Statsketball: Updating the Western Conference Playoff Odds

Well, All-Star weekend is officially over, and I'm so glad that the Western Conference won and that Dirk scored literally every point in the game. That is what happened, right? (In case you missed it, that isn't what happened).

But, now that the game is over, it's time to get back to thinking about the end of this season, and there's no better way to re-gear than to take a look at how the Mavericks are doing in the playoff race, and what they need to do to make the postseason.

The Playoff Picture so Far

Since the last time I did a playoff projection with Statsketball, a lot of things have happened in the Western Conference: the Minnesota Timberwolves failed to take advantage of an easy January schedule, and they currently stand 3 games under .500 and sliding down fast.

Meanwhile, the Memphis Grizzlies have rocketed forward in the standings after getting Marc Gasol back from injury, picking up D-League superhero James Johnson mid-season, and getting 3-and-D king Courtney Lee in a trade with Boston. Between the three, Memphis was actually the best team in January by point differential, and they're trying hard to climb into a playoff birth, despite a temporary setback after point guard Mike Conley hurt his ankle.

Now, the Mavericks find themselves actually in the 6th seed in the West - above Phoenix and Golden State - but every team in the bottom 3 of the playoff bracket finds itself in a bitter battle for dear life between all the other playoff teams and Memphis.

Of course, the Mavericks are in the 6th seed primarily because their last week of games has been particularly easy - the game against the Pacers aside - while their last month is very, very difficult. In fact, despite that 6th seed, the likelihood that the Mavericks make the playoffs are not particularly better now than they were a month and a half ago since they're as likely to find themselves with 5 game loss streaks as anything, come March.

To that point, it's probably worth noting that Hollinger's Playoff Odds currently have the Mavericks finishing the regular season in the 7th seed, above Phoenix, with a 78.5% chance at the playoffs - compared to Memphis' 40.3% chance. Putting Phoenix in the 8th seed seemed unfair to me, and I've been bothered by how everyone assumes that they'll drop off at some point this season: they have an easy-ish schedule coming up, and Eric Bledsoe will be returning any day now. The Suns have proven that they're a legit team, and a real threat to Dallas' playoff hopes.

As one final prep before we move on to my current projections, perhaps we should take a look at the Projection that I had over a month ago. The visualization here was done by my friend Adam Jacobs, who has been so incredibly gracious in doing the visualizations for both projections. You can follow him on twitter at @NBAviz.

via cdn0.sbnation.com

So, all that said, what better time could there be to project how I think that the Mavericks are likely to do?

My Playoff Picture

For the full explanation of how I go about doing my playoff projections, you can read the projection piece I did a month ago, but to recap: the idea is that I take the average win% for each Western Conference team as the season has progressed. For these projections, I prefer the average win% to current win% to insure that if a team is in the middle of an abnormal win or loss streak when the projection is calculated, it doesn't affect the result.

Once I've calculated the average win% for each team, I run a model that projects the season 5,000 times, and uses each team's win% and their remaining schedule to see how they are likely to end the season. The schedule I publish here is the average number of wins that each team ends the season with, and each team's best and worst possible win total is included.

As well, there's one big change that I made to the model for this edition as opposed to the last one: it occurred to me that, first, the model did a very poor job for accounting for teams that are on impressive rallies and win streaks or collapses and losses; or, to put it another way, because I'm using average win%, the model does a poor job accounting for teams that are genuinely getting worse or better. Second, it does a poor job with injuries. Luckily, I had a solution.

Over the last 20 games (the last large sample size of games that any given team has played) I can do some math to see any team's rate of improvement or regression over those games, and use the team's particular rate of growth or regression to weight their win% for my model.

Basically, if a team is getting better or worse by a significant margin, I can find out how much better, and use that knowledge to give them due credit when I run my projection. This accounts for both random improvement in a team (the team "gelling," or whatever else you want to call it), as well as players getting injured and returning from injury.

However, it also necessitates that the calculation be done after a relatively stable period, when a major player both hasn't been injured in the last 10 or so games (though players returning from injury is fine, because the win% from the period before the injury can be used).

So, for example, in the period between when Marc Gasol returned from injury and the All Star game, Memphis has had a 6% increase in their average win%. That is a big deal. That's an increase from winning 48.5% of their games and 51.4% of their games. It's probably worth mentioning, too, that Gasol seemed to be hurt in the game prior to the All-Star break, but latest reports seem to indicate that he is ok, and so I did my projections under the presumption that Gasol will remain healthy.

So, without further ado, here are my projections for the rest of the season, starting from right after the All-Star break:

I also have a visualization of the new projections (this time in color!), again from Adam Jacobs (@NBAviz), who I thank profusely.

Notes About the Projection:

• I also started this project trying to either prove or disprove Andy Tobo's theory that the Mavericks are the most predictable team in the league. So far, he seems to be emphatically correct. First, there was only a variance of 5.48% to the Mavericks' average win%, which is more than 4 times less than the variance to everyone else's (except the Clippers, who are apparently also quite predictable). Similarly, over the last 20 games, the Mavericks' win% has only changed by 0.6%: as much as ten times less than the Grizzlies or Pelicans, and even fairly stable teams like the Clippers and Warriors have changed three times as much as the Mavs. Andy is right. They're incredibly predictable.

• You may notice that the Grizzlies and the Mavs project to be tied. The Grizzlies have a smaller win% to this point, but they have a much easier schedule to end the season, and have been a force of late. It's going to be a hell of a race between these two teams to end the season, as only one team needs to win or lose one unexpected game to totally shift the balance of power. Of course, if they do end the season in a tie (which is quite unlikely) the Mavericks already own the series tiebreaker and would make the playoffs.

• If the Mavericks and Grizzlies are the two teams that end up vying for the final spot, it's worth noting that there's a 30% chance that the Mavericks end with more than 50 wins, but there's a 40% chance that the Grizzlies end with more than 50 wins. On the other hand, there's only a roughly 50% chance that the Grizzlies end with above 48 wins, but there's a 70% chance that the Mavs end with more than 48 wins. Basically, the Mavs are fairly stable, and so, there's a fairly good chance that the Mavs end with wins in the high 40's, but they're not likely to win too many more games than we would expect. Meanwhile it's still kind of a toss up as to whether Grizzlies end the season that well, but the Grizzlies have better odds of winning an unexpected number of games.

• It's also interesting how large the dropoff is from Grizzlies to the Timberwolves. The Wolves do, in fact, still project as an above .500 team for the season - they have a really easy March - but they were supposed to be well over .500 at this point in the season, too. I'm not entirely sure where the perpetual insistence/faith that they're a good bet to make the playoffs comes from, given that there's only a 4.85% chance that the Wolves end the season with over 50 wins, now, and 50 wins is increasingly looking like the threshold needed to make the playoffs.

• I'd also like to take a moment to say how crazy it is that the Suns project to end the season with 52 freakin wins, in the 6th seed, above the Warriors. No one, before the season started, had them over 25 wins; not a single person. But, now they're projecting to above 50 wins. Unreal.

• Just in case, I did model a projection in the instance that Gasol's pre-All Star injury is serious. It does not appear that it is, particularly, but I will include that projection here, if you're curious.

• Finally, there's the usual disclaimer that this is all massively subject to change. Things like injuries are, kind of, taken into account, but any major unforseen injury still throws the system off pretty majorly. As well, no one, 2 months ago, could have predicted that the Wolves would completely fall apart through their January or that the Grizzlies would surge into playoff contention. Random stuff happens. This model does its best, but it can only account for so much random stuff. In fact, sometimes, these kinds of projections are most interesting for what they tell you about each team now, and less about what they tell you about each team's future.

Statsketball is a registered trademark.