and 3 forwards (imagine Center being absorbed). Agree that there have been changes in roles and multiple switching - but I am rather old-school in this - I still think the 5 player roles is quite neat for Bball.

Is there a way for me to attach the excel file to our pages? I am still trying to figure out how to post a table here. Or can I simple cut and paste?

 

Thanks just a word - your points are valid, our challenge is a difficulty to find basic stats to support some of them and be able to use them for analysis. Agree that somewhat for RPG some of it is affected by shooting FG% - the difficulty of course is determining how much of it is contributed by poor shooting (there is a % problem for us would be this % is likely to vary across games) - one approach may be to adjust the final score by a factor. I am not too overly concern at the moment since we are just doing peer comparison and at the end of the day we are ranking each position against the rest of the 29 teams - we are looking at relative performance (so hopefully some of the hidden factors cancel each other out - but we are not sure). At the moment, I am using total rebound.

Is there an easy way for me to get the other ratios you mentioned broken on a per game basis? I think they could be useful but at the moment might be a hassle to incorporate.

There is in fact another method of tracking I am toying with - but it's mighty hard to do for me at the moment - I call it the Points Contribution Analysis - it requires me to analyze all moves (and for that I need to keep reviewing every single play) - it basically attempts to analyse Each player's contribution per points gain/loss/missed/saved. Say Harden score a FG on a fast break with an assist by Lin off a defensive rebound from Asik - so essentially the 2 points that the team gained is actually contributed by Harden-Lin-Asik from PCA perspective they should really share the points (i.e. 2/3 real points or RP each). And for example when the opponent attack and it's a 3 on 1 fast break and scores - the loss point (LP) should be attributed to not only the single defender but also the other 2 defenders whom have failed to get back in time (i.e. 2/3 LP each). Conversely, if the sole defender was able to prevent a score - he would have garnered 2 SP (saved points). Missed points (MP) are attributed not to the only person who have missed the ball but should also include others whom may have contributed to the miss (trying to work out whom should be included - at the moment, I think those who manage to get themselve free from their defenders but did not get a pass do not need to be included in the MP calculation)

So a combination of RP, LP, SP and MP should give us a PCA view of each player's contribution to the team. I am also thinking about a weighted split for the points i.e. for shooting the actual shooter may get more than the other contributors, for LP defenders who lose their men attributed more, shooters also attributed to more MP, actual blocker getting more SP etc.

But it would take quite an effort to do the PCA (I need the proper feeds to review the videos and lots of time which I don't really have... sigh).

Let's try this current method and see how it goes

 

Thanks for the 82games link will check it out and see whether we can make use of it.

Not sure about the combined ppm that you have in mind at the moment but I still think it is best to separate offensive measures apart from defensive measures I.e. reality is that in nba we have players who are good in both or good only in one (if you're bad in both one wonders why they are still in nba).

There are other help activities that I really love to see being tracked eg helping to pull defender away for an open lane, a good screen, denial of passing opportunities leading to shotclock v or a steal etc.

 

Thanks for the link to the article - interesting read - but too complicated for us armchair analysts to implement.... I am just waiting for a few more teams to hit their 16 games mark to release my next batch of analytics - and we can see if the measures could still be useful to gauge the performance of our guys.....