CARMELO, not Carmelo
538 has released its annual projections, based on their “CARMELO “ forecasting system, for NBA players. Such projections are never perfect, but as far as I can see they are objective. That makes them extremely useful; of course if you ask me, I’m going to tell you that Joel Embiid will be easily the best center in the NBA next year. That’s partly because Joel is awesome, and partly because as a huge Sixer and Joel fan, I am heavily biased toward him. If CARMELO says he’ll be the #1 center — or #2, or #5 — then that may be incorrect, but it isn’t a consequence of the 538 computer’s fandom.
Before we look at the numbers, let’s note a couple things:
- CARMELO projects per-minute production; the figures below don’t give extra credit for players expected to see more court time. 538 also has value projections that combine per-minute production with minutes played, but here I focus on the per-minute.
- The units of these projections are points per 48 minutes, just like BPM and RPM. Basically CARMELO is an attempt to predict a mix of 2/3 RPM and 1/3 BPM. If a good, full-time player is replaced by an average NBA player, the team would be expected to lose around 3/4 of his CARMELO figure per game, since starters play around 36 MPG and 36 is 3/4 of 48. If you lose a +3 player and have to fill in with a -1, that’s a loss of 4 points per 48, which is 3 points per 36. One point per game tends to equate to around 3 extra wins/losses over the course of a full season, so the loss of a +3 player might cost his team 3*3 = 9 wins, perhaps a bit more since the promotion of the -1 backup to starter status would also weaken the bench. For a sense of scale, here are some 2018-19 CARMELO forecasts, rounded to the nearest whole number:
The Sixers and CARMELO
OK, here are the offensive, defensive, and total CARMELO forecasts for the Sixers for the upcoming season, sorted from top to bottom:
(Editor’s note: Upon converting the CARMELO data to the table above, Robert Covington’s total CARMELO rating was previously incorrectly presented as 4.1. It has been corrected in the table to reflect Covington’s total CARMELO rating, which is 2.8.)
Sadly, CARMELO does not have forecasts for Furkan Korkmaz or Jonah Bolden. I’m not really sure why, since they do include the following forecast: Andzejs Pasecniks -0.3 (OFF), +0.5 (DEF), +0.2 (TOTAL).
Yes, that’s right, CARMELO thinks the much-maligned AP would be a decent bench contributor if he were to come over this year. Not sure why the model likes him so much more than a casual look at his traditional stats would suggest; if I have insight to share on that at some future point, I’ll write a piece on it!
Before we evaluate the team, let’s do a quick comment on the method. Regular readers will note that this list ranks our players in pretty much exactly the order I place them in when I write about the team, an order that is quite different than that employed by most fans and pundits. I refer in particular to the high positions of Robert Covington and Amir Johnson, especially when compared to JJ and Dario, and also to the low position of Wilson Chandler. I recognize that many readers think that only irrational attachment or a brain virus can explain my making statements such as “Robert Covington is a top-30 player in the NBA.” And yet here is CARMELO, which is a computer program so seemingly immune to both bromances and viruses, and it ranks our Cov top 25 in the league! I know people suspect I love Amir and his old-man game so much only because of some connection between me and Amir, maybe I’m old too! And yet here is the CARMELO computer program, only a few years old itself, ranking Amir as (per minute) superior to Saric and Redick! In truth I think all this is simple to explain. If you don’t like the ranking above, ask yourself if you like this ordering of the key non-Embiid players better:
This list looks about right to an awful lot of people as a ranking of current hoops prowess. But what it actually is, is a list of the players ranked as 538 does, except IGNORING DEFENSE! All the pundits who think Chandler is better than Amir and JJ/Dario are better than Cov are just putting zero weight on the defensive side of the ball.
In terms of individual players I’d say:
- All the rankings feel a little low, and I think I understand why. As fans we basically pay attention only to players who are actually doing something out there. But a certain percentage of players just fail miserably in a given year, often but not always as a result of injury. Take JJ Redick at -0.9. Last year JJ was between 0 and 1, and he probably will be again. A drop to -0.2 or something is reasonably likely, but I’d happily bet at even money that he’ll be over -0.9, and I’m guessing you would, too! What CARMELO is recognizing is the skewed distribution here. There’s a decent chance, say a 20-30% chance, that JJ at age 34 just falls off a cliff and becomes a -2.5 or -3 player, i.e. that he falls to the level guys like Marco Belinelli and Jamal Crawford have played at in recent years. Whereas there’s basically no chance he’s a +2 or +3 player this season. So the mean may be -.9 even if the median is more like 0.
- Similarly, Dario’s number looks a little light at +0.4; 538 is projecting him to be no better than last year. Chances are Dario will improve, but there’s probably more room on the downside for him than the upside. Early last season he was playing at a -3 level; he probably won’t go back to that, but the computer is probably thinking he has some chance of returning to -3, but almost no chance of leaping to +3.8 (All-NBA-type player), which would be an equal move in the opposite direction. Thus his median is probably close to +1, but the mean is lower.
- The machine is projecting enormous mean reversion for Covington, whose RPM was well over 5 last season, and I think that is reasonable. Again, at the +2.8 they project, he’s the #24 player, a hair behind Jayson Tatum and a bit ahead of Paul George. If he’s that good, maybe he’ll finally get some respect (but probably not!). If he is up over +4 again in RPM, RAPM and the like, which would put him near the top 10 for a second year in a row, you can be sure I’ll be posting about it!
- What about Joel and Ben? You can see why Sam Hinkie and Brett Brown are obsessed with stars. As fans we think a lot about guys like JJ and Dario, but the gap between them and scrubeenies is dwarfed by what the superstars provide. As you can see 538 is very bullish on Ben. A bad week against Boston has caused many to underappreciate the astonishing things Ben accomplished as a rookie. Sometimes it takes a cold, logical computer to see what’s right in front of our face: Ben Simmons is absolutely amazing, and will probably be a lot more amazing in 2018-19. The model isn’t quite as enthusiastic about Joel; it ranks him 17th in the league for next year (vs. 10 for Ben). We’ll see if Drew Hanlen’s work will make that projection too conservative.
- As for the young guys, no big surprises. Rookies like Zhaire, Shake and Shamet rarely contribute much, indeed Zhaire’s -1.6 is, unless I missed one, the highest CARMELO projection among this year’s first-round non-centers. Which is extra-impressive given Zhaire’s extreme youth. I continue to see good things ahead for Zhaire! But we shouldn’t expect more than OK backup play from him as a rookie.
- And then there’s Markelle. The computer doesn’t know about the injury or the yips or Drew Hanlen or three-a-days or any of that. It doesn’t separate out Markelle’s awful few games before being held out for most of the season. The model just sees that, overall, he didn’t hit shots, so understandably the model predicts he won’t be all that good this year. Since the model has so little to go on, it’s predictions are even more uncertain than usual. We’ll just have to see what happens. But, I said I’d do an objective analysis, and so there it is: the CARMELO model does not have high expectations for Markelle this season.
Looking at the offseason, well, it started out amazing, with the addition of Zhaire and a valuable future pick. Since then it has not been overly impressive. We haven’t seen much Landry Shamet yet, and I certainly hope for the best from him, but personally I’d rather have seen that pick packaged with Bayless to obtain Kyle Korver than used on a guy who would likely have been there in the second round. We spent $24M to obtain an OK player in JJ Redick and a not-really-OK player in Wilson Chandler. $24M is a tremendous amount of money, and I do believe we could have done a lot more with it. This isn’t the place to go into all the alternatives, so for now suffice to say that the -0.9 and -1.6 projected for these two $12M players is a far cry from the +1.5 the equally-paid Tyreke Evans is projected to deliver. I liked the Bjelica signing a lot, but it didn’t stick; on the plus side the Muscala deal was a nice recovery, as we added a quality player and cleared a needed roster spot. Overall a solid offseason, but something of a letdown after the tremendous Zhaire trade.
Sixers vs. Celtics: The virtual competition
How about the team as a whole next year? Actually, it’s tremendous — I sort of “buried the lede,” as the journalists say. Let’s keep the title’s promise by comparing Philly and Boston. Here is the Sixers’ starting lineup again, with total CARMELO forecast:
- Ben Simmons: +4.8
- Joel Embiid: +4.1
- Robert Covington: +2.8
- Dario Saric: +0.4
- JJ Redick: -0.9
- Total: +11.2
That is a phenomenal starting group! For comparison, let’s do the vaunted Boston 2018-19 starters:
- Al Horford: +2.4
- Jayson Tatum: +2.9
- Gordon Hayward: +1.7
- Jaylen Brown: +0.6
- Kyrie Irving: +2.6
- Total: +10.2
Angry Celtic fans, please direct your rage not to me but to:
Of course this doesn’t mean the Sixers will outplay the Celts this season, there’s bench play, coaching, fit/chemistry, etc. etc. I’ll discuss the Sixers’ potential bench issues below, without doing an equally deep dive on the C’s, but suffice to say that Boston’s bench, which includes fine players like Marcus Smart, Terry Rozier, and Aron Baynes, is projected by 538 to be far superior. The point here was not to bury the C’s, but to praise them — we all know the Celtics have a terrific starting lineup for next year, and yet the objective 538 system thinks we should expect more from the Philly starters than from the Bostons. Celtics fans, if you’ve written a diatribe about how scoring points and preventing the other team from scoring and all that fancy nerd math stuff doesn’t matter because the game is really determined by fit and chemistry and character and tradition and Leprechauns, please re-read this paragraph before posting! I’m not saying that stuff doesn’t matter, it’s just not being analyzed here because 538 doesn’t have a CARMELO model for chemistry and fit, and I’m not capable of analyzing those things objectively, and this piece is an objective analysis.
Moving on... as a further comparison, the 538 projection for Golden State’s group of Steph-Durant-Klay-Dray-Looney is a bit over +15. If the Sixers could have pulled off Dario, Chandler and picks for Kawhi, we’d be at that same level, but it was not to be. Did I mention that San Antonio passed on our value-laden package to trade Kawhi for a negative-value asset in the overpaid DeRozan, and that I’m working through the stages of grief on that? Almost to acceptance! Anyway... if DeMarcus Cousins comes back strong the Golden State figure gets close to +20; yikes!
As discussed above, according to 538, the Sixers are starting a below-average-in-expectation player in JJ Redick. The model assumes aging and JJ is pretty old; he was a little above average last year and we can hope he doesn’t decline, but even if we get that lucky he just is not all that good, because defense counts too. In terms of on-court contribution, re-signing him was probably a mistake, but perhaps he’ll add enough through continuity, teaching the young guys, and all that to justify his compensation. Of course the fantasy is that Markelle steps up and becomes a key contributor. Fultz is still very young and, frankly, if his jump shot fully returns and he plays at a +0 level, that would be tremendous, a strong sign that he is on his way to being a frequent All-Star. If you look at other young PGs who became major stars, guys like Wall and Kyrie and Harden and Westbrook and Dame, you’ll see it generally took 2-3 years before they were meaningfully above average as contributors. I am bullish on Fultz long-run, and bullish on him returning to form this year. But fairness compels me to say that even had he never experienced his setbacks, even if he were right on path for the Hall of Fame, a +0 or +1 year would be the most we should reasonably expect for 2018-19.
What about the bench?
The Sixers’ bench, despite the offseason overhaul, is still not impressive. Well, the backup bigs are good, with Amir being at the quality level of an above-average starter and Muscala a low-level starter; plus bringing very different skill sets so as to match up with different types of opponents; that’s a nice group. But suppose our dreams come true and Markelle becomes a quality starter, relegating JJ to the bench. Then our key backups are:
- Amir Johnson: +1.2
- Mike Muscala: -0.2
- Wilson Chandler: -1.6
- JJ Redick: -0.9
- T.J. McConnell: -1.3
An average starter is around +1, an average backup is around -1. So as you can see, we have below-average backups, in the aggregate, at the 1-2-3 spots even in the Markelle-succeeds scenario. This isn’t a disaster for several reasons:
- Most teams don’t play 10 deep, so when we say an “average backup” is -1, that average is skewed toward the 6-7-8 players on a team, not 9-10
- We don’t have projections for Furkan or Bolden; that probably wouldn’t change this but it might!
- We’re not that far below average in this scenario
- We have several guys competing at each of these spots and we get to play whichever of them turns out the best; out of Wilson Chandler, Zhaire Smith, and Furkan Korkmaz the best one will probably be better than -1.6
Still, given that our starters were, by this method, the best in the East prior to the Kawhi trade, we are legit Finals contenders, and so it’s a shame for us to have a bench that is just ordinary. That’s not a problem that needs to be solved right away; unless a terrific player becomes available the team is likely best served to start the season with the guys they have and take the Fall to see how good their young players are and which combinations are working well together. At that point they can see what trade-deadline or buyout deals would most benefit the team. But the short version is this: we could get lucky and have a strong bench with the group bon the current roster. That would require JJ to be as good as last year, and Markelle to step up as both a player and a shooter, or, alternatively, one of the other players rated below -1 (probably Zhaire, or perhaps Chandler) to play at a positive level by playoff time. Plus everyone being healthy. Frankly the more likely scenario is that we need one more good player. Which is why the Wilson Chandler deal killed me; we had $12M to spend to get a good player, and instead we got a bad player and a second-round pick. If we stunk that might be OK, but as 538 makes clear, it’s not just homerism that makes me believe we have a real shot to make the Finals this year. It’ll be a shame if the Chandler deal leaves us short the one solid bench player we need to get over the hump.
But I don’t want to end on a negative note. There is a strong, objective case that the Sixers’ starters are the best in the East. A Toronto group of Lowry-Kawhi-Anunoby-Ibaka-Valanciunas would, like Boston’s lineup, have a 538 projection below Philly’s; I think that Toronto may be able to get a number a bit above Philly’s 11.2 if they go double-point-guard by starting VanVleet and Lowry together in the backcourt and benching Ibaka; whether that would work in real basketball I leave to your judgment. In any case the point is that despite the failure, so far, of the “star hunt” Philly’s lineup matches up with the very best in the East, according to a model with no dog in the fight. And that’s right now, with Markelle Fultz not among the starters and with the Sixers having $42M to spend next offseason, something our Eastern rivals can’t match. In a future post I’ll take a look at the 538 long-term value estimates, which are quite interesting. But we’re past the point where we are forced to focus on the long term because the present is disappointing; the Sixers should be headed for a big year in 2018-19!