At Basketball Prospectus, we've made log5-based previewing something of an institution during our coverage of the college tournament season. While I've tried to drop log5 forecasts into our NBA playoff coverage the last few years, it's not as much of a centerpiece, which is fine. Nevertheless, I always run the projections in order to set a baseline for my own expectations of the bracket. Today, I share the results of those log5-based simulations with you.
Others do log5 projections. I think mine are different in two ways. First, the inputs I use are the team power ratings (POW) generated by NBAPET, my projection, evaluation and analysis system. The ratings adjust a team's Pythagorean win percentage for home/away success, strength of schedule and record in close games. I don't give added weight for recent performance for a simple reason: I have not been able to find significant evidence that suggests you get a more accurate playoff predictor by weighing team performance by recent results. As is almost always the case in quantitative analysis, the bigger the sample you use, the better off you're going to be. I express my adjusted Pyth% in wins per 82 games, which makes interpreting the results a simple matter. My past research indicates that POW results are fractionally more accurate in forecasting playoff success than point differential and quite a bit better than using raw won-lost record.
Anyway, I never presented the final NBAPET Hoops List, so let's take care of that right now.
TEAM, POW TEAM, POW
1. San Antonio Spurs, 60.0 16. Phoenix Suns, 40.4
2. Chicago Bulls, 59.4 17. Atlanta Hawks, 40.4
3. Miami Heat, 57.2 18. Utah Jazz, 38.1
4. Los Angeles Lakers, 57.1 19. Indiana Pacers, 36.7
5. Dallas Mavericks, 55.4 20. Milwaukee Bucks, 36.2
6. Boston Celtics, 55.1 21. Golden State Warriors, 35.5
7. Orlando Magic, 53.4 22. Los Angeles Clippers, 32.1
8. Oklahoma City Thunder, 53.0 23. Charlotte Bobcats, 31.2
9. Denver Nuggets, 52.2 24. Detroit Pistons, 29.5
10. Portland Trail Blazers, 48.1 25. Sacramento Kings, 26.4
11. Memphis Grizzlies, 47.9 26. New Jersey Nets, 22.6
12. New Orleans Hornets, 46.2 27. Toronto Raptors, 22.3
13. Houston Rockets, 45.6 28. Washington Wizards, 20.7
14. Philadelphia 76ers, 42.0 29. Minnesota Timberwolves, 20.5
15. New York Knicks, 41.8 30. Cleveland Cavaliers, 18.1
The Bulls spent the last three months of the season closing on the Spurs, but just couldn't quite get there. The primary factor in this was strength of schedule--Chicago's weak Central Division brethren resulted in the Bulls having the easiest record in the league. That really only matters in a comparison of teams from different conferences. The Bulls' slate might have been the softest, but coming right after them was Miami, Boston and Orlando. NBA schedules are fairly uniform and these teams didn't have to play themselves. The other component is conference strength and, once again, the West was the more powerful circuit. So the Bulls' schedule doesn't work against them when comparing them to Miami or Boston, but it does set them back a tick against San Antonio, which was in the middle-of-the-pack in schedule strength. (Minnesota faced the league's toughest slate. Again, they didn't get to play themselves.)
I mentioned that there were two primary differences between my log5 forecasts and others that you will find on the interwebs. The first was the POW inputs; the second is that I take those inputs and run them through a Monte Carlo-style simulator 1,000 times in order to get a feel for each team's probability for advancement. The results are listed in the following table. I listed the actual number of simulations in which a team advanced to each round. So the 961 under R8 for Chicago means that the Bulls advanced to the second round 961 out of 1,000 sims. That means Indiana actually knocked them off 39 times. I left out the percentages to avoid clutter, but if you care about something called "log5 projections" you can do the math yourself.
log5 PROJECTIONS, 1000 SIMS
Sd Tea POW R8 R4 R2 R1
E1 chi 59.4 961 704 455 272
W1 sas 60.0 836 619 413 232
E2 mia 57.2 872 520 245 125
W2 lal 57.1 810 461 223 107
W3 dal 55.4 712 389 171 78
E3 bos 55.1 835 412 163 77
E4 orl 53.4 839 271 125 57
W4 okc 53.0 551 174 72 29
W5 den 52.2 449 142 58 12
W6 por 48.1 288 85 22 6
W8 mem 47.9 164 65 25 4
E5 atl 40.4 161 19 6 1
W7 nwo 46.2 190 65 16 0
E6 nyk 41.8 165 35 4 0
E7 phi 42.0 128 33 2 0
E8 ind 36.7 39 6 0 0
The results aren't quite chalk--the Bulls won more overall sims than the top-ranked Spurs. This is because they got to the Finals more often and because, once there, they had the homecourt advantage in any head-to-head matchups with San Antonio. The Bulls' path is a little easier because of how weak Indiana is as an opening opponent, which gets them into the second round far more than any other playoff entrant.
I'm always fascinated by seeing which teams sneak away with a championship or two further down the ladder. In this case, the Blazers (.6 percent chance of winning a title) and the Grizzlies (.4 percent) found multiple paths to the title, which I suspect would be a shocker for anyone who follows the NBA. Can you imagine the "World Champion Memphis Grizzlies"? The Hawks won a sim as well, but any team can win one. That's an outlier.
Do I really think the Lakers really have just a 10.7 percent chance of winning the title? Not really. You can't always trust regular-season results when it comes to veteran teams with championship pedigrees. As a matter of fact, I suspect a "past championship" factor might be in order as an adjustment to regular-season POW. Or not. I'm not about to chase my tail again in another investigation for an NBA playoff secret sauce. Anyway, this isn't about dead accuracy. It's about a baseline of expectation. Now that we know when to be surprised and how surprised to be, let the games begin.
This free article is an example of the kind of content available to Basketball Prospectus Premium subscribers. See our Premium page for more details and to subscribe.
Follow Bradford on Twitter.
Bradford Doolittle is an author of Basketball Prospectus.
You can contact Bradford by clicking here or click here to see Bradford's other articles.