Half-Season Lineups' Stats – What worked for us, and what didn't

It's been exactly half a season, and that's a good time for another stat review.

Since TTNN made such a great thread for individual players' stats, I thought I'd focus on lineups.

Selection Problems (Sample Size)

Injuries and coaching decisions (minute distribution) have made the lineups' stat analysis very problematic in terms of sample size…

Some examples to show just how problematic:

* Let's filter by number of games, shall we?

Spoiler

You may think it would be a good idea to include only lineups that were used in, say, half of the games this half-season (20 games of 41). Good luck trying – NO ROCKETS LINEUP was used in more than 18 games!!
Even if you draw the line at a 1/3 of the games (lineups used in 14 games or more), you will be left with only 3 lineups, featuring only our current starters + Lin & Casspi…
In order to have a nice set of 10 different lineups you need to include lineups that were only used in 7 games!

* OK, shall we filter by overall minutes?

Spoiler

So far we played a total of 1,936 minutes, and I first thought of setting the bar at about 15% or 20% of team's overall playing time (only including lineups that were used for 290 or 387 minutes, respectively).
Well, guess what – NO ROCKETS LINEUP was used that often!
If you draw the line at 10% of PT (183 minutes) you are left with 2 lineups (Starters with Bev, Starters with Lin).
For a nice set of 10 different lineups you need to go as low as 30 minutes overall! (Less than 2% of the time!).

* Shall we perhaps select the most frequent lineups in a way that includes all of our main rotation?

Spoiler

Also not easy!
Look at Garcia, for instance – Even though he played in 40 games (!) and got his nice share of minutes (20.3 a game on average) – Still, you will have NO lineup featuring Cisco that played more than 36 minutes overall!
(Same with AB – He played 33 games and had 17.5 minutes a game on average, but none of his lineups played more than 33 minutes combined!).

Introducing the lineups for analysis

At the end, to get a meaningful set of different lineups, I chose to simply select the 10 most frequent lineups in terms of overall use (total minutes), even if they were only used in very few games, and I marked them as lineups A to J,
Please bear in mind that the further you go down in the alphabet (certainly from lineup G and on) – the less reliable the stats are (SAMPLE SIZE).

Spoiler

10 Most Frequent Lineups for Analysis (A-J)

I will also display the stats for the next set of lineups, down to #26 (sample size getting as small as 19 minutes overall) separately – not because they're performance has any real significance, but only for the sake of further experimenting: If any lineup, that was used very little, stands out as surprisingly promising – we may want to try it more and expand the sample to a more reliable size.

* Sadly and surprisingly, even this 2nd set of not-as-frequent lineups still barely features D-Mo (He's in lineup O only. until last game he was missing entirely) it totally overlooks Greg Smith and Canaan.

Spoiler

16 Next-in-Frequency Lineups (K-Z)

Mark the letters for your lineups of interest!
First actual stats will follow shortly...

(all stats from http://stats.nba.com/)

 

[Alert! I noticed that I swaped the descriptions between lineups F & G! Sorry...]

OK then!

First let's look at some basic scoring raw data tables.

Here are the scoring figures for the 10 most frequent lineups -
* Shots attempted and made from all ranges
* Total points scored by lineup, and points scored per 48 minutes
* Overall +/- achieved by lineup, and +/- per 48 minutes

Spoiler

Some notes:

1. Our starers* (A) seem to do very well, relatively, in scoring percentage.
They are ranked second best in FG% and in 3PT%.
The starters are also quite high (3rd) in overall scoring per 48 minutes.

2. Our worse shooting lineup by far is I (the TT+Bev).
Their awfull 3PT% is probably also skewed by CP's shooting,
(as you may remember - in the TT times he was still struggling from long range)
3. Two of our best +/- lineups (H, ranked 1st & C, ranked 3rd), that did better than the starters in this stat, are lineups with a stretch-4 (Omri Casspi in both cases).
4. Not surprisingly, two of our worst main lineups in terms of +/- are the TT lineups.
5. In the Lin vs Bev debate, the +/- stats show an advantage to Bev's lineup (though both are positive)

 

... and here's the same table for the 16 less frequent lineups:

Spoiler

0. ALL stats here should be taen with many grains of salt! Samplesize is problematic! Still:

1. Note that once again - all the best +/- lineups are stretch-4 lineups!
(W with Casspi is +78.3 points, U with Casspi is +50.5 points, & Z with Parsons is +48.0 points)
These 3 lineups are our best scoring lineups by far, in a big margin, both in stellar shooting %
and in overall scoring per 48 minutes (156.6 points, 144.0 points & 144.0 points respectively).
BTW - In overall scoring they are followed by the TJ + DMo lineup (130.9 points).

2. Unfortunately, our 3 worst +/- lineups are also built with a stretch-4 (Though the negative figures are much smaller)
In the two lowest lineups, K (-25.7) & Y (-17.7), it is Parsons at the PF, and in lineup Q (-11.4) it is Casspi,
so it looks like CP should stick to the superb SF he is, and leave the PF position (in small-ball) to Omri.

3. Not surprisingly (or perhaps it is ) - the lineup with Brooks, Garcia, Parsons & Casspi (lineup U)
was simply exquisite from 3PT - Making 10 of their 12 attempts!! (83%)
But note that this lineup also has the worst FT%! (I guess it was Asik who went to the line...)

I'm just glancing at the data for the first time.
Please add your thoughts about is as well.

Next - some advanced lineup stats, counted per 100 possessions

 

Here are the Offensive Ratings, Defensive Ratings and Net Ratings
for our 10 most popular lineups:

Spoiler

With no comments for now - I'm trying to watch tonight's game...

 

So here are the Offensive Ratings, Defensive Ratings and Net Ratings for our 10 most popular lineups:



The findings:

OffRtg

1. Surprise! Our best offensive lineup doesn't have Harden in it! Lineup H, featuring Brooks, Garcia, Parsons, Casspi and Howard, has scored 122.5 points per 100 possessions!!
--[Alert! This lineup's stats may be less significant due to smaller sample size]
2. Our starting lineup (A, the one with Beverly) is our 2nd best offensive lineup, but it's quite far behind, with 112.9 points per 100 possessions.
3. Our worst offensive frequent lineup is J, with Beverley, Harden, Garcia, Jones and Howard (Before Jones's beasting last night...), with only 82.6 points per 100 possessions.
--[Alert! This lineup's stats may be less significant due to smaller sample size]
4. The 2nd and 3rd worst offensive lineups are the 2 TT lineups, E & I - one featuring Lin, with Harden, Parsons, Asik and Howard, and the other featuring Beverly with the same 4. Those 2 lineups are practically tied with 88.3 and 88.6 points per 100 possessions, respectively.

DefRtg

1. Surprise again! Our best defensive lineup DOES have Harden in it! Lineup F, featuring Lin, Harden, Brewer, Jones and Howard, has allowed only 92.2 points per 100 possessions!
2. The only 2 lineups featuring Casspi - C, the one with Lin, Harden, Parsons and Howard, and H, with AB, Garcia, Parsons and Howard, are tied as 2nd best defensive lineup, allowing only 95.0 and 95.2 points per 100 possessions. (Remember that H is #1 in offense as well!)
--[Alert! Stats of lineup H may be less significant due to smaller sample size]
3. Our worst defensive lineup is surprising as well, since it's a Twin-Tower lineup! (with Beverly) - lineup I, which as already mentioned as one of the worst offensive lineups. This lineup allowed 112.7 points per 100 possessions!
--[Alert! This lineup's stats may be less significant due to smaller sample size]
4. The 2nd worst defensive lineup is J, also mentioned as a weak offensive crue. They allowed 109.0 points per 100 possessions.
(Interesting enough, these 2 worst D-lineups have P-Bev in them, though he is considered a good defensive force)
--[Alert! This lineup's stats may be less significant due to smaller sample size]

NetRtg

1. This follows naturally from above -
The best NetRtg lineup is H. This lineup is very stron in both D and O, and it creates a difference of 27.3 points per 100 possessions!!!
2. The worst of all lineups are the aforementioned miserable I & J who trail by 26. and 24.2 points per 100 possesions!!