Is The Game 1 Score Predictive Of The Rest Of The Series?

Here will be a very short post that will take a stab at the question in the title.

With the eighth seeded Trail Blazers upsetting the Lakers last night (less surprising) and the eighth seeded Magic upsetting the Bucks earlier in the day (more surprising!), the natural question is what can we expect going forward. Should we have higher expectations for our underdogs in future games of the series?

As it turns out, yep, a bit higher expectations. But we should be cautious.

With the data I had handy, all playoff game results from 2003-2018 for a total of 240 series, I built a simple linear regression. The inputs: the home team’s margin of victory in game 1 (INPUT 1) and the difference in strengths of the two teams in the series, as measured by the difference in their regular season point differentials (INPUT 2). The output: the average margin of victory for the team with home court advantage in the rest of the series.

(To be clear, we are measuring the remaining series margin of victory for the “favorite”, the team with home court advantage in a theoretical game 7. So we would be projecting the Lakers’ margin of victory in the remaining series games against the Blazers.)

In other words, we really want to know if the score in game 1 (INPUT 1) adds something meaningful to our prediction of the scores in future game of the series, after controlling for the strengths of the two teams (INPUT 2). Because it’s a quick morning post, here is the grizzly R output of the linear regression.

Screen Shot 2020-08-19 at 8.43.04 AM

The result? Well, game 1 adds something to our prediction for future games. The coefficient for this term is about 0.09. This means that if the away team wins the first game in the series by 10 points, we would expect them to do about 1 point per game better in future games of the series than if they had lost game 1 by a point. Also, we see that the coefficient for the game 1 score is statistically significant at the 0.05 level (the p-value is the “Pr(>=|t|)” column). In other words, that 0.0869 value does seem to be different than 0, though it’s not significant at the 0.01 level for what it’s worth.

The (statistical) significance of the game 1 result actually kind of surprised me. I mean, we are often cautioned not to read too much into small sample sizes for good reason, and one game is indeed a small sample. Fluky things can happen. Our favorite can have an off shooting night, like the Lakers last night when they shot 15.6% from beyond the arc.

I still think we should not go overboard here. An underdog winning a close game 1 portends them being only fractions of a point better on average for the rest of the series than if they had lost game 1. Also, I think this result somewhat reflects the shortcomings of regular season margin of victory as a measure of team strength. Sometimes we are dealing with clearly different playoff squads than regular season ones and the game 1 score simply reflects this (think Cavaliers vs. Raptors in 2018).

And of course, it goes without saying that banking a win is good for the underdog’s odds of pulling the upset beyond any predictive power it might have on future scores in the series. They now only have to win 3 out of 6 rather than 4 out of 7.

The moral of the story for game 1’s? Adjust your priors for the series, but not too much.

If you want to replicate this analysis, the code is posted to Github.

 

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s