Dallas Mavericks 13-game winning streak: John Hollinger, etc.

John Hollinger is a basketball columnist in a way that I am not a basketball columnist—lots of people read his stuff. And he gets paid more than enough to buy, say, a coke. Secondarily, he has a lot of interesting analysis about the kind of stuff that you or I (well, I, anyway), wouldn’t think to wonder about.

But, of late, he’s made bad predication after bad prediction regarding the Mavericks. To the point that a surprising amount of time in John Hollinger’s life would seem to have been dedicated to justifying himself. Aside from his columns grading trades—in which he felt that the Mavs had not improved significantly, using his “trade machine” to generate an additional one win for the Mavericks over the course of the season (which has to have already been proven false)—and his Power Rankings which has yet to put the Mavs above 13th, he’s written two other ones in a month without a hint of suggestion he might be missing the boat here.

In fact, just today he wrote a third column addressing the problem again. Off the top of my head, I can’t remember Hollinger writing so many columns for just one team in such a short span. My guess is he’s taking a lot of heat for this in his inbox, and is probably kinda praying the Mavs’ luck turns. Because he’s human.

In fact, today’s column has to be the most infuriating Hollinger bomb to date, in which he explicitly lists the teams ahead of Dallas in the rankings in terms of whether he thinks Dallas could beat them in a seven game series. He thinks Dallas could beat the Celtics, the Spurs, MAYBE the Lakers, and possible the Hawks. He thinks they’d lose to the Magic, Jazz, Cavs (will, yeah, probably), the Suns, the Nuggets, the Thunder, the Trailblazers, and the Bucks. His most bizarre point is that the Power Rankings have only been wrong three times in predicting the Mavericks' games during the streak--against Orlando, Atlanta, and Charlotte. Why, exactly, the Lakers v. Mavs qualified as a draw and why Hollinger didn't think the Suns, holding down the fifth spot in his ranking, wouldn't beat the Mavs is another question, but yeah, it is true that besides those five games the Mavs have played teams they should have beaten anyway. I don't think, however, that you can punish a team for not beating teams that they did not play. 

How a team could be so bad and be second in the West must surprise John. Especially because the Mavs, heading into their 10th straight playoff berth, and 10th straight fifty-win season, with only positive changes having occurred between last year and this, are the furthest thing from a statistical anomaly.

Rob Mahoney thinks it’s no big deal, and he’s certainly right. It’s true that earning respect is part of the fun of being a fan, but Marc Stein has the Mavs at 2 right now, so that’ll do. More than that, we’ll take victory, any day (and the love of Steiny-Mo). It's true, too, that Hollinger’s statistical formula is pre-determined. Aside from the other columns he’s written, he’s just writing down what the numbers say. The other columns he’s written simply showcase his nearly religious faith in the numbers which, given that it’s propelled him to a position as an ESPN columnist, probably isn’t the worst idea.

But he’s NOT right, of course. And that's besides the fact that the trade has presumably already helped the Mavs win more than one extra gem, and it’s only been 14. It’s one thing to have a predictive model, like the trade machine—which I myself have used to trade such a Lebron, Durant, and Dwight Howard to the Mavericks, together, and gained only something like ten wins—and it’s quite another to write an actual column in which you agree wholeheartedly with the results of the predictive model. Two is what you should have a problem with, if you have a problem.

After all, it’s not like it’s rocket science to predict the Mavs being significantly better when they upgrade at two starting positions, and go from having one NBA-caliber center (so, 30 minutes or so of NBA-caliber center play, less against the better centers because of foul trouble), to two, one of whom is a phenomenal defender. If someone put that to you in a sentence, just like this, you’d have no trouble saying yes, obviously, the Mavs will be significantly better.

After all. The Mavs finished sixth last year and won a playoff series. They got better this year at four positions. Marion is an upgrade, Butler is an upgrade, Haywood is an upgrade, and Beaubois is an upgrade(on the backup side of things). Only the young teams—the Thunder, the Bucks and the Hawks say—got better. Only the Cavs improved significantly at the trade deadline, besides the Mavericks. The Mavericks have been successful for a decade, and again, improved significantly, so it’s not as if this year snuck up on anybody.

What am I missing?

The debate over statistical analysis has particular resonance for the Mavericks, and particularly for this week. Henry Abbott and others, coming off this year’s dorkapalooza have been happy to announce the arrival of advanced statistics on the scene, and it seems very much as if this is the case. “The Stats Movement Goes Establishment” was the Truehoops headline. The reporting on the events at this conference was especially interesting to Mavs fans for two reasons, each of which demonstrates a side of the stats revolution.

The first of these was Avery Johnson’s admission that his much-pilloried decision to go small against Golden State was based on the recently fired Wayne Winston’s statistical analysis. The second, showcasing how much upside statistics can offer, was Cuban’s discussion of his analysis of Kidd’s “win-time” performance—certainly something not all statheads agree upon. As you’ll remember, the Kidd for Harris trade was routinely denounced, by everyone, in its first year. It’s pretty obvious now that there was something to what Cuban saw in the stats. And there was something really awful in Avery Johnson not looking at a tiny team and thinking, hey, I can clog the lane and get almost all of the rebounds.

And that’s, I think, the way this works. Advanced stats can be enormously helpful. They can also, however, as in the case of Hollinger’s stubborn insistence indicate, be misleading—in a case where simply using your eyes would put you right. The obvious answer to Hollinger’s complaints about win-differential and strength of schedule and so on are not even intuitive points. They’re simple, factual, the kind of thing statheads should be comfortable with.

If we have a predictive model for how good a team will be, shouldn’t that model reset when a team changes as dramatically as the Mavericks have? I recognize that this is, to a degree, the point of emphasizing the last 25% of games played, but that’s simply not good enough. If, say, to use an extreme example, Lebron James goes down with an ankle injury and will miss the first round of the playoffs—is your predictive model any good at suggesting how well the Cavs will play in the first round, if it still has those games that James played with them on the dossier?

Likewise, what good does it do to try to predict the Mavericks’ success when you’re factoring in the time the Mavericks played with Josh Howard and Barea sharing time at the two, and Dampier and Dirk sharing time at the 5? 25% of the season is approximately 21 games. For the Mavericks, that goes back to Tuesday, January 28th a 106-112 loss to the Suns. They’ve gone 15-6 over that span, sure—because it includes their 2-5 skid before the all-star break—before the trade. Is that useful, statistically? Would anyone, even someone not nearly as bright or mathematically oriented as John Hollinger, think so, if simply asked whether, if one team changes its identity by slightly more than 40% (Butler and Haywood as starters, Beaubois as a competent backup, Najera and Stevenson as enforcers off the bench), using data from before that change would be particularly useful?

This is, actually, easy. And if someone wants to point out that the Mavericks margin of victory isn’t elite, well, I’d like to point out that SINCE the trade, they’ve actually won by an average of 6.2 points—7.7 if you drop the Thunder game, the first game the new team had together. 7.7, incidentally, is actually better than any margin of victory any team in the league has in Hollinger’s latest rankings. Any team. Even 6.2 ranks third, behind the Magic and the Cavs.

And though the strength of the Mavs’ schedule will continue to get worse—in case someone wants to talk about that—in the first seven games of the streak, the Mavs played the teams ranked 1, 4, 5, and 9 in Hollinger’s ranking. And beat every one.

In other words, Mavs 2.0 has a margin of victory that’s in the top 3—or top 1, depending how you calculate it—and a strength of schedule up there with anybody’s—and they have far and away the best record over the last 10, and they have the best streak, and they have the best division standing and they have the second best conference standing (to use all of John’s categories). Does it really make sense to try to figure out how good these Mavs are any other way than by looking at how the current, the vastly different from previous, team is playing?

So Hollinger’s model fails not only the smell test—that is, the variance between what you see and what the numbers tell you you get--but also the math test. As long as reason is applied to math, in this case.
Now, Doctor Mahoney says this doesn’t matter—and he’s absolutely right, because it’s a power ranking. It’s not an AP poll. The NBA champion is not decided by John Hollinger. But I think there is a way in which this matters.

John Hollinger is the most famous basketball statistician in the world. And he’s wrong. And rather than either letting the numbers speak for themselves or attempting to explain the disconnect between his numbers and reality, he’s vocally, publicly, and frequently, made the case that all of this—all of this sustained excellence for more than a decade, buoyed by self-evident and easily explicable recent improvements—is an illusion. Now, right now, when the public is being exposed to the relevance of statistics in a way they never have been before. The same week, in fact, that Henry Abbott announced the arrival of the stats movement—in that week, John Hollinger is being publicly and loudly wrong. For reasons that are obvious, and should be obvious to him.

As my friend Eoin (who’s from Ireland and a Trail Blazers fan) put it: "The thing is, if Dallas goes on a deep run, he's actually going set, not only his formula, but the entire statistical movement, back quite a bit in the eyes of the casual fan.” And that’s the danger. I think advanced stats are good. Its most public prophet has a big public problem. And it’s not just Dallas fans who think so. This is a credibility issue for John--and, to the casual fan, very likely a credibility issue for advanced statistics. And that's the problem.

The West has been a beast for a long time. I’d be interested to see if its number 2 seed, this late in the season, had ever shown up so poorly on whatever numbers Hollinger is working. The irony here is that it doesn't appear to be his numbers that are at fault, so much as his refusal to apply them in a way relevant to the current Mavericks team. That isn't necessarily what it looks like, however, and he would seem to be doing a grand disservice to the stats movement by refusing to be flexible.

This is, of course, not to say that the Mavs are as good as their streak shows, or that they won’t come down to Earth a little bit. The Lakers have gone 5-5 in their last ten, and they’re more stacked than the Mavericks are. It happens—and there are also bound to be some letdowns. Of the remaining 16 games, only three are against what I’d consider title contenders (Boston, Denver, and Orlando). Memphis, who they play twice, is an athletic, young team, the kind that gave the old Mavs trouble and Portland is a feisty young team, and they play OKC again and so on, but that’s only three and actually two games that you’d think it’d be normal for the Mavs to lose. And they’re going to lose more than that, I’d guess. They’ve been playing on one leg for some time now. Even though Dampier and Haywood both appear to be back, and Terry shouldn’t be out too long, at some point the magic won’t be there.

But if you’re trying to figure out how good the Mavericks are, I’d suggest using the relevant statistics, and using your eyes.  And if you’re a fan, I’d suggest, more than anything else, just to enjoy it, no matter what John Hollinger thinks.

The best teams--the best teams for the foreseeable future--will be those that most successfully blend their numbers and their eyes. Sometimes, when you play by the numbers, you end up going small against the Warriors when you've won 65 games another way. Sometimes, though, it'll grab you Jason Kidd even though Devin Harris would SEEM to make more sense. Blending is the issue, and there are plenty of good blenders out there. Henry Abbott, Kevin Arnovitz, Darryl Morey and Mark Cuban come to mind. It's probably overstating it to suppose the stats community will need to do some public relations work to repair this--all I'm saying is, the best will know how to keep things in perspective.

Incidentally, I am not suggesting that Hollinger redo his calculations every time an injury or trade happens. I'm merely suggesting that when he chooses to provide color commentary to flesh out the whys, he might take some time to mention the reasons he may be wrong as well, rather than simply the reasons what appears to be an issue with the numbers is not an issue. And in terms of stats in general, I'm suggesting that they should be taken, as they almost always are, I imagine, with a grain of salt.