With a break in the Nuggets schedule I have been roaming around on the new NBA Stats site recently. With the ability to search on 2,3,4 and 5 Man Lineups the NBA Stats site has the ability to give a large amount of lineup information back. The filters available can help to narrow the numbers down, but there is no way that I know of to cut and paste results, link queries together, or even to display the results in a graph form.
However, by typing the results myself directly into some other programs I was able to come up with some infographics to help share what I found after exploring and breaking down Denvers lineups with some lineup based statistics:
the Offensive Rating (OffRtg), the Defensive Rating (DefRtg), and the Net Rating (NetRtg)
Player On Court Ratings
Note: As these rating statistics reflect the success of the lineups that a player was in, individual ratings will be affected by who they were played with. On the other hand basketball is a team sport and they can provide a good starting point and may spark some questions that can be tested with further lineup combinations.
For context, the team ratings (all lineups) for Denver (after 74 games) are the following:
Off Rtg 107.4 DefRtg 102.1 NetRtg 5.2
I have only included the nine regular rotation players in the ratings charts below as the other Denver players only get irregular playing time (which leads to skewed ratings). Wilson Chandler had the lowest minutes for an included player at 841 mins, the next Nuggets player after him was Jordan Hamilton with only 379 total mins.
These are the individual ratings for the nine players in the regular rotation:
The players are organized by the NetRtg number (= the difference between the bars) which is displayed on the horizontal axis along with player surname.
Nice to see all the players here have a positive NetRtg. Lawsons ratings were probably the most surprising to me.
This graph suggest that a lineup of :
Unfortunately since this lineup has only played 10 mins together this season it is worthless to look at the NetRtg for this lineup (if you are wondering, the NetRtg after 10mins is a completely insane +81.2).
Player Off Court Ratings
The graph below displays how well Denver rates on Offense and Defense when a particular player is on the bench. In this graph the lower the NetRtg the more Denver is suffering while that player is on the bench. The order came out different than the above On Court graph, but the first five players (looking from left to right) are the same five players as above. The only lineups on either graph that achieved under 100 on DefRtg (at 97.9) are those playing while Lawson is on the bench.
As the Rating statistics are best used to look at lineups rather than individuals, I used NBA Stats to run multiple queries to see if I could find and then connect any lineup patterns in the data.
The following table summaries the limits I applied and results I analyzed
Summary comments on Lineup results
I was trying to focus my limits so I would get around 10-20 results for each lineup number as this would give me more chance of finding patterns.
5 Man Lineups Huge outlier was the standard starters lineup (consisting of Koufos, Faried, Gallinari, Iguodala, Lawson) with 797 mins together as a 5 Man Lineup. The 2nd place lineup has only 124 mins together on court.
4 Man Lineups Of the 12 results: 5 lineups each had over 800 mins, while the other 7 all under 400 mins each.
3 Man Lineups Of the 24 results: there are 10 lineups with over 900 mins while 11 of the lineups have less than 500 mins.
2 Man Lineups Because of the large number of combinations of 2 Man Lineups, I loosened the time filter and then restricted the results to only the bottom 24 NetRtgs. Even after loosening the time filter to >266 mins, only two of the 2 Man Line ups had a negative NetRtg.
I was quite surprised by the low number of negative NetRtgs at each level, but it gave me the idea to try and link the negative lineups together to look for patterns.
This is what I found.....
The lineup infographic works as follows:
The red box shows (one of the two) Denver 5 Man Lineups with a Negative NetRtg.
Below the red lineup are any of the other negative component lineups which form part of the 5 Man Lineup.
Above the green line are some component lineups which form part of the red lineup but have positive NetRtgs.
The lineups above the yellow line (and the yellow boxed players) are additions that the Ratings suggest will do well.
Lineup 1 Patterns
The red lineup in this graph is the ‘starters’(Faried, Gallinari, Iguodala, Lawson) plus McGee (instead of Koufos).
McGee has not started this year so this lineup hasn’t started any games, but whenever it played it did not work. By removing Javale the lineup goes from a 5 Man Lineup rated at -12.8 to a 4 Man Lineup with a +5.1 rating. If Koufos is added back for the full starters lineup (whiteboxed lineup) the 5 Man starters lineup is +6.5.
If we remove Faried instead of McGee, the 4 Man lineup is rating better than the original red 5 Man Lineup but is still -5.9 NetRtg.
Both of the negative 2 Man Lineups of Denver also form part of the red lineup with Faried and McGee at -2.6 and McGee and Lawson at -1.0.
The data suggests Javale replacing Koufos and playing with the ‘starters’ lineup at any time is a bad idea.
The data does suggest however that smallball lineups with the starting core but playing Faried at center (and having Miller or Brewer come into the lineup) might be worth trying a bit more as these lineups have produced good net ratings.
Patterns suggest more time for these lineups
Kenneth Faried, Danilo Gallinari, Andre Iguodala, Ty Lawson, Corey Brewer
Lineup 2 patterns
This red lineup involves the ‘bench’ plus Ty lineup: Chandler, Brewer, Miller, Lawson, and McGee.
This lineup goes from -11.4 to +3 if Lawson is removed from the lineup, and if McGee is removed also the 3 Man Lineup rating for Chandler, Brewer and Miller is +5.3.
The lineups above the yellow line indicate that it might be worth keeping McGee in the lineup and adding Gallinari (for a +13.2 NetRtg) or Iguodala (+8.2 NetRtg) in place of Lawson. The lineup with Gallo has a very high OffRtg of 108.6 while the AI lineup has a great DefRtg of 88.5. The minutes played by these lineups is pretty low, so these could change significantly with more court time, but seem like promising lineups.
Patterns suggest more time for these lineups
Wilson Chandler, Corey Brewer, Andre Miller, Danilo Gallinari, JaVale McGee
Wilson Chandler, Corey Brewer, Andre Miller, Andre Iguodala, Javale McGee
I have not delved deeper to try to find the root causes of these negative 'Red lineups', I may at some point, but for now this was more a post for thought, and I hope it may generate some thinking or discussion on lineups and court combinations among the readers of Denver Stiffs.
A note on lineup total times and methodology
Time is equivalent to sample size in lineup analysis (the more time a lineup plays together the more ‘samples’ of that lineup you have to judge it on). So the higher the minutes played by that line up the surer you can be about the results. Some lineup will not have played enough time together but after 70 games some patterns are emerging.
Because I do not know the exact amount of time a lineup must play together to give reliable results, I used the minutes played filter on NBA Stats for the lineups. I used this filter to try and find the sweet spot between getting rid of noise and removing good information.
For example: with no time filter, line ups that have played 1 minute+ together are returned. The ratings for these low time line ups come from too small a sample and so are meaningless. I wanted to get rid of this ‘noise’ so I could see any patterns in the other lineups more easily.
Example of results for all 5 Man Lineups (with minutes>47)