Unit Analysis: Curry, Ellis, Wright, Lee, Player X

In lieu of assembling the discrete stat detail, I compared the EZPM100 and its 3 components for this core 4 with Rad vs one of the other 3 non-Rad alternatives. Here is the with Rad - without results (possessi0n weighted, if I did it right): WARP EZPM O D REB Monta Ellis -1.0 5.1 5.5 0.1 -0.5 Dorell Wright -0.1 7.4 3.5 4.6 -0.7 Stephen Curry -0.2 3.5 5.3 -2.3 0.4 David Lee -0.9 -3.9 0.2 0.5 -4.6 Rad as 5th guy -0.2 2.3 1.5 1.3 -0.6 EZPM100 improves at 4 of 5 spots, O100 at 5 of 5, D100 at 4 of 5, Reb100 at only 1 spot. That is still quite good in favor of Rad as the 5th guy for this core 4, especially since 3 of the 4 rebound declines were small. (The WARP decline puzzles me a bit.)

Confidence intervals could be calculated to give a better estimate of the bounds for good results in the future.

Your initial graphic showed that Rad was by far the best piece to match with the core 4. I later showed that Rad was in the best quad and other strong performers. If they let him go elsewhere, I will just shake my head. Are they going for better than a proven double digit positive per 48 minute winner? The odds of getting better than that are slim. Maybe the odds of repeating the performance fully at bigger minutes aren't that strong either but the odds of getting at least moderately positive results would probably be at least pretty good. Are these Rad quads and 5 man lineups just fit for specific rare circumstances? That could be, would have to look at the opponent file and tape further. But you'd have to prove that argument rather than just assume it.

Of the 5 best on +/- per minute, Curry was 4, Ellis 3. Wright 4, Williams 1, Radmanovic4, Lee 4, Udoh 0, everyone else 0. The best 5 would make Curry Ellis Wright Radmanovic Lee. How did that lineup do? +13 per 48 on raw +/-, +12 on Adjusted +/- ! Used just 156 minutes for the season. What the "l" do they want?

Curry, Ellis, Wright, Lee is actually the least positive of these 10 quads at just +2 per 48 minutes. That is a pretty powerful argument against making it the core 4 going forward. Sure it had more 3 times as much trial and some of the other quads might have benefited from small sample variance or weaker average opponent quality... but all of them were better. Were all of them advantaged by chance and / or circumstance? I doubt it.

Yeah for all the lineups that get used situationally on purpose or at random there are often lineups that don't get use that probably should have at least been given a decent look or a bigger look. Curry Wright Radman Lee just happened to have the best cumulative +/- for a quad for the season and by far the best +/- per minute (at +19.8 per 48) but hey why give that more than 192 minutes in a season? There were 9 other quads with 150+ minutes that were positive but they must not have been used enough either. (The quads are of course overlapping and not completely unique so the total minute count is probably not as high as it might seem.) Of these 10 quads, Curry was in 8, Ellis 7. Wright 8, Williams 1, Radmanovic4, Lee 8, Udoh 4, everyone else 0.

The 6 best lineups on Adjusted +/- used over 40 minutes for the season had about 4.3 guys from the core 4 and Rad with the other guy being mostly Williams. The question would be whether these lineups that worked well were in really better than the lineups pulled together with 4-5 guys from the same 6 that did poorly.

Substitute in Wright for either Curry or Ellis and that would be worth a look too. Or take Lee out and put Rad in. Or maybe other stuff.

Ellis and Wright benefit the most from being with Rad. He is Curry's best pick for the 5th guy too. The only one who is worst with Rad is Lee but he is bad with 3 of the 4 5th guys so it may be more a general issue with him in this core 4 and system than being with Rad. I am surprised that a Curry Ellis Radmanovic Lee Udoh lineup only got 5 minutes all season. Based on logic and these numbers, it should have at least gotten a look. But then again I am not surprised.

Biedrins-Lee is getting beat by the best bigs because this lineup is not performing well on offense against them. Lee-Udoh is getting beat equally on 3 fronts, at least by the best opponents. Radmanovic-Lee at least does well on offense. Admundson-Lee is god on rebounding as expected but that is it. If they don't keep Radmanovic, Udoh is a pretty clear pick for the 5th spot over at least these 2 alternative but I'd probably keep Rad for at least situational use. There would be other options if one were willing to change from the core 4.

This set of data is a useful display and allows some further analysis. The average position of the best 10 opponents for Biedrins is 3.4 (a bit bigger than SF), Udoh 3.8 and Radmanovic 3.0 are all higher in the position order than the worst ten opponents and by an average of 1.2 positions. The position is close to equal between best and worst 1o for Admundson, with the worst 10 actually being a little higher on the position order. For Biedrins the best opponents are far better on defense. The worst are very bad on both offense and defense. For Udoh the best are equally good on offense, defense and rebounding. The worst are worse on defense. For Radmanovic the best opponents are best on offense and rebounding. The worst are worst on defense. For Admundson the best opponents are very good on offense and defense, a bit more on offense. The worst are terrible on offense. Radmanovic's lineup with the core 4 is a very powerful performer in modest use. I might try to retain him if the core 4 stays the same, especially if the style of play stays similar with the core 4. He has good traditional APM for this season as well. I do think he is a good example of a guy who is easier to dislike / dismiss by just boxscore stats than enhanced boxscore stats with lineup context and Adjusted +/-. His type of stretch 4 is often associated with positive team offensive impact.