During this past season, I looked into how adding home ice advantage would change my model for ranking teams. To summarize: there is a home ice advantage in DI women's hockey, but it does little to change the top 8 ranked teams using my model.
I did some more work in preparation for a talk I am giving at the 2009 Joint Statistical Meetings in DC next week, and I wanted to share it and ask for your thoughts.
Using the past 4 seasons of data and looking at just wins, losses, and ties (no game scores), if I assume that HIA (home ice advantage) is the same for every team the estimate of HIA is 0.167. What does this mean in context? If two equal teams are playing, the probability of the home team winning jumps from 41.9% to 48.5%. The probability of a tie is also calculated, so that is why both of these probabilities are less than 50%.
Roughly speaking, if two equal teams played 50 games, a team would win (on average) 21 at a neutral site and 24 if they played at home. Another way to think about it is that playing half of the 34 games in a season at home earns a team 1.1 more wins that if all those games were played at a neutral site.
To investigate things further, I tried to estimate each teams individual home ice advantage. I will skip the details (you can ask if you really want to know), but here are the results for each team, ranked by best home ice advantage, for the last four seasons of games.
Putting these number in context:
When Clarkson plays at home vs. neutral site, the probability of a win increases from 41.9% to 60.2% (assuming equal teams), a gain of 3.1 wins per season (on average).
When Princeton plays at home vs. neutral site, the probability of a win DECREASES from 41.9% to 26.9%. In other words, Princeton losses 2.6 wins a year playing a home vs. neutral ice (on average).
Now, you may argue with the magnitude of these individual home ice advantages, but what I am interested in is the rankings.
Clarkson (very northern New York) and Maine (middle of Maine) have campus locations that can lead to long and difficult travel in winter, while UNH has Olympic sized ice. So these three teams being on top make sense.
It is the lowered ranked teams that I have trouble explaining. Quinnipiac moved to a new rink during this time frame, so we can ignore them for the moment. Robert Morris plays off campus, which explains the lack of home ice advantage.
But what about Princeton and Cornell? Why don't they perform well at home? I have ideas, but I would love to hear your thoughts.
I did some more work in preparation for a talk I am giving at the 2009 Joint Statistical Meetings in DC next week, and I wanted to share it and ask for your thoughts.
Using the past 4 seasons of data and looking at just wins, losses, and ties (no game scores), if I assume that HIA (home ice advantage) is the same for every team the estimate of HIA is 0.167. What does this mean in context? If two equal teams are playing, the probability of the home team winning jumps from 41.9% to 48.5%. The probability of a tie is also calculated, so that is why both of these probabilities are less than 50%.
Roughly speaking, if two equal teams played 50 games, a team would win (on average) 21 at a neutral site and 24 if they played at home. Another way to think about it is that playing half of the 34 games in a season at home earns a team 1.1 more wins that if all those games were played at a neutral site.
To investigate things further, I tried to estimate each teams individual home ice advantage. I will skip the details (you can ask if you really want to know), but here are the results for each team, ranked by best home ice advantage, for the last four seasons of games.
Code:
Maine 0.4635
Clarkson 0.4603
New Hampshire 0.4587
St. Cloud State 0.4352
Syracuse 0.4319
Connecticut 0.3991
Colgate 0.3950
Harvard 0.3615
Wayne State 0.3588
Providence 0.3434
Boston 0.3434
St. Lawrence 0.3214
Minnesota State 0.3071
Rensselaer 0.2809
Wisconsin 0.2743
Bemidji State 0.2679
Mercyhurst 0.1826
Boston College 0.1524
Yale 0.1244
Ohio State 0.1011
Dartmouth 0.0723
Union 0.0715
UMD 0.0581
Brown 0.0567
Niagara 0.0535
Minnesota 0.0531
North Dakota 0.0091
Northeastern -0.0716
Vermont -0.0889
Robert Morris -0.1460
Quinnipiac -0.1834
Cornell -0.3664
Princeton -0.4101Putting these number in context:
When Clarkson plays at home vs. neutral site, the probability of a win increases from 41.9% to 60.2% (assuming equal teams), a gain of 3.1 wins per season (on average).
When Princeton plays at home vs. neutral site, the probability of a win DECREASES from 41.9% to 26.9%. In other words, Princeton losses 2.6 wins a year playing a home vs. neutral ice (on average).
Now, you may argue with the magnitude of these individual home ice advantages, but what I am interested in is the rankings.
Clarkson (very northern New York) and Maine (middle of Maine) have campus locations that can lead to long and difficult travel in winter, while UNH has Olympic sized ice. So these three teams being on top make sense.
It is the lowered ranked teams that I have trouble explaining. Quinnipiac moved to a new rink during this time frame, so we can ignore them for the moment. Robert Morris plays off campus, which explains the lack of home ice advantage.
But what about Princeton and Cornell? Why don't they perform well at home? I have ideas, but I would love to hear your thoughts.
