Yeah, that's fun. IT is projected to improve both offensively and defensively, nearly doubling his total +/-, and has a worth of $28 mil! Love it.
Not sure about Aldridge, though. Took him some time to adjust in SA, but he was playing like a bona-fide all-star in the playoffs there. Monster numbers, seemed unstoppable. Derozan has to be a "borderline all-star," too, if Horford is -- he's a two-time all-star now in his prime.
Smart as a "future all star" -- man, that's aggressive. CARMELO loves him.
Are we looking at the fine print for IT? if so he's projected to get worse slightly worse offensively and a decent amount worse defensively.
As for Derozan his ownly saving grace is how much he gets to the line.
You need to look very carefully at the 'fine print'.
First off, the model takes things like draft slot and height information as input and Isaiah naturally takes a hit, fair or not, in it's projections.
Second, and more important, Isaiah posted a 6.8 WARP in 2014 as a full-time starter, then that dropped to 3.6 in 2015 as he got stuck in Phoenix' 3-PG limbo and then came off the bench for us after the trade. Based on those two seasons as it's most recent input, CARMELO projected Isaiah to post a 4.0 WARP for last season. Well, obviously, he blew that out of the water and posted a 7.8 WARP.
So when it projects now that he will post a 4.9 WARP for next season, keep in mind that that number is derived with the 2015 data still as part of it's inputs. I.E., it is dragging down the 'regression line' artificially. UNLESS, of course, you think that Isaiah will no longer continue to be a full-time starter, that projection makes no sense.
Now, to be fair, their 'confidence interval' for Isaiah spans from anywhere as low as about 2 WARP all the way to just under 10 WARP. They are fairly confident he'll end up somewhere in that span! LOL!
A more 'confident' estimate is probably to eyeball the 6.8 and the 7.8 and guesstimate that, assuming he'll continue to be a starter, he'll post another season with a WARP somewhere north of 7.0, which is a lot better than "good starter".
Similarly, because the prior two years of Jae Crowder's data were as a bench player, it predicted 1.9 WARP for him last year, and he blew past that with a 6.1. If Jae continues to be a starter, he almost certainly will again blow past the 4.5 that they are projecting for him this coming season.
If you examine CARMELO results for last year, it did best for players that have continued to get the same amount of playing time and usage both for the preceding few years and through the projection year. For players that missed time due to injury (like Noah, Irving) this last year, it over-estimated them tremendously. For players that have one or two anomalous injury seasons in their recent history (like Al Horford) or were under-used compared to their current role (like Isaiah and Jae) it under-estimated them tremendously.
As always with such projections, take them with a HUGE grain of salt. Look at the error bars for the projection. Look at the data it is using as inputs and make a judgement call as to whether the projection makes any sense.
For young players (rookies and second years) especially beware because it includes a lot of college data in the inputs that doesn't always map that well into the NBA. For example, one of the (several) reasons Smart has underperformed his CARMELO projections is because his FT frequency in college was wicked good. In the NBA it has been just 'a little better than average'. As the college data ages out of the weighted average, the projections should start to get better for Marcus.