Re-examining Homecourt Advantage

Posted on 11/07/2010 by

17



“I came out of Bataan and I shall return”-Douglas MacArthur on arriving in Australia in 1942

One of the interesting side effects of changing jobs is losing my laptop. I turned it in on Friday and since then I’ve felt like I’m missing an arm. I’ve been working on my wife’s laptop (which is a less than ideal scenario for all involved) and it’s kinda sucked. Yesterday’s piece took about four times as long as normal, while I adapted to my new writing/number crunching scenario reality. But I adapted and I persevered.

Let’s get back to work.

You Big Baby

Probability of Home team winning a game (Win %)

= (Projected Wins Home Team-Projected Wins Road Team)/82 +.606

=Win %: (Proj. Home Team Win% – Proj. Road Team Win%) +HCA(.606)

You’ve seen this before on this blog (Go here for the Basics). This is the simple equation I came up with for the home team winning  a single game (see here for detail). I’ve used this  for the evaluation of our predictions among other things (see here) and it does a pretty good job.

But you know by now that pretty good just doesn’t cut it.

We're all about excellence

When I first worked out Homecourt advantage I simply went looking for available data. The data set was the percentage from regular season games  from 1999 thru 2008 (in which the home team wins 60.6% of time) and as I said before this was good and worked fairly well. A number of my readers have made many suggestions. Here’s the list of things that I’m going to do here:

  • Add in the effect of rest days and back to backs.
  • Add in the effect of altitude
  • Use a more recent data set.

The first thing to do is build the data set. For that I turned to our old friend over at  dougstats .

Dude, thanks for the info

I downloaded every game for the last five years (and before you ask see here ). And I went about adding the rest days and altitude. I did that as follows:

  • For rest days, I chose three levels:
    • 0 for back to backs
    • 1 for day of rest between games
    • 2 for >2 days of rest
  • For Altitude, I also chose three levels:
    • Zero to low elevation (430′ or below): Boston, LA, Memphis, Miami, New Orleans, Sacramento,New York City,Orlando,Dallas,New Jersey,Toronto,Houston,Seattle(gone but not forgotten),Portland,Golden State and Washington (hi Ted!)
    • Some elevation (430′-1117′): Charlotte (Jordan with just some elevation seems wrong somehow), Milwaukee, Chicago, Cleveland, Atlanta, Indiana, Detroit, San Antonio, Phoenix, Minnesota
    • Nosebleed Country (5300-9400′): Denver and Utah

Now, If I simply work out the percentages here I won’t account for the strength of the opponent. Luckily, we already know the formula for two opponents on a neutral court:

Probability of  Team A beating Team B at a neutral site = (Raw Win Production Team A (Sum of ADJP48*MP/48 for all players on team) -Raw Win Production Team B (Sum of ADJP48*MP/48 for all players on team))/82 +.500

So if I use the total raw win numbers for each team for the season to calculate the win probability for each game if it was played at a neutral site, I can average it out (through the magic of large sample sizes) and figure the homecourt advantage in each scenario over playing at a neutral site. That  table looks like so:

If I add in .500:

Some interesting findings. Both, altitude and rest days affect the Homecourt advantage and they interact with one another. Average HCA is at 59.9%.  Altitude is directly proportional to HCA . Rest days are a little stranger.  Altitude directly interacts with rest. Denver and Utah kill teams at home if they have a rest edge but they get killed themselves if the other team is coming in with at least a two day rest edge.

Tomorrow, I give you power rankings.

About these ads
Posted in: Uncategorized