All-Star Thoughts and the Introduction of PAWS

Posted on February 20, 2007 by


The Junior All-Star game was not a very competitive affair. Looking at the teams beforehand, though, it did not look like this game had much chance to be competitive. The game played on Sunday, though, did not appear to be so lopsided. Yet, the West took an early lead and coasted to an easy victory.

If we look at the composition of these teams one would have expected the West to have an advantage, but the East certainly looked like a pretty good team. The average player on the Eastern squad had a Wins Produced per 48 minutes [WP48] of 0.177 in the first of this season. Given this average, one would expect this team to win about 72 games in an NBA regular season.

The average WP48 for the West was 0.237, which translates into 97 wins in a regular season.

A Digression

Okay, stop right there. The season is only 82 games long. How can you win 97 games? Well, you can’t. If this team played an entire season some of these players would likely play worse (given the law of diminishing returns). But which players would decline? Shouldn’t we build a model that answers that question? For that matter, shouldn’t we be concerned when a player produces a negative quantity of wins? After all, you can’t win less than zero games.

In some circles, the issue of teams winning more than 82 games or less than zero has been raised. For me, though, this is not much of a problem. What I want a model to do is predict an outcome under a scenario that is likely to be observed. It is unlikely that an actual NBA team will assemble an All-Star team. Likewise, it’s unlikely that a team will be assembled consisting entirely of very bad players. So I am not concerned about the prediction of the model in these circumstances, since these circumstances are not likely to be observed.

Furthermore, what I really want a model to do is tell me when a team is likely to be good or likely to be bad. An All-Star team is likely to be good, and Wins Produced tells us this. A team of all negative players is likely to be bad, and Wins Produced tells us this also.

Other models, though, can tell us that a player is “good” when in fact he is not. For example, as noted frequently, NBA Efficiency exaggerates the impact of scoring. Consequently, NBA Efficiency can tell me that a player is “good”, when in reality, that player is not helping very much. Okay, enough on that issue. Certainly this has been discussed in this forum in enough detail. Let’s get back to the All-Star game.

PAWS and PAWSPER at the All-Star game

In my post earlier today I mentioned “Win Score per minute adjusted for position played.” I also suggested that this needs an acronym. Within minutes the readers of this forum came to my rescue. The winning suggestion is PAWS or Position Adjusted Win Score (PAWS is a popular acronym, used by many animal organizations. So this is not original. Still, it sounds neat I think we shoud go with it).In discussing the per-minute evaluation, the name PAWSPERM was suggested, or Position Adjusted Win Score Per Minute. So from now on, in discussing performance in a single game, we will note PAWS and PAWSPERM.

In the All-Star game, how did each player perform in terms of PAWS and PAWSPERM? Let’s start with the losers (players are ranked in terms of PAWSPERM).

  • Dwight Howard: 8.8, 0.418 (PAWS, PAWSPERM)
  • Chauncey Billups: 3.9, 0.243
  • Chris Bosh: 5.6, 0.242
  • LeBron James: 7.4, 0.231
  • Richard Hamilton: 2.1, 0.139
  • Joe Johnson: 2.2, 0.122
  • Jermaine O’Neal: 2.0, 0.085
  • Gilbert Arenas: 1.4, 0.068
  • Shaquille O’Neal: -1.3, -0.078
  • Dwyane Wade: -3.6, -0.151
  • Vince Carter: -2.8, -0.177
  • Caron Butler: -3.9, -0.246

For the East, Dwight Howard was the leader in PAWS and PAWSPERM. Four players, though, were in the negative range – which for this stat simply means that they were below average. Again, as noted earlier today, WP48 centers around 0.100. So marks below 0.100 in WP48 are less than average. PAWS and PAWSPERM center around zero, so marks in the negative range are below average.

Turning to the West, we see that the MVP – Kobe Bryant – was the leader in PAWS. But Amare Stoudemire took top honors in PAWSPERM. Here are the results for all the Western Conference All-Stars:

  • Amare Stoudemire: 10.8, 0.515
  • Carmelo Anthony: 10.7, 0.428
  • Kobe Bryant: 11.8, 0.422
  • Shawn Marion: 8.3, 0.376
  • Tim Duncan: 5.6, 0.375
  • Kevin Garnett: 4.9, 0.352
  • Tracy McGrady: 5.2, 0.289
  • Dirk Nowitzki: 2.6, 0.160
  • Ray Allen: 1.8, 0.087
  • Tony Parker: 1.8, 0.076
  • Mehmut Okur: 0.1, 0.009
  • Josh Howard: 0.1, 0.003

Every single star for the West was above average. So it’s not surprising, given the performances of these players, that the West won so easily. Still, if we look at what the players did in the first half of the season, we would have expected the East to put up a better fight.

One needs to remember that this was not a “serious” game. Players were doing things that they would never do in a regular season or playoff game. And it was also only one game. So we should not look at the All-Star game as evidence that the West is better than the East.

To make that argument, all we have to do is look at the quality of the teams in each conference. When we simply look at the standings it’s very clear that the West is better than the East this year. This does not mean the Western champ will definitely win the title in 2007, but certainly we expect that team (be it the Mavericks, Suns, or Spurs) to be the favorite.

– DJ