How good is home court advantage? Some (rough) stats
Posted: Wed May 15, 2019 8:12 pmOne of the things I've noticed playing the game is that home court advantage can be a big swing in games, and that it's pretty easy to have a high win rate at home. But I wanted to know more, how much of an advantage is home court advantage? So I decided to find out using my rusty statistical skills and basic understanding of Microsoft Excel.
If we want to find out how good home court advantage is we need to control for as much other stuff as possible. So by modifying one of the database files from NCAAhoops' full college mod I averaged out:
-Coach skills
-Coach philosophies (everything set to the middle of the bar)
-Team prestige
-Team academics and facilities
-Stadium capacity
Of course the most important thing to even out is the players themselves, so I made a rough mock of a generated team and cloned it 351 times to make a veritable clone army (Note: some of the translation between database and game values is not 1:1 so this was a very unbalanced clone army). The main thing left that I couldn't control was strategy as that was entirely in the hands of the AI.
So now I just needed to sim some games. The methodology I opted for was as follows:
1. Start a new association with the modded rosters
2. Skip the summer and put AI control on everything
3. Sim the first day and write down the results of all non-tournament (read: neutral site) games
Thus I could generate a "random" sample of games. One flaw worth noting is these samples were not all the same size, which may be a source of error. Ultimately I ran this process 20 times simulating a grand total of 288 games. Out of those 288 games the home teams won 223, which is a win rate of about 77.43%. On average the away teams scored 108.47 points while the home teams averaged 117.14 points.
For some more detailed analysis here's some graphs:
A histogram of the home win rates for each of the samples: https://imgur.com/a/BH71Mz5
So the way to read this is that for example, for 5 of the samples or times I ran this process, the home teams had a win rate of .7-.799%
A histogram of the average points scored by the home teams for each of the samples: https://imgur.com/a/u50ijQ1
A histogram of the average points scored by the away teams for each of the samples: https://imgur.com/a/PefeUYV
A histogram of the average points difference for each of the samples: https://imgur.com/a/vsN84Ru
I hope this provides a slightly more insightful illustration of the impact home court advantage. Going forward I'm interested in tuning the clone army to make it more reasonable and I'm probably going to see what else I can experiment with in terms of numbers.
If we want to find out how good home court advantage is we need to control for as much other stuff as possible. So by modifying one of the database files from NCAAhoops' full college mod I averaged out:
-Coach skills
-Coach philosophies (everything set to the middle of the bar)
-Team prestige
-Team academics and facilities
-Stadium capacity
Of course the most important thing to even out is the players themselves, so I made a rough mock of a generated team and cloned it 351 times to make a veritable clone army (Note: some of the translation between database and game values is not 1:1 so this was a very unbalanced clone army). The main thing left that I couldn't control was strategy as that was entirely in the hands of the AI.
So now I just needed to sim some games. The methodology I opted for was as follows:
1. Start a new association with the modded rosters
2. Skip the summer and put AI control on everything
3. Sim the first day and write down the results of all non-tournament (read: neutral site) games
Thus I could generate a "random" sample of games. One flaw worth noting is these samples were not all the same size, which may be a source of error. Ultimately I ran this process 20 times simulating a grand total of 288 games. Out of those 288 games the home teams won 223, which is a win rate of about 77.43%. On average the away teams scored 108.47 points while the home teams averaged 117.14 points.
For some more detailed analysis here's some graphs:
A histogram of the home win rates for each of the samples: https://imgur.com/a/BH71Mz5
So the way to read this is that for example, for 5 of the samples or times I ran this process, the home teams had a win rate of .7-.799%
A histogram of the average points scored by the home teams for each of the samples: https://imgur.com/a/u50ijQ1
A histogram of the average points scored by the away teams for each of the samples: https://imgur.com/a/PefeUYV
A histogram of the average points difference for each of the samples: https://imgur.com/a/vsN84Ru
I hope this provides a slightly more insightful illustration of the impact home court advantage. Going forward I'm interested in tuning the clone army to make it more reasonable and I'm probably going to see what else I can experiment with in terms of numbers.