Surprised that people don't rate Kemba - he was still fairly good last season and I expect him to make the contract New York signed him on the best bargain deal in the NBA.
The guy is a defensive turnstile with a banged up knee and on the wrong side of 30. I was ok with the Celtics trading him for a bag of chips.
Defensive turnstile who posted a positive value in most adjusted plus-minus metrics this year and still pumped out fringe All-Star value on offence (especially after he got his sea legs back) - here are Kemba's numbers after his dismal first 10 games last season:
Box stats - 20.6/+1.6% PPG to rTS, 23.6%/11.1% assist rate to TO rate
Adjusted plus-minus - +2.6 OEPM (94th percentile), +0.1 DEPM (63rd percentile), +2.7 EPM (85th percentile), +3.5 ORAPTOR (27th in the league), +0.4 DRAPTOR (tied for 16th among point guards)
Before you start nitpicking the list of names in those metrics to dismiss them as stuff made up by nerds that are infinitely worse than the basic box score in measuring a player's value, there's definitely some noise to single-year APM (it only estimates a player's situational value in that given year, which can either go wrong because of wonky lineup data or poor situation/fit for the player in question), but the fact that Kemba produced such numbers in a situation that everyone and their grandmother on this forum screamed was far from ideal speaks volumes to his quality. And before the narratives of statistics lying about how good a player is start flooding in, the numbers jive with the eye test. Kemba was still a dangerous PnR ballhandler who strains defences with his ability to pull up for three on high volume and good efficiency (he shot 37.4% on 8.5 attempts after his first 10 games), drive to the basket for layups or floaters, counter defences running him off the line and sitting in the paint with a quality in-between game and find open teammates if defences collapse on him. On defence, he's active both as a man and team defender, makes good rotations (his ability to take charges is elite) and does a decent job at staying in front of his man as a point-of-attack defender. I completely understand scepticism about him if someone drafted him to start for their team, but how many backup PGs in this format are better than Kemba Walker as a basketball player? Good offence and decent defence, what more do you want for someone to run your offence for roughly 10-15 MPG?
I do in fact think most of those metrics are lists made up by nerds that don't really hold up. There are so many outliers that the use of those metrics is essentially meaningless. And, I think the "decent defence" claim strains credibility. He played below-average defense his first few months here. Then he hurt his knee, and he was downright bad. I would argue sub-IT, sub-Kyrie bad.
But, those observations aside, he's a nice player to have on a team regarding his leadership and for his overall good offensive play. But, what confidence can anybody have that he can play in back to back games, or be healthy for the playoffs? The Celts babied him all last season, and he was still unhealthy when it counted.
Yet it's still by far the most reliable metric we have for evaluating defence - the amount of outliers aren't that crazy when you start filtering out the low-minute guys and know what happened to the few high-minute ones (eg. nothing normal happened in Toronto last season).
And I think the 'below-average defence' claim is the one that truly strains credibility.
You might be able to get away with that if we hadn’t all watched him play 100 games here. If a metric says he’s better than below average, that metric is garbage.
Raptor tells us Jabari Parker was a better defender last year than Marcus Smart. Kemba and Smart were equals. Kemba is significantly better at defense than Timelord, Jaylen or Tatum.
Looking at another team many of us have watched closely, we learn that Pat Connaughton is a better defender than Jrue Holiday.
Boogie Cousins, a favorite of mine: best defensive player on the Clippers. Kawhi is 7th on the team. But, perhaps that is not surprising, as Boogie is better than Giannis, as well. Of course, a lot of players are better than Giannis, who ranks 93rd. Ben Simmons? 84th.
Does any of that make sense to you after watching games?
Or maybe it tells you that you're missing something in your analysis. Half of the examples you brought up were low minute players that the stat tells you to filter out (Parker, Boogie, a couple of the guys ahead of Kawhi on the Clippers, etc).
Most of the rest makes sense: anyone who followed the Celtics closely will have noted that Jaylen/Tatum cratered on defence last season (even Smart took a dip, although more on him later), and it isn't weird for good defenders to post similar or worse numbers in their down years to average/below-average defenders having the season of their life on that end of the floor. Does that mean that'll be the case next season? I think all of us can agree that both of them will bounce back with a proper offseason to rest, less tension in the dressing room, etc and just dwarf Kemba in these metrics like they have in previous seasons when they start playing defence again come October/November.
Simmons and Giannis are in the 30s after filtering out low minute players. This obviously underrates Giannis, but the stat has ranked Giannis near the top of the league in his DPOY year when the Bucks were actually going full throttle in the regular season (they definitely coasted this year on defence, they were 10th in DRTG). As for Simmons, I actually don't think that his current rating is
way off: he's not far off from the likes of Kyle Anderson, Robert Covington, OG Anunoby, etc - all strong defensive wings while he was bunched up together with Paul George, Draymond Green, Jrue Holiday, etc in 2020.
You make valid criticisms of the metric for Smart, Connaughton and Robert (although in fairness to RAPTOR, Smart and Pat this year are probably a result of some really weird stuff - it nailed their defensive value in his past few seasons), but this is where data triangulation comes in. Estimated Plus-Minus reverses its evaluations for Pat and Jrue compared to RATPOR (it has Jrue among the cream of the crop while Pat is good but nowhere close to being great) and rates Robert/Smart highly.
To conclude, using multiple members of the adjusted plus-minus family and applying context/team situation/eye test/etc to them gives you a fairly accurate picture of how good a player is. It's definitely something that most fans aren't used to, but it's certainly much better than looking at stuff like DWS and have truly wonky results pop up even when you filter out players whose data might be noisy (is anyone ready to crown Jokic as the fifth best defender in the league over Giannis last year?
)