The NBA playoffs are upon us! There are a hundred different ways to predict who will win. In the first round, in particular, it's a fairly straightforward task. That's especially the case this year, where according to advanced metrics, the home team is quite a bit better than the away team in each series.
For a couple years now, we've looked at the playoffs by borrowing from Bill James' log5 method used in baseball. (It's always been popularized for the NCAA Tournament by Ken Pomeroy.) There are two pieces of the log5 method that need explanation.
First, instead of using straight winning percentage in the calculations, we use Pythagorean win percentage (which we'll call Pyth). This uses points scored and allowed to present a more accurate assessment of team quality. Margin of victory -- which is a proxy for Pyth -- can tell you the difference between teams with similar win-loss records. That's the idea. Its record is strong.
Where the log5 method really stands out is in giving us a probability for every which way a series can end. Basically, using the teams' Pyth and a standard NBA home court advantage (the home team will beat an equal opponent 60 percent of the time), we come up with probabilities the higher seed will win any given game at home and on the road. Then we use those numbers to calculate the probability of each series result: a four-game sweep for each team, a loss-win-win-win-win result, a win-win-loss-loss-win-loss-win, etc. We then add up the probabilities for every series ending in five-, six- and seven-game victories for each team. And now, we present them!
Here's how to read this chart: the percentages in each column correspond to that team's probability of winning the series in the number of games listed in the left column. So according to log5, there's a 19 percent probability the Thunder will beat the Rockets in four.
All in all, the Thunder are heavily favored. As the No. 1 team in the league by Pyth, this is expected. The Rockets would have stood a much better chance against the Spurs, though San Antonio would have been favored. The most likely scenario appears to be a gentlemen's sweep by the Thunder.
Spurs (2) vs. Lakers (7)
Interestingly enough, the Lakers have the exact odds of beating the Spurs that the Rockets have of beating the Thunder. Which is not good. Of course, log5 doesn't exactly know that San Antonio has had some roster upheaval, nor that Kobe Bryant is done.
As in the Thunder-Rockets series, the most likely scenario is a five-game win for the higher seeded team. But we shouldn't necessarily be surprised if it goes to six or seven -- there's a 47 percent probability it lasts that long.
Despite a pretty fantastic record most of the season, Pyth never really liked the Warriors much. Golden State is actually considered the worst West playoff team by this measure. While Denver isn't graded as highly as the Thunder or Spurs -- or Clippers -- they do get a big ol' nod over the Warriors in a seven-game series.
The most likely result is a five-game victory for Denver. I would be willing to wager that if that were the case, that one Golden State win comes at home.
Clippers (4) vs. Grizzlies (5)
Given the similar records, you might guess this would be a toss-up. And it still might be. But log5 likes the Clippers by a solid margin.
Worth noting: the Clippers didn't win home court advantage in this series until Wednesday night, when they edged the Kings. If they would have lost HCA, however, they still would have been favored 60-40 in this series. Pyth really likes the Clippers.
So ... the Bucks.
I don't remember seeing a lower log5 projection for a playoff team in recent years, despite the Bucks not being the worst playoff team we've seen. (Word up to the 2007-08 Hawks ... who, of course, took the eventual champs to seven.) There's an 80 percent probability that this series will end in five games. And a 0 percent chance that happens in Milwaukee's favor.
Many Knicks fans are nervous about the Celtics in the first round. The log5 system and Pyth do not agree.
log5 is pretty judgmental of Boston's late-season performance, or lack thereof. That said, this is the projection I'm least confident about. Paul Pierce in Madison Square Garden? Kevin Garnett? I think this is a long series, though log5 gives New York a big advantage.
Pacers (3) vs. Hawks (6)
Another series not projected to be close. Indiana is heavily favored by log5.
There is a 42 percent probability the series will last six or seven games, but the odds heavily favor Indiana at that point. Again, log5 doesn't really know that the Pacers have struggled over the past few weeks. Adjust accordingly.
The closest first-round series, according to log5, will be Nets vs. Bulls. Brooklyn has a 65 percent probability of winning the series, but the odds of it getting to seven games are 31 percent.
The complexion of this series depends heavily on Joakim Noah's health. In other words, if Noah is not healthy enough to play and perform well, the Bulls' chances fall even further.
Of course, these are all just predictions: we know anything can happen in the playoffs. I, for one, can't wait.