Nerdnumbers avatar

Westbrook or Love?

In the 2008 draft Love was drafted 5th by the Memphis Grizzlies. The Grizzlies then traded him to the Timberwolves for O.J. Mayo. That's five teams that royally messed up and one of those teams was the Oklahoma City Thunder -- who went by the Seattle Supersonics at the time. With the fourth pick they selected Russell Westbrook. The Thunder went on to extend a max contract to Westbrook, and many consider him an integral part of the Thunder's mammoth success. Kevin Love has had historic seasons but had struggled to get a top deal or into the All-Star game. Obviously we believe Kevin Love is a much superior player. Using the numbers doesn't even make this difficult.

Use the Boxscore Geeks Comparison Engine to see why Love trounces Westbrook.

Stats taken from 2013-12-31

Per 48 Minute Stats Kevin Love Russell Westbrook
Points over Par 7.7 0.3
Points 34.9 31.1
Defensive Rebounds 13.2 7.1
Offensive Rebounds 4.8 1.6
Assists 5.5 10.2
Turnovers 3.2 5.8
Blocks 0.4 0.2
Steals 1.1 2.6
Fouls 2.4 3.1

There are only two places Westbrook knocks Love. The first is assists. However, Westbrook is a point guard, and is expected to pass more. Relative to a standard power forward, Love gets 2.6 more assists per 48 minutes. Westbrook gets 1.8 more assists than the standard point guard. What's more Kevin Love's assist to turnover ratio is 1.72, which is just slightly below Westbrook's 1.76. It's arguable that Love is a better passer than Westbrook! I'll give Westbrook his impressive defensive rebounding and steals. However, typically this is not why people think Westbrook is such a star. 

Yay! Points!

Per 48 Minute Stats Kevin Love Russell Westbrook
Two Point Attempts 16.1 19.8
Three Point Attempts 8.6 6.4
Free Throw Attempts 10.1 8.8
Two Point Shooting Efficency 51.5% 46.2%
Three Point Shooting Efficiency 38.3% 30.9%
Free Throw Shooting Efficiency 83.0% 79.3%

Well, Love wins this one pretty handily. Even from the ever famous "usage argument", Love only gets 1.5 shots fewer per 48 minutes than Westbrook. And he gets 1.3 more shots from the line per 48 minutes. When it comes to shooting efficiency Westbrook is slightly above average than the average point guard at two point shooting. However, he completely loses this edge when it comes to his poor three point shooting and below average (for a point guard) free throw shooting. The net result: Westbrook should not be taking the most shots on his team given his percentages. Love on the other hand? I'd actually argue he should try shooting more!

Big beats small

A big part of the Wins Produced formula is comparing a player to their respective position. Guards have different responsibilities than centers. Of course Kevin Love's stats trounce Westbrook's right? Well, there's two keys to this. Even counting Love as a power forward, he comes up almost four times as good as an average power forward! Westbrook comes up right about average. But the other key is that big men are inherently more important in the NBA. Every NBA team to succeed repeatedly in the playoffs has had a top big (and yes, LeBron counts.) Point Guards have a lot of responsibilities, but aren't key to success. The Lakers made multiple finals with an ancient Derek Fisher at the helm. Chalmers is a competent guard, but I doubt people point at him as the driver for the Heat's success. And this makes the Thunder's original decision that much more unforgivable. If there's a good guard or a top big in the draft, you go with the big! Years later Love is filling the stat sheet, and doing in the right way. Westbrook is scoring lots of points, but don't tell me the Thunder wouldn't look amazing with Love next to Durant. Love or Westbrook isn't really a question. At least, it shouldn't be.

But, but, but ... Westbrook *looks* like a great basketball player! He's athletic. He can jump. Etc., etc., etc.

Even after Love's weight loss, who wants a guy that looks like a 6'10" version of a rec league player?

And, you know, that's how we really evaluate basketball players in the NBA, right?
Oh, one other comment. If Seattle had taken Love with the 4th pick in the draft in 2008, they probably wouldn't have taken Serge Ibaka later on in the first round. They'd probably have taken a PG at that spot, no? And the next 2 PGs taken after Ibaka were George Hill and Mario Chalmers.

If you were OKC, who would rather have - Westbrook & Ibaka or Love & Hill/Chalmers? Probably Love & Hill/Chalmers, but it's a much closer question, no?
But westbrook pretends like every mid range jump shot he makes in the second quarter just won the nba finals and he must scowl.
Only issue with Love over Westbrook is games played but thanks to Patrick Beverly that is now being evened out.
I don't think you ALWAYS take the big though. You have to look at what Love was as a rookie, his weight loss has made a huge difference. He is a different player than anyone projected he would be. RW pretty much is what people thought he might be. A combo guard masquerading as a PG. Who loves scowling
Def. picking Chalmers and Love. So far this year Mario has been outperforming Westbrook; plus, Mario has recently proven to be far more durable. Love is far superior to Ibaka at the PF/C positions.
I don't think it is clear that Love is far superior to Ibaka. This year, for sure, Love has been playing out of his mind - and we know from Love's 2011 season that dude is capable of a >.300 WP48 full season. But over, say, the last 4 years combined, has Love really been far superior to Ibaka? I don't think so.
The thing about Ibaka is he was a mystery box at the time. I say the Thunder either take both (never a problem stockpiling bigs) or have Love and Durant and be content.
Is OKC better without Westbrook?
Of course not! They've done a terrible job getting our point guards (Fisher, really?) and Westbrook is average to good. Just understand that Kevin Love is legendary and Westbrook is overpriced and overrated.
There is a least one team I can think of that didn't need a quality big to win - Jordan's bulls. Of course they were playing with a triangle offense so not sure how well that figures into that, but they won 6 champs without a quality big.
You could also argue Isiah's Pistons needed Isiah way more than Laimbeer
Imagine they got both Love AND Ibaka. Ibake is undersized at the 5 but who cares? He still blocks shots and rebounds.

Durant, Love, could play Nick Young and Norris Cole next to that front line and still contend.

Of course they had a quality big. They had Dennis Rodman. He was one of the best bigs of all time.
xitongzou ,
For their first three they had Horace Grant, who made it back to the finals with Orlando. For their second three they had Dennis Rodman.

You could also argue Mario Chalmers was more key to the Heat's success than LeBron James. Not sure I ever would.
Dre, I thought it was worth kicking around looking at the way Lamb, Jackson and Roberson have been playing.
I think most people agree a healthy* Kevin Love is a better per-minute player then Westbrook, but I don't like how this is slipping into a bash-Westbrook-athon.

For instance, I don't think Westbrook is overrated or overpriced. Most people consider him to be a top-10 player, and many NBA metrics (except WP) indicate that he is. Considering his staggering (until recently) durability, and the types of players getting max-money, I don't think he's overpriced at all.
With apologies, it's not a popularity contest. I don't go for "each metric gets a vote." Many of the popular metrics have major flaws, and Westbrook gets away with being a chucker, and gets rewarded just for being on a good team.

No apologies needed, and I wasn't trying to make it a popularity contest. I'm just pointing out that metrics like XRAPM, ASPM, and Win Shares, with great correlations to past and future wins, indicate that Westbrook is an extremely valuable player.

All metrics have flaws, which is why I look at all of them, as long as they show an ability to explain/predict wins, which these all do.

Being a "chucker" in terms of taking a lot of shots, isn't necessarily a bad thing. Also, all metrics have instances where they reward players for being on a good team.
"chucker"? It's like I'm really on /sp/. Why not just run a team of all bigs and no guards to get that team FG% up to 55+% and win every game?
It's an interesting question. I do think you need some ball handling. However, check out the 1996 Bulls. While they didn't have the conventional 7 foot center, they did actually have one of the tallest teams in the league by inserting "oversized" guards.

With the "all metrics have flaws" you again go for the "equal footing" which drives me nuts. As to your points about the APM variety metrics, I can just say I know you haven't dug into the math or inner workings as much based on how you cite them.
Without the phrase "small sample size," then, how would you explain OKC's struggles to score without Westbrook in the playoffs last year? If they were losing such an inefficient chucker who "gets rewarded just for being on a good team," why did they pretty much immediately cease being good?
I agree Love is better and that Westbrook's shot selection could use some improving, but I also certainly think there is something about how he contributes to wins that WP misses.
I recall me saying that they could replace westbrook with jackson and still contend. They are arguable the same player with westbrook having the assist edge but as a given.
Dre, saying "I can just say you haven't dug into the math" is as bad as saying "all metrics have flaws, so let's listen to all of them".

How do you know that? Because you have, and you don't think it's valid? If so, a better rebuttal would be to explain why, but even then, you're not the final authority on math and statistics, or you guys would have by far the most precise, predictive model, rather than a good one.

This is that tone crap that people keep bringing up. In any other domain, if you read or heard a comment along the lines of "I can just say you're not as knowledgable as me, so you're wrong", without any support, I assume you'd think it was a rude, wrongheaded thing to write/say.

All that said, I can't figure out what everyone else loves so much about Westbrook, especially the refs. Take away the FTAs and he's Brandon Jennings with more rebounds. And he's almost as egregious a bad-call magnet as Harden. He's even got his own flop move; where Harden has the two-arms-on-ball rip-through-and-stumble, Russ has the 45 degree face-first leap into a big guy 10 feet from the basket.
Please see the 6+ years at Wages of Wins, 2+ years at Arturo's Silly Little Stats, and now 2+ years here. We've written tons on this, we've shown the math. And I know people like Hoopdon are regular readers. Apologies if my patience for repeating the same arguments in the comments is not huge.

It's fine if you want to disagree with my tone, or say I come off as rude. However, none of that means you've actually done the work of analyzing the metrics.

The main issue with +/- variants is their ambiguity, both in terms of calculation, and how a high score can be consistently achieved.

However in terms of raw player evaluation, they are excellent, giving a WP like correlation to past wins, with an even better correlation to future wins.

If you know of something else I should be considering, please share.
I've chugged through most of Arturo's old site and WOW, and there is next-to-nothing on XRAPM, ASPM, and even Win Shares. Almost all "in-depth" critiques of metrics are of PER and raw/adjusted plus/minus. If I'm missing something, please point it out.

Also, your tone doesn't bother me, though I can't speak for other readers.
The "math" they cite is just that plus/minus is too small a sample size to get anything worthwhile.

The math they ignore is the xRAPM/ASPM both consistently predict and retrodict better than WP.
Has there been any WP lineup studies done? Players affect on WP of total lineups? I would love to see articles and stats on things like that.
It is just so hard for me to watch basketball and not think that if Tim Duncan is on the floor (Or Dirk or Howard) there are more corner 3's available if the floor is spread correctly and things like that.
Have you guys done articles on that before?
Westbrook is a player that I posit wins produced consistently underrates. Wins produced doesn't like him because of turnovers (which are in his case largely a result of being so high usage; wins produced does not handle stats that co-vary with usage well) and because he shoots too many 3's; I'll happily ding him for the latter but the former is an artifact of how the model is set up.
I'd like to see Dre tackle Hamthony's question. Regardless of what numbers help us to understand, context is still necessary to help us understand those numbers.
"That's five teams that royally messed up"

One of those teams would be Minnesota, which took Mayo, and traded him with three other players, for Love and three other players who they (presumably) preferred, right?

Another of those teams would be Chicago, which took Derrick Rose. Health issues aside, that's worked out pretty well from what I can see. And obviously, Love has had his own health issues.
Dre, it's not about the math behind WP, it's about the math for the systems you're critical of. I've been reading WoW since 2009, and, far as I can recall, you guys have never dug into the flaws in the other advanced stats models. You weren't dismissing Hoopdon's understanding of WP, you were dismissing his understanding of the other models, and making a claim that you've done the math, and they're wrong.

And hey, that could well be true, but (and this is a weak claim, given that I love the site and come here to read what you guys have to write), since you've never demonstrated it, it's not the kind of thing I can take on faith.
I've been wanting to comment for a long time in many different posts but never did because I am still learning so much and I don't have an analytical mind like most readers here. Patrick, Dre and Arturo are like professors to me and I love this site so much.

Having said all that, It would be so much of a relief to me, if I start to discuss more here even if I feel really unqualified! I can't talk NBA with nobody.

I would pick Kevin Love in a heartbeat, I think it's very obvious that he is the more valuable player, however I do feel people in this site and "statheads" in general bash on RW too much. He is overrated in an economic sense, but sorry to be a romantic here, he does bring a lot of the magic and intangibles that are too dynamic to be measured. Along with that he is incredibly durable, not including last postseason and just recently, I remember seeing that he played every game for 4 seasons, including playoffs, that's pretty huge.

I do think he is extremely annoying most of the time, but it works for him and his team I guess. Point is, I wish more analytical people can be less biased with players like RW, he is not a role player, he's one of the most important player, if not most important player on his team (not the best player, it's Kevin Durant(no brainer)) because he has the ball most in a game. Role players don't have the responsibilities of star players to take risks and create something out of nothing when opponents are focusing all their attention on them, they will miss a lot of shots and commit turnovers. Being a role player in a professional sport team is a lot easier to play it safe and hit their quotas/numbers/projections and even increase their efficiency.

All in all, sorry for the long essay without having numbers to back it up, I know most people will troll me, but I enjoyed all the comments and debates in this post even though it's very lopsided, KLOVE no doubt but I still have respect for players like RW.
The concept of the latest family of stats is really solid - we have lineup-level plus-minus data, and using that as the DV in a regression of box score stats could capture aspects that Wins Produced misses. I like this direction quite a bit.

It's development so far has been, well, these metrics are chimeras that have been cobbled together step-wise, and are methodological messes as a result. APM uses data on players from the previous year to help stabilize its results, for example; that helps the evaluations from any particular year to be more reliable, but when examining a sample of several years - each depending on data from the previous year, also in the sample - your sample is going to have inflated correlations from that double-counting.

Then there's the whole issue of ridge regression - don't get me wrong, I love regularization for dealing with small or ill-posed data sets - but what you get after regularization is a biased posterior. It's great when that bias can reduce the size of your CI, but what you have there is a model, not empirical data - and I don't know why you'd think it was a good idea to run a regression with that model as a DV instead of the data.

Again, I like this direction quite a bit. It has a lot of promise. It's also a much more complicated multilevel model that needs to be put together carefully and purposefully. Stats like ASPM are cool proof-of-concept models, kludges using what is readily available; when this is put together correctly from the ground up, I'd speculate that it'd be the best model out there.

We're just still at the stage where we're scribbling on napkins.

Sign in to write a comment.