July 30, 2020 - by Jason Lisk
Do teams like the Milwaukee Bucks have lowered championship odds by not getting home court, and taking time off? (Photo by Jevone Moore/Icon Sportswire)
The NBA season is set to resume today, and the season restart presents a whole host of unusual factors:
All these things could affect what we see in the NBA restart, and how much the results match what we would have expected under normal circumstances.
Our research suggests we could see an increase in some NBA playoffs upsets for a couple of reasons.
One is the home court factor, and how that impacts the likelihood of the better seed winning a series.
The other is the time delay, and what that means for how we should project teams.
In a typical NBA playoff series (before the NBA Finals), the better seed gets the first two games at home. They then go on the road for two games, before alternating home-away-home. That means that the better seed has a net advantage in the number of home games — the first two games in the series and Games 5 and 7.
In the current playoff format, there is no such advantage for the better seed. It may be a subtle factor, but that increases the lower seeded team’s chances of winning the series in an upset.
Consider a hypothetical series where the better seed has a 60% chance of winning each game on a neutral court, a 70% chance of winning each game at home, and a 50% chance of winning on the road.
Here’s how the odds of that series being decided in X number of games shakes out:
NORMAL SERIES | Decided in 4 Games | Decided in 5 Games | Decided in 6 Games | Decided in 7 Games | Overall Chance of Winning |
---|---|---|---|---|---|
favorite | 12.3% | 24.5% | 17.7% | 19.8% | 74.3% |
underdog | 2.3% | 4.5% | 10.6% | 8.5% | 25.9% |
NEUTRAL SERIES | Decided in 4 Games | Decided in 5 Games | Decided in 6 Games | Decided in 7 Games | Overall Chance of Winning |
---|---|---|---|---|---|
favorite | 13.0% | 20.7% | 20.7% | 16.6% | 71.0% |
underdog | 2.6% | 6.1% | 9.2% | 11.1% | 29.0% |
The chances of a sweep are about the same. The chances the better team wins in 5 games goes down, as it’s no longer a home game. Overall, the better seed in this scenario is expected to win a series 74.3% of the time when played under the traditional format.
The same team is expected to win 71.0% of the time under the neutral court situation.
That 3.3% difference is almost entirely because of the Game 7 factor, and it not being a home game for the better seed.
The last time we saw NBA teams play was in early March. Teams that were ramping up for a final push toward the playoffs have now had more than four months between competitive games.
Normally, teams only see that kind of time off between the end of the playoffs and the start of the next season. They also haven’t been able to practice as a team for most of that time.
How teams will react to the delay is a great unknown. There simply are not any similar situations to truly compare, when trying to project what will happen.
What is more likely is that the best teams will be a little less dominant than they would have been if the games were played sooner.
We aren’t just making this statement based on opinion, though. If you want to see some more detailed charts and data to support “regressing toward the mean” because of the time delay between games, you can see that in a section below.
If you want simpler explanations, here they are:
That’s why, for example, our power rating for the Milwaukee Bucks will be +8.4 points better than an average NBA team when the NBA restarts tonight. That is, our ratings would expect the Bucks to beat an average team by 8.4 points on a neutral court.
When the season was postponed in March, we had the Bucks rated +9.2 points better than an average team. That decrease of less than one point may not seem like a big deal, but it does impact the Bucks’ projected odds of winning a bit.
The Clippers, Lakers, and Bucks will still be the favorites for the NBA title, but the uncertainty from the lengthy time off reduces their outlook a bit.
When the NBA season resumes, we’ll use ratings that are a little lower for those top teams compared to our March ratings the day the season was suspended. We’ll then adjust their power rating from there based on individual game results, with games after the restart having a bit more weight than games earlier in the season.
Our initial projections when the NBA resumes should be tighter than they would have been had the games been played in the spring. Playoff games on neutral courts also creates a slight relative advantage for the lower seeded teams.
So the net result is a slightly increased likelihood of upsets, as ratings tighten up slightly, and home court advantage is absent.
The below chart shows data from past seasons, and looks at the standard deviation of NBA margin of victory (MOV) predictions based on our predictive power ratings of the teams as of March 11 of a given year. (March 11 is when the NBA shut down this year.)
The horizontal axis shows the time elapsed in months from March 11:
The chart shows that the variance of outcomes increases the further away in time you move. In other words, the MOV predictions based on our power ratings get less accurate the farther in the future you go.
Some of those dots represent months with much smaller sample sizes. Months 3, 4, 15, and 16 are playoff months with fewer total games. Month 8 represents the small number of games at the start of the next season.
That’s why you see outliers from the general trend line for those months.
This next chart chows the coefficient you should use to adjust the March 11 ratings when best trying to predict the Game Score in a given month in the future.
This means that if you want to predict a team’s average Game Score in the first month of the next season — so, the point at (7, 0.77) — you use 77% of their March 11 rating from this season.
Now, the start of the next season usually comes after an offseason involving players switching teams. So part of that regression is due to those factors.
If you fit a model to the points in this graph, and add a flag indicating that there has been a season change, you get -2.4% more regression towards 0 for each calendar month, plus an extra -10.5% for the season change.
Printed from TeamRankings.com - © 2005-2021 Team Rankings, LLC. All Rights Reserved.