Sunday, April 21, 2013

The Point Spread, Home Court Advantage, and the Playoffs

At Wages of Wins, Ed Feng posted recently on the impact of home court advantage in the playoffs.  He found that home court advantage seems to have a bigger impact in the playoffs, even after correcting for relative team strength (e.g. 1 seeds vs. 8 seeds, etc.).

Feng found that home court advantage is worth 4.5 points in the playoffs, compared to 3.2 points in the regular season.  I had noticed a similar phenomena in the Vegas point spreads during last year's playoffs.  When a playoff series would switch venues, I was expecting a 6.5 point swing in the point spread (the betting market values home court at 3.25 points in the regular season).  Instead, the spread was swinging by 7 to 9 points.  For example, in last year's Lakers-Nuggets first round series, the Lakers were favored by 4.5 points in their first two games in LA.  But when the series moved to Denver for game 3, the Nuggets were favored by 4, instead of the expected 2 points.

In this post, I will quantify how much the betting market appears to value home court advantage in the playoffs.  I will also take a deeper look into the stats to see what is driving this home court advantage differential.

Impact on the Point Spread


To measure the impact of home court in the playoffs, Ed Feng compared scoring margins within each playoff series, and measured the difference between the two venues (this implicitly adjusts for relative team strength).  I have taken the same approach below in comparing the point spreads.  The chart below shows the average differential in the Vegas point spread for both regular season series and playoff series.


There is a clear gap between the regular season and playoff point spreads, consistent with Ed Feng's findings on scoring margin.  When averaged over the past 10 seasons, it amounts to about a half point.  I'm not sure what to make of that spike in 2007, nor the subsequent decline in the following years (note: I'm labeling each season by the year in which it began, so the 2007 data point is for playoff games played in 2008).

So, regardless of what is causing it, the betting market appears to have at least partially caught on to the amplification of home court advantage in the playoffs.  I say partially because the impact on the point spread is a half point, but in terms of actual scoring margins, it appears to be worth 1 to 1.5 points.

Based on this analysis, future publications of my team rankings and playoff simulations will assume home court advantage in the playoffs to be worth 3.75 points, compared to the 3.25 points I use for the regular season.

Home Court and the Four Factors


So what could be causing this gap in scoring margin?  The tables below summarize how the impact of home court advantage shows up in the Four Factors: Effective Field Goal Percentage (eFG%), Turnover Rate (TO), Offensive Rebounding Rate (REB), and Free Throws per Field Goal Attempt (FT).

In the same way that I measured the swing in point spread and scoring margin, I can measure the home/away split of any stat I choose.  The table below summarizes the difference in each of the four factors when comparing a team's home vs. road performance (within each series):

Table 1: The Impact of Home Court on the Four Factors

FactorRegularPlayoffsDiff
eFG1.34%1.54%0.20%
TO-0.40%-0.84%-0.43%
REB1.41%1.91%0.50%
FT1.12%2.12%1.00%

In the regular season, teams at home complete 1.34% more of their field goals, have a 0.4% lower turnover rate, grab 1.41% more offensive rebounds, and make 1.12% more free throws (per field goal attempt).  In the playoffs, all of these stats get better for the home team, with field goal percentage improving an additional 0.2%, turnover rate an additional 0.4%, rebounds 0.5% and free throws 1%.

To show how each of these stat differences ultimately affect scoring margin, I built a simple linear regression model that correlates four factors performance to the scoring margin.  Here are the coefficients from that model:

  • eFG% : 132.8 
  • TO : -121.9
  • REB : 39.62
  • FT : 25.98

By multiplying these coefficients through the stat differences in Table 1, we can see how each ultimately impacts scoring margin.

Table 2: Four Factors and the Impact on Scoring Margin

FactorRegularPlayoffsDiff
eFG1.782.040.26
TO0.491.020.53
REB0.560.760.20
FT0.290.550.26
Total3.134.371.25

In the regular season, more than half of home court advantage can be attributed to improvement in eFG% (1.78 out of the 3.13 points).  What is interesting is that the gap between the regular season and the playoffs is being driven most significantly by turnover rate, having twice as much impact as any of the other three factors.

I did a quick check of the stats and the improvement in turnover rate does not appear to be related to steals.  Unfortunately, the dataset I'm using doesn't allow me to dig deeper than that.

Referee Bias?

In his post, Ed Feng theorized that the larger effect of home court advantage was due to referee bias.  Crowds are larger and more vocal in the playoffs, and various studies have shown that referees, either consciously or sub-consciously, are affected by those crowds and skew their calls in favor of the home team.

It's hard to draw any firm conclusions, but these results seem somewhat consistent with that theory.  Turnover rate and free throw rate both double in impact in the playoffs, where field goal percentage and rebound percentage show smaller relative growth.  Of the four factors, I would expect turnover rate (which includes violations and offensive fouls) and free throw rate to be more susceptible to referee bias than the other two.

11 comments:

  1. Nice work. Since you have shown a significant gap between actual playoff home court advantage and how Vegas lines reflect it, have you looked at the ATS record of playoff teams playing at home?

    ReplyDelete
    Replies
    1. Thanks Ben. That actually came up in the comments to Ed Feng's article. Home teams are 53.6% against the spread in the playoffs going back to 2004.

      Ed followed up on this here.

      Delete
    2. Thank you! Do you have any interest in determining how much impact being ahead or behind by X games in a series have on HCA (via actual margins and Vegas lines)?

      BTW, Cantor has just released lines for the whole 2013 NFL Season. The post at predictionmachine.com seems to have the formatting most friendly for regression analyses: http://predictionmachine.com/NFL-lines-week1-week16

      Delete
    3. I've tried to tease out patterns in the spread as a function of how the series developed, but couldn't come to anything conclusive. There did seem to be evidence of teams in a 0-3 hole "giving up". It showed up both in the point spread and in the margin (but only about a point).

      Thanks for the link to the opening lines. I had taken a preliminary look but haven't had time to write it up. Cantor Gaming has the Broncos, 49ers, Seahawks, and Patriots all bunched up at the top (around a 4.5 GPF).

      Delete
    4. Yep, I noticed the same thing. I also separated each team's lines per home/away - it's interesting to discover that Vegas already thinks they've got HFA's for every team figured out. Not sure about the rest of teams, but I recall when I crunched point spread predictions a few years, NE had the worst HFA (or among the bottom 3). Pretty interesting that Vegas still sees things that way. Also, seems that Denver is expected to have the best HFA.

      Delete
    5. BTW, if I did my regression right, when averaging away + home ratings, Denver is a point ahead of NE. But if you look at team ratings only when they're visiting (in order to remove HFA bias), it's the way around - NE 1 pt above Denver. Make of it what you will.

      Delete
  2. BTW, when will you post the MLB ratings? :)

    ReplyDelete
    Replies
    1. Probably in another week or two. The model needs about a month of data because of starting pitching rotations (each team is more like 5 different teams from the standpoint of setting the lines).

      Delete
    2. The weighting method you use has a 'fat' past - last year's data for a particular pitcher is ostensibly worth about as much as a game four days ago - do you continue to use it, or go for a complete reset?

      Delete
    3. I do a complete reset. This season's rankings will be based solely on the 2013 season.

      Delete