What does FiveThirtyEight’s CARMELO projection system think of the Utah Jazz? And can the Wins Above Replacement projection it produces tell us anything about how productive individual Jazz players will be in 2018-19?
Because the system doesn’t spit out predictions for raw numbers like points, rebounds and assists, I took a different path to try to find those things. Under the WAR projections for any player you search in the system, you’ll find that player’s “10 most comparable players,” which FiveThirtyEight explained in June:
The basics of the system are largely similar to previous years, with the backbone of CARMELO remaining an algorithm that compares current players to past ones who had statistically similar profile through the same age. For instance, Utah Jazz phenom Donovan Mitchell is similar to players such as Gilbert Arenas, Ray Allen, Stephen Curry, Ben Gordon, Victor Oladipo and O.J. Mayo through this early point in their respective careers. Some of those players (Allen, Curry) became superstars, while others (Gordon, Mayo) didn’t really pan out. The combination of those good and not-so-good outcomes gives us a probabilistic forecast for the rest of Mitchell’s career.
To project raw production, then, I took the following steps:
- Looked up the season after the one shown in the system for all 10 comparable players;
- Totaled points, rebounds, assists, steals, blocks and minutes for each season;
- Multiplied each total by FiveThirtyEight’s similarity scores;
- Calculated per-36-minute averages for the five major categories, to give us a nice, level playing field.
Following are the results for Utah’s projected starters (the bench can be found here), along with some comments on which numbers I think may be way off.
2017-18 Numbers: 15 PTS, 11.9 REB, 2.6 BLK, 1.6 AST, 0.9 STL per 36 minutes, 62.2 FG%, 68.2 FT%
2018-19 Projection: 12 PTS, 10.7 REB, 2.1 BLK, 1.8 AST, 0.9 STL per 36 minutes, 53 FG%, 63.6 FT%
We’re at least in the right neighborhood here, but it’s hard to see Rudy Gobert taking this big a step back across the board. The biggest miss is probably on field goal percentage, where the system may just have a hard time accounting for a modern center who puts up a huge portion of his offense in the form of dunks.
The projection here isn’t egregious, but I see Gobert finishing 2018-19 closer to last season’s numbers than the ones posited here.
2017-18 Numbers: 15.8 PTS, 9.2 REB, 1.7 AST, 1.4 BLK, 0.9 STL per 36 minutes, 56.3 FG%, 22.2 3P%, 65.1 FT%
2018-19 Projection: 13.2 PTS, 9.8 REB, 1.5 AST, 1.3 BLK, 0.8 STL per 36 minutes, 50.6 FG%, 22.6 3P%, 72.7 FT%
Derrick Favors’ comparisons are sort all over the place, but the numbers wound up being pretty believable. Like Gobert, this forecast expects a step back in scoring for Favors. That may well happen, but I think it would have more to do with a drop in usage than a drop in efficiency.
Donovan Mitchell could see an uptick in shots, and Dante Exum might finally break out. So, that could mean fewer shots for Favors, but I expect the ones he takes to be open and generally around the rim.
If he’s added a legit three-point shot, though...
2017-18 Numbers: 13.1 PTS, 5.5 AST, 4.8 REB, 1.3 STL, 0.3 BLK per 36 minutes, 46.7 FG%, 44 3P%, 79.5 FT%
2018-19 Projection: 12.5 PTS, 3.7 AST, 5.4 REB, 1.3 STL, 0.6 BLK per 36 minutes, 45.7 FG%, 37.8 3P%, 80.3 FT%
Dropoffs this steep in assists and three-point percentage for Joe Ingles would be pretty stunning. Sure, he’s already in his 30s, but he has the kind of game that should age well.
I suppose a decline in playmaking is conceivable with Mitchell taking on a bigger role there and Exum having a full season, but FiveThirtyEight’s system doesn’t account for those things.
This is another one I see ending up closer to the 2017-18 numbers than the projection.
2017-18 Numbers: 16.1 PTS, 6.5 AST, 5.6 REB, 1.9 STL, 0.2 BLK per 36 minutes, 41.8 FG%, 35.2 3P%, 86.6 FT%
2018-19 Projection: 16.6 PTS, 7.4 AST, 4 REB, 1.7 STL, 0.3 BLK per 36 minutes, 44.9 FG%, 34.2 3P%, 78.1 FT%
Other than the decent uptick in assists and field goal percentage, this one looks pretty close to last season. And that wouldn’t be much of a surprise.
Rubio was unleashed as a scorer in a way he’d never been before last season. He averaged 15 points after the All-Star break and logged four of his five career 30-point games with Utah.
2017-18 Numbers: 22.1 PTS, 4 REB, 4 AST, 1.6 STL, 0.4 BLK per 36 minutes, 43.7 FG%, 34 3P%, 80.5 FT%
2018-19 Projection: 19.9 PTS, 4.6 REB, 3.7 AST, 1.4 STL, 0.5 BLK per 36 minutes, 43.8 FG%, 36.6 3P%, 80.3 FT%
As is the case with a lot of these projections, it appears to be pretty conservative with the estimates on Mitchell.
If I had to guess, I’d say his numbers look closer to what the projection would’ve said if I’d only used his top five comps: 20.6 points, 4.3 rebounds, 4.3 assists, 1.4 steals and 0.4 blocks per 36 minutes, with a 44.5 field goal percentage, a 37.9 three-point percentage and a free throw percentage of 84.1.
Even then, I don’t think that quite captures what Mitchell will do as a scorer next season. It may sound like a lot, but I see Mitchell somewhere around 25, five and five per 36 minutes.