A lesson on What Not To Do

Bill Simmons links to a "spirited defense of Shawn Bradley" in one of his latest Twitter posts. It's actually a bit of an interesting read, but I think it's also a great example of the use of bad statistics (specifically a single argument I've grown to hate over years of reading boring arguments about who should and shouldn't be in the Hall of Fame). It also is a pretty cool opportunity to play with Basketball Reference's search feature.

Now, don't get me wrong: I pretty much agree with Mr. Shawn-Bradley-Lover. Bradley wasn't a great center, but he was alright. He developed into a pretty good defender as time went on, and (so long as you weren't building a franchise around him) he was a good guy to have on your team. He was average.

Anyways...

In reality, not only did Shawn Bradley have a solid eight year run to begin his career, but the three-year stretch between 1995-1996 and 1997-1998 were pretty remarkable. Even Tommy Boy and the "one and a-half percent of his brain" that he uses can understand that the following Bradley stat is pretty impressive:

Between 1996-1998, Shawn Bradley was the only Center in the NBA to average at least 11.4 points and 8.1 rebounds while appearing in at least 64 games each season (click here for the basketball-reference.com supporting data.)

Now, if there was anything really valuable in life that reading Bill James teaches you, it's that when somebody says "Only Willie Mays and Player X have career statistics Y and Z," you need to immediately question pretty much everything about that statement.

So, when Mr. Bradley-Lover says, "Only Bradley averaged at least 11.4 points, 8.1 rebounds, and 64 games from '96 to '98," we need to ask a few things:

1) Why did Mr. B-L choose such specific numbers?

2) How close were other players?

3) Does it matter that other players don't make those cutoffs?

So, why did Mr. B-L choose 11.4 pts, 8.1 rebs, and 64 games? Because those are Bradley's minimum stats in each category from the '95-'96 to '97-'98. Now, that's not really shocking, and there's nothing wrong with this. It's just that we need to know that what Mr. B-L is really claiming here is "No other player produced as much as Bradley from '96 to '98."

We can answer #2 and #3 simultaneously. Essentially, these cutoffs are arbitrary. They don't mean anything. Let's pick another player: one who appears twice on Mr. B-L's search query on BRef. What about our friend Dikembe Mutombo?

From '96 to '98, Dikembe Mutombo never appeared in less than 74 games in a season. He never averaged less than 11 points per game, nor less than 11.4 rebounds. If we search for that combination, only four individual seasons fit our criteria from '96 to '98: Mutombo's three seasons, and David Robinson's '96 season.

What about Hakeem? Well, he was injured in the '98 season, so his numbers are off. Still, he never averaged fewer than 16.4 pts/g, 9,2 rebounds, and 47 games. And the only other players who meet those criteria in all three seasons are Shaq and Alonzo Mourning. Pretty good company, right (of course, that's to be expected. Both are almost certain HoF-ers, and Hakeem was pretty good, too)?

Sabonis: never fewer than 13.4 points, 7.9 rebounds, and 69 games. And he's the only one to do it during those years.

The point is this: if you take enough of a pretty good player's stats, and you set them as the bare minimum for some arbitrary group across a long enough time period (corresponding to that player's peak), you'll probably narrow it down to a group of about four players, at most. With starting centers, the group is so small that you can probably narrow it down to one or two almost every time. Things get dicey with guards, but whatever.

For instance: which forwards averaged at least 80 games, 9.9 points, 4.1 rebounds, and had at least a .442 FG% from 2005 to 2007? Only Shane Battier, Tayshaun Prince, Shawn Marion, and Memo Okur. Ah, but who did all that on less than 8 FGA/G? Only our man Shane. Clearly, Shane Battier was having a truly remarkable three years. Even Tommy Boy could see it.

So we can answer the questions I asked above. How close were other players? Pretty close, actually. If you reduce pts/g by 0.1, you can add Dikembe into the Mr. B-L's group, and if you set the criteria as Deke's points and rebounding stats, then you throw everyone else out, including Bradley.

And it doesn't matter that the players who come close don't make the cutoff, either. I showed that you could isolate Sabonis, Deke, and even Battier (ten years later) if you set the averages right. It only matters if the criteria are meaningful.

When someone says, "Tracy McGrady was the only forward to average at least 67 games, 25.7 pts/g, 6.0 reb/g, and 5.5 ast/g from 2003 to 2005," that's impressive and meaningul because the criteria are impressive and meaningful.

But Bradley's 11.4 points and 8.1 rebounds are not particularly impressive stats for a center, and so the group created by those stats in that time period isn't impressive. Despite what the linked article states, the fact that Bradley produced at least 67 games per year of 11.4 points and 8.1 rebounds from '96 to '98 is not "pretty impressive." It's underwhelming. It's evidence of mediocrity. It's evidence that maybe such a player should not have been taken #2 in the '93 draft. But the argument sounds good because it's presented in the right way. Look at Sabonis's numbers again. Yeah, he was the only guy who had at least 69 games of 13.4/7.9 basketball, but does that scream "impressive" to you? If it does, it's only because it's presented alongside "and he was the only one to do it." Without some more context, that doesn't matter much.

So, the lesson to take from this is that, by carefully manipulating the data you select and wording your argument properly, you can make almost any player look like an all-star. Please use this knowledge only for good.