Predicting the 2018-19 NBA Season, from the Warriors and Rockets, to the Knicks, Suns, and Kings.10/16/2018 
With the NBA back in action this week, we are launching our inaugural NBA Predictions to see how the league will shape out this season. While the Warriors are the obvious favorite to lead the the league in wins, since a majority of the league’s 30 teams make it to the playoffs, whether you agree with that or not, the battle for the last few seeds always goes down to the wire. Here's how we generated our 2018-2019 NBA playoff and win predictions:
We used the same methodology that we developed and tested against the 2017-18 NBA season a few weeks ago. By looking at casino over/unders and expert panel picks, we can generate an expected, or average, win total consensus for each team. Then by looking at the average team variance in their rolling 5 year win percentage, we can determine an expected variance for each team. With a unique average/variance combination, we can use the Method of Moments to generate hyperparameters for a Beta distribution unique to each team.
Using data and domain expertise to estimate the hyperparameters for a prior distribution, in our case a Beta distribution, is known as Empirical Bayes. For full transparency, the win projection estimates were taken from the following sources:
The consolidated version of this initial data can be found on our Github here. Let’s use the Warriors as an example to illustrate this methodology. Our 9 casinos and experts estimated that the Warriors would finish with 62.5 wins this year, or a win percentage of 0.762. This is our mean, or average. Over the past 5 years, each NBA team’s winning percentage on average had a variance of around 0.009. We will use this as a constant going forward, since some teams varied wildly with their 5 year winning percentage. We made the assumption that this value of .009 is comprised of two variances, the variance of luck, and the variance of true talent. If the Warriors have a projected win percentage of 0.762, then the variance of luck over an entire season of 82 games can be calculated by (0.762*(1-0.762))/82 ~= 0.0022. This means that the Warriors have a variance of true talent of 0.009 - 0.0022ish = 0.0067. Sticking 0.762 and 0.0067 into the following function gives: Alpha = 19.82 Beta = 6.18 Beta(19.82, 6.18): 
For comparison, here is what the Rockets prior distribution ends up looking like: 
Since there is some overlap, there is a chance that the Rockets will have a better winning percentage than the Warriors this season, but of course, the most likely scenario is that the Warrior will win the Western Conference. While expected wins can be determined by just taking the average of each team’s prior distribution and multiplying by 82, playoff odds are a little more complicated. For each team in each conference, we take one sample from their priors, and then determine the top 8 in win percentage. We repeat this 100,000 times. The proportion of times that a team makes the top 8 equals our estimate of their playoff odds. The 100,000 repetitions is important because it takes into account the slight chance that the Kings will have a better record than the Warriors this year. 
Without further ado, here are our NBA preseason Playoff Odds and Predictions:
Western Conference: 
Interestingly, the Western conference is the one that FiveThirtyEight and our predictions diverge the most. For example, we have the Spurs at 44 wins, with 57.5% chance of making a 22nd straight playoffs, while FiveThirtyEight has them at 37 wins and only a 25% chance. We are also more bullish on the Trail Blazers, and more worried about the Timberwolves than FiveThirtyEight. While we give the Clippers only a 16% chance, last year we gave the Pacers a 17% chance of making the playoffs at the beginning of the season. Surprise upstart? Locks: 1) Warriors 2) Rockets 3) Jazz 4) Thunder Probables: 5) Lakers 6) Nuggets 7) Pelicans Cusp: 8) Timberwolves 9) Spurs 10) Trail Blazers Long-Shots: 11) Clippers 12) Mavericks 13) Grizzlies Out: 14) Suns 15) Kings Eastern Conference: 
Similar to last year, we expect the Eastern conference to be less competitive than its Western counterpart. While we gave the Western conference ten teams with over a 50% chance of making the playoffs, the Eastern conference only has 8, with an astounding 5 over 90%. FiveThirtyEight is more bullish on the Wizards and Nets than we are, while we like the odds of last year’s upstart Pacers in making a repeat playoff appearance. While we only give the Knicks a 9% chance of sneaking into the playoffs, our estimate of 30 wins is 7 more than FiveThirtyEight’s total… a discrepancy as large as our estimate for the Spurs! Locks: 1) Celtics 2) Raptors 3) 76ers 4) Pacers 5) Bucks Probables: 6) Wizards 7) Heat Cusp: 8) Pistons 9) Hornets Long-Shots: 10) Nets 11) Cavaliers 12) Magic 13) Knicks 14) Bulls Out: 15) Hawks Finally, here are our win estimates with 80-90-95-99 confidence intervals. 
In case you’re wondering, here’s how our preseason confidence intervals did last year. Pretty much nailed it. We expect to see similar results this season. 
Which upstart team do you think will sneak into the playoffs? Do you have any questions about our methodology? Let us know in the comments below! As always, our code, data, and visuals can be found on our GitHub. Be sure to check back regularly as we update our NBA playoff odds and expected wins daily! The SaberSmart Team P.S. If you enjoyed this article, and need something off Amazon anyway, why not support this site by clicking through the banner at the bottom of the page? As a member of the Amazon Affiliates program, we may receive a commission on any purchases. All revenue goes towards the continued hosting of this site.
Comments
|
Archives
January 2019
Categories
All
|
RSS Feed