Rondo vs the Elite PGs: By the numbers

by drza44 on Jan 31, 2011 2:44 PM EST

There was a thread this morning asking why Rondo doesn't measure out as well in John Hollinger's PER stat as we might think he is...that Rondo ranked 47th in PER, so what is it about his game that doesn't translate to that advanced stat? Well, that got me thinking. I'm really into the APBR-metric stats in basketball, as I think they give us a LOT of information that either isn't apparent or else just flat out isn't in the box scores. But in order for them to be effective, IMO, you need to a) understand a bit about how they work, b) understand a bit about their strengths and weaknesses, and c) look at a large enough cross-section of the "advanced stats" to get a full picture of a player.

So I wondered: how do our players measure out vs their peers, according to the nerd stats? We have a lot of internal debates around here as far as how good our guys are, and because of the way the Cs play their impact isn't readily obvious in the traditional points/rebounds/assists summary/Sportscenter blurbs. So, what do the advanced stats say about our guys?

I'm going to look into all 4 of our All Star caliber starters, but I'll start it off today with Rondo since he was the subject of the post that got me on this. I'll compare him against 9 other newsworthy point guards (7 that I thought of by name, then the best point guards out of each of the last 2 draft classes by reputation). I'm going to look at 5 different stats: Hollinger's PER, Basketball-reference's win shares, Dave Berri's Wins Produced, 82games.com's Roland Rating, and BasketballValue's 1-year adjusted +/-. I'll give a brief blurb about each stat based on my experience with them, and then at the end we'll look and see how those stats would rank our guy vs the other bests at his position.

Today we're looking at Rondo vs Chris Paul, Deron Williams, Steve Nash, Derrick Rose, Russell Westbrook, Tony Parker, Ray Felton, Steph Curry and John Wall.

PER: Hollinger's stat, probably the most popular of the "advanced stats", favorable (compared to other advanced stats) to volume scorers and players that generate a lot of free throws; generally ranks those considered "great" by the general public well, though also will tend to have role players with good scoring-per-minute very highly. Here is how each of our PGs ranks in PER:

	PER
Paul	26
Nash	24.05
Westbrook	24.02
Williams	23.08
Rose	22.89
Parker	21.06
Curry	20.67
Rondo	18.55
Felton	17.31
Wall	15.35

Win Shares: From Basketball-reference.com, shooting/scoring efficiency; loves points per shot (thus values FTs drawn). To account for different minutes played, we're going to look at Win Shares per 48 minutes played.

	WS48
Paul	0.284
Nash	0.195
Rose	0.189
Williams	0.184
Parker	0.179
Westbrook	0.155
Rondo	0.151
Curry	0.145
Felton	0.097
Wall	0.038

Wins Produced: Dave Berri's controversial stat (most likely to be trashed on an APBRmetric board) is also the one seemingly growing fastest in popular usage; wins produced values what he defines as possessions, so loves rebounds, steals, and blocks and doesn't like TOs; doesn't value shot creation, but does value assists. We'll look at Wins Produced per 48 minutes.

	WP48
Paul	0.4
Nash	0.335
Rondo	0.31
Williams	0.229
Westbrook	0.208
Rose	0.194
Parker	0.184
Curry	0.15
Felton	0.128
Wall	0.104

Roland Rating: 82games.com's Roland Rating is based upon a combination of PER and +/- stats. It looks at the individual PER of each player, the PER of their primary defensive assignment, and subtracts the 2 for a 1-on-1 value then they combine that 1-on-1 value with a team-impact based on-court/off-court +/- stat to get the rating. Tends to produce fewest "what???" rankings, because players that rank out highly in both the 1-on-1 and team stats are almost universally who we consider to be among the best in the game...though the order at the top isn't always what you'd expect.

	Roland Rating
Nash	15.8
Paul	12.5
Rondo	9
Rose	7.1
Williams	4.1
Parker	4
Curry	3.5
Westbrook	2.9
Wall	-2.9
Felton	-5.2

1-year Adjusted +/-: This is Basketballvalue.com's APM calculation. I don't love it because APM is so incredibly noisy that a single year (or less) doesn't give conclusive answers. Even 2 years may be too short for an APM calculation. I like longer APM calculations, 4 years or more, to really clean up the noise and give a robust effect. Nevertheless, we're talking about this year so this is both the 1-year APM as well as the (huge) standard errors for each guy:

	APM (basketballvalue)	APM SE
Rose	17.81	10.76
Paul	17.46	9.55
Williams	11.66	9.34
Nash	10.75	9.68
Curry	9.09	5.69
Rondo	5.53	6.83
Westbrook	-0.3	10.11
Parker	-1.49	7.31
Felton	-10.52	6.81
Wall	-12.33	6.01

Overall Rank orders: Giving each of our 10 guys a '1' through '10' ranking based on where they ranked in each stat, here is a summary of how each guy did. I'll add an average across the 5 stats (with standard error) to give us a better idea how our seat-of-the-pants-advanced-stat-cross-section-view ranks Rondo with respect to 9 of the best point guards in the NBA:

	PER	WS48	WP48	Roland Rating	APM	Average	Std. Error
Paul	1	1	1	2	2	1.4	0.24
Nash	2	2	2	1	4	2.2	0.49
Rose	5	3	6	4	1	3.8	0.86
Williams	4	4	4	5	3	4	0.32
Rondo	8	7	3	3	6	5.4	1.03
Westbrook	3	6	5	8	7	5.8	0.86
Parker	6	5	7	6	8	6.4	0.51
Curry	7	8	8	7	5	7	0.55
Felton	9	9	9	10	9	9.2	0.20
Wall	10	10	10	9	10	9.8	0.20

Conclusions:

Just about every advanced stat, no matter how calculated, agreed that Chris Paul and Steve Nash have been the 2 best point guards in the NBA this year. Likewise, just about every stat concurs that John Wall and Ray Felton have been clearly the worst of this top-10.

Rondo is smack-dab in the middle of the rest of those vying for a top-spot. With standard error considered Rose, Deron Williams and Rondo all overlap with each other for the 3 - 5 slots. Likewise Westbrook, Parker and Curry all overlap each other for the 6 - 8 slots. But Rondo is so well clustered that his ranking would overlap with everyone from Rose (nominally 3rd) to Parker (nominally 7th).

On the whole, I'm satisfied with that ranking for Rondo. He's not quite as good on the whole as Paul or Nash, but he's right there with Rose, Williams and the rest with a strong argument for top-5 PG in the NBA this year. Sounds about right to me.

Be respectful and keep it clean. Thanks.

Tweet Comment 18 comments | 2 recs |

Do you like this story?

Comments

Great idea to combine the ratings

I actually think you could build a very useful web-based service for this.
I find it amusing that Nash is on top of the Roland Rating, but I guess he usually plays defense on the least productive of the opponent’s guards.

by European NBA fan on Jan 31, 2011 4:20 PM EST reply actions

Thanks for the compliment...

…as for Nash, his Roland Rating value is almost entirely tied into offense. His on-court/off-court +/1 of +21/48 minutes is off the scales, and if you look into it the Suns are essentially +21 on offense when Nash is around while the defense is the same either way. Basically, Nash is irrelevant to the Suns’ defense but mega on offense. But since point guard is the position that generally impacts defense the least anyway, differences in defense among point guards aren’t really big factors.

by drza44 on Jan 31, 2011 4:36 PM EST up reply actions

I just would have imagined that his reputed lack of defense would show up in the statistics. But it makes sense that it doesn’t add up to much

Thanks for posting the link below. To bad the guy behind the site got a real job with basketball statistics instead of developing it further ;)

by European NBA fan on Jan 31, 2011 5:10 PM EST up reply actions

Also...

…I’m not the one that first came up with the idea to combine the rankings. The first time I saw it was back in ‘08, when someone did it. I’ve got a link somewhere that I probably should reference…ah, here it is.

http://www.countthebasket.com/blog/2008/05/28/comparing-player-ratings/

by drza44 on Jan 31, 2011 4:41 PM EST reply actions

ratings

What a joke these stats are. The real thing that counts is the 13 assists a game, 1st, 2.4 steals, 2nd, 52% FG Percentage 1st, and all defensive 1st team, that other crap is just BS.

by angloamer on Jan 31, 2011 11:22 PM EST reply actions

rpndo losses points in these types of stats becuase he doesnt shoot the ball 15-20 times a game.. if he took some of the lay ups he has instead of trying to get players like davis, perk, semih, shaq, kg, pierce, ray allen, jo, nate, von wafer going offesnively.. then he would average 18 points 11 assists 2 steals 50% fg, and for my money- i’ll take his 10 pts 14 assists 2 steals 51%fg.

by jdunn123 on Feb 1, 2011 12:42 AM EST reply actions

sorry- rondo! not rpndo

by jdunn123 on Feb 1, 2011 12:43 AM EST up reply actions

Not really true...

…again, that’s the point of looking at a wider body of the “advanced” stats out there. Yes, PER would prefer that Rondo shoot more, so it isn’t a great stat to capture what he does. But Wins Produced doesn’t value shot creation/volume scoring much at all, so it is more in-tune with what Rondo does. Also, Win Shares has a defensive component that Rondo measures out well in. And both Roland Rating and Adjusted +/- deal with defense, and APM ignores shots completely.

In other words, it just isn’t true that Rondo shooting more would help him in every advanced stat. Just the opposite, in fact. MOST advanced stats aren’t as concerned with volume shooting as PER…that’s a big point of looking at them, to note things about a player’s contributions that you wouldn’t get from a quick points/highlights level of analysis.

by drza44 on Feb 1, 2011 8:10 AM EST reply actions

win shares does, however depend on your relative share of scoring

And offensive win shares definitely reward volume scoring.

Win shares also score lower for a player on a ‘balanced’ team.

Rondo’s tendency to defer scoring in favor of assists, for example, pumps up the offensive win shares of Paul & Ray at the cost of his own.

by mmmmm on Feb 1, 2011 2:24 PM EST up reply actions

That sounds reasonable...

…and again supports why we should look across the full body of available stats as opposed to just picking one or 2 that might be measuring the same things. Based on your argument, it’s logical that Rondo did better in Roland Rating and APM than in win shares and PER. All 4 looked at different things, but some show Rondo’s areas of strength better than others. And in looking at all of them we get a more complete picture than looking at any one alone.

by drza44 on Feb 1, 2011 2:40 PM EST up reply actions

True

but I am not sure you get a better ‘overall’ picture than just looking at multiple discrete stats.

I don’t like the ‘average rank’ across these because I do not weight them equally. And indeed, the fact that several of these ‘roll up’ stats DO weight volume scoring heavily means that that ends up getting weighted multiple times.

Sort of like trying to diversify by buying 5 mutual funds, not realizing that each fund is invested 40% into the same stock.

I also am not fond of adjusted +/- very much for comparing players from different teams for a bunch of reasons, although I DO see a lot of value in raw +/- as part of assessing the value of a given player to HIS team.

So far, basketball doesn’t have anything quite as useful as WAR in baseball. None of these roll-up stats used here does a very good job, imho.

To return to my point above, I could tell more clearly that Nash does a phenomenal job at offense by simply looking at his point creation stats (points + 2.2 * assists). Ditto for D. Williams, Rondo and Paul. That would show that offensively those guys are all way above everyone else in offensive impact.

And on the other end, I could simply look at their defensive stats and see that of THOSE guys, Rondo and Paul are clearly better than the rest at generating steals, blocks and rebounds. Its not by accident that both the Celtics and Hornets have two of the best defenses in the NBA. Its not in spite of having those two. Those two are clearly two of the best defensive PGs in the game and are both significant contributors to their team’s defense.

Based on that simple, quick ‘roll up’ (or is it really a ‘parsing down’ of players) I have no hesitation in simply saying that Paul and Rondo are probably the best two overall PGs in the game.

Beyond that, any ranking is arbitrary, depending on how you want to weight the various stats.

by mmmmm on Feb 1, 2011 5:56 PM EST up reply actions

Reasons

As a Celtics fan I really appreciate what Rondo does for the team, and I can’t imagine anyone doing a better job than him. But it doesn’t always show in the statistics.
There are more reasons than these two, but they come first to mind:
1) People like Paul, Williams, Nash and Rose do not only run their teams offense, they ARE the offense. Not a lot happens, if they don’t do it. That’s also why they are in the MVP talks, which Rondo probably won’t be until the Big 3 retires. On the Celtics team, Paul Pierce or Ray Allen will often take over as playmaker if needed, and sometimes KG will make a play from the elbow. But as the Lakers game showed us, the Celtics are actually a better team, when Rondo is in command.
2) As a side effect of the above, there are games where Rondo sort of fades out. I believe this happens more frequent against bad teams. The C’s are actually able to be quite effective even when they are just going through the motions, but Rondo is not in control and it will be harder for him to take over when the game is on the line. So the ball goes to Paul or Ray, and Rondos assists will be a bit random. And the C’s won’t hit as many shots.

by European NBA fan on Feb 1, 2011 9:30 AM EST reply actions

There's some truth in that, but again...

…that’s why we look at a whole lot of different advanced statistical approaches. The box-score based approaches are generally based on what the stat-creator believes is a historic trend, often based on some type of regression analysis over time. In other words, general analysis over NBA history suggests that things like scoring efficiency and passing efficiency are both usually more indicative of winning basketball contributions than most other factors, so pretty much every boxscore-based “advanced” stat incorporates offensive efficiency in some way.

So the fact that both Nash and Paul have been much more efficient as both scorers AND passers than Rondo this season plays a large part in why they measure out better in the box score metrics. And their efficiency shouldn’t be helped by them being the offense…in fact, intuition would argue that they should be LESS efficient than Rondo because they have to do more and have lesser caliber players to play with.

Rondo measuring out below them in both the +/- based and the box-score/efficiency-based stats indicates to me that the reason they’re grading better can’t be tied very strongly to teammate quality. They’re just playing better so far this year.

by drza44 on Feb 1, 2011 12:37 PM EST up reply actions

Actually my view is, that Rondo isn’t playing up to his potential this season, probably because of the injuries. There are too many games where the C’s are just going through the motions on offense. But since the C’s are otherwise very efficient on offense, and there are others who will make the plays – though not as good as Rondo – the team survives most of those lapses.

I don’t agree that it’s counterintutive, that Nash and CP3 are more efficient, if you look at my argument. If all three have a positive impact on their team, it makes perfect sense that Rondo is neither very productive or efficient, when the Celtics are going through the motions, and it is up to Paul Pierce or Ray Allen to make the plays. Rondo will simply not be involved very much, and it will show in the stats.

by European NBA fan on Feb 1, 2011 1:48 PM EST up reply actions

In that case...

…I’m not sure you could really fault what the analysis is telling us. The numbers can’t prove that Paul or Nash are actually better players than Rondo, but they can show that Paul and Nash are producing more and more efficiently on the court this year than Rondo. Now, if I understand your post correctly, your argument is that Rondo COULD be better if he were fully healthy and going all out, but that since he isn’t playing to (your perception of) his full potential his stats aren’t as good.

If I’m understanding you correctly…essentially we’re both agreeing that Paul and Nash have been better so far this year. Your caveat is that you believe Rondo could be better if he had to be, but that since he doesn’t (and isn’t fully healthy) he hasn’t been as good, thus his stats aren’t as good. Is that right? If so, I think we’ve at least got common ground to work with. I think.

by drza44 on Feb 1, 2011 2:46 PM EST up reply actions

Yup :) I forgot to point out in the first post, that I agree that Paul and Nash are the superior point guards so far this season. The great thing about using the combination of rankings is that they actually reflect very well, how I perceive the hierarchy of point guards.

It also makes it quite difficult to make the case for Rose as MVP, even if you take team records into account. If it should be a point guard, imo it should definitely be CP3.

by European NBA fan on Feb 1, 2011 3:00 PM EST up reply actions

an assist is an assist is ... not an assist?

One problem with the numbers discussed so far is that assists are less easily quantifiable than points.

When Rondo makes a pass to KG, that is potentially an assist.

KG could pass it along.
KG could score.
KG could miss.
It could be turned over (bad pass or mis-handled catch).

One thing that would be interesting to factor in is to try to measure just how effective a PG is at generating assists. That is, how efficiently his passes turn into points.

There are dependencies. In the above, suppose KG is a terrible shooter. Rondo’s assist numbers would suffer from that. Conversely, if KG is a GREAT shooter, then Rondo’s assist numbers go up.

On the other hand, if Rondo makes bad passes, maybe that drives KG’s FG% down. Or flip it.

The point is there are so many dependencies between the PG and his teammates.

I do know that no other team seems to be able to get the ball to the rim as consistently as the Celtics do. The Nuggets score a tiny fraction of a shot more at the rim per game (17.1 to the C’s 16.8), but take more than 3 exrta shots to do so. The Celtics shoot an incredible 68% FG% at the rim because an impressive 60% of those shots at the rim are assisted. Both those percentages are tops in the NBA. The Celtics team FG% goes up in games where they have higher assist numbers.

This suggests that Rondo’s passing is a very important part of why the Celtics are able to attack the rim so much and also that, well, he’s very good at it. The net result is the team leads the NBA in TS%, FG% & eFG% in large part because of Rondo’s passing.

by mmmmm on Feb 1, 2011 2:48 PM EST up reply actions

I should add, of course that all that passing also contributes to a high turnover rate.

The Celtics unfortunately have the 5th highest TO rate. Not all because of passing (moving screens and traveling are just as big of culprits) but the high number of passes does contribute to it.

The net benefits of very few wasted shots far outweighs that. But it is a negative that one can’t ignore.

The Celtic’s TO rate is the one factor that keeps them from being at the top of the offensive efficiency ladder. Because on a typical possession, they are the most accurate at converting that first shot into points. TOs hurt because they rob you of the chance to take that shot.

by mmmmm on Feb 1, 2011 2:54 PM EST up reply actions