The Wizards are winning. How excited should we be?
Very! But also, it's early.
There is a feeling that anyone who has spent any amount of time rooting for the Washington Wizards can likely relate to. It’s the feeling of unbridled, slightly1 delusional optimism. The feeling usually sets in after the team wins two or three games. Though, I have felt it after a single win. It might also show up in the middle of a game and then leave. Sometimes, it sets in when they don’t even win, but there’s a general sense that they could have won, and if that happened, oh, man, look out. I could go on, but this meme pretty much sums the feeling up:
We are at a prime “the feeling” moment in the Wizards season. This is historically early for the feeling to set in, but that makes it even more exciting. The Wiz are ranked first in the Eastern Conference. They have only lost three of their 11 games. They have won against good teams and bad teams. They have won games they should have lost. And they’ve won games they should have won, but in past years still would have lost. It’s all exciting and confusing and amazing. So, how excited should we be?
Do Early Season Wins Predict Overall Win %?
I decided to take a look at regular season data from 2000 to 2019 to see if a team’s win percentage for their first ten games or so was at all related to their overall regular season win percentage.
As shown in the scatter plot, there is a pretty clear relationship between how a team starts and how they do overall. The figure suggests that early season wins predict about 31 to 38 percent of overall win percentage. There’s also a lot of variation depending on the team.
Based on the historical relationship shown below, we would expect the Wizards to finish this season at around .585 given their current record. Overall, though, from 2000 to 2019, early season wins were not that strongly predictive of season win percentage for the Wizards (it’s about 33 percent for the Wizards on average). The data for the Wiz are kind of noisy. But the relationship between start and overall season wins were actually very strongly related for the Spurs, for example. Then there are teams like the 2007 Celtics, who won their first 10 games and finished the season with a win/loss percentage of 0.293. But overall, it looks like how you do in the first ten games is slightly predictive of how you do for the season.2
Still, the relationship is not as strong as you’d think based on “the feeling.” A first-place ranking in November is exciting. It’s great. I’m telling anyone who will listen that they need to pay attention to this current Wizards squad. But, the great thing about the NBA is wild stuff happens every year. We don’t know how Rui will do under a new coach and new teammates. We have no idea what’s going to happen when Ben Simmons finds a home or the Jokić brothers start waiting outside of arenas for opposing team’s players. We don’t know if the top of the East from last season will suddenly wake up, or if it’s all the ball’s fault.
This is just one figure and doesn’t take into account all of the other things that matter for both early season and overall wins. I think it’s both a good sign that the start of the season is not super strongly correlated with overall success and that it still does appear to be related to overall success. Teams have slow starts and recover, but good teams tend to be good across much of the season—beginning, middle, and end.
If the Wizards go on to lose their next eight games, I’ll probably lean on the “the correlation is there, but still pretty noisy” interpretation of that figure. If they keep the winning this season, the data suggest it is at least partly related to early season success. I like data, but I don’t watch the Wizards based on data. I watch them because I like the feeling that maybe, just maybe, this is the season or game or quarter where it all works out.
I originally wrote “mostly” here, but that seems kinda strong.
I don’t want to turn this into a technical discussion, but starting win% explains about 20 percent in the overall variation of regular season win %, with coefficient of 0.35 on start of season win%. This is estimated from a multi-level model estimating the relationship between seasonal win% and start of season % that accounts for variation across teams. Code can be found on Github here.