Tuesday, March 27, 2007
Is there really less depth at the top in the women's game?
The general consensus seems to be that there is a greater dropoff between the best teams in the women's game and the middle tier teams, compared to the dropoff in the open division. I am not really going to try to dispute that viewpoint here. What I am going to do is dispute that the scores in open games, versus those in women's games, actually suggest that.
Take this example:
In an open game, team A commits 14% fewer turnovers than team B over the course of the game.
In a women's game, team A commits 16% fewer turnovers than team B over the course of the game.
Based on this information alone, would you conclude that the talent gap between teams A and B is much smaller on the open side than the women's side? I doubt it. It's more likely you would conclude that the teams are similarly evenly matched and that the 2% difference in the gap is too small to draw conclusions from.
Now, here comes the twist:
In the open game, team A committed 6 turnovers to team B's 7, and won by 2.
In the women's game, team A committed 26 turnovers to team B's 31, and won by 6.
Based on this information, you would be tempted to conclude that the open game was a tight match between equals, while the women's game was a blowout between two teams that were playing on different levels. But this is really a statistical artifact of the different turnover rates in the women's game versus the open game. Because there are fewer turnovers in open ultimate, there is less opportunity for teams to pile up a high margin of victory based on a slight edge in scoring efficiency.
Again, I'm not arguing that there is not a larger dropoff in the women's game than in the open game. What I'm arguing is that the scoring gaps alone don't suggest that. They are just a reflection of the different turnover rates in women's and open. What the above does suggest is that the results in women's are less random than open, since a small efficiency gap has more chance to snowball into a safe lead.
Some may have already figured this out, but teams A and B are the winners and losers of the 2006 UPA club finals.
Take this example:
In an open game, team A commits 14% fewer turnovers than team B over the course of the game.
In a women's game, team A commits 16% fewer turnovers than team B over the course of the game.
Based on this information alone, would you conclude that the talent gap between teams A and B is much smaller on the open side than the women's side? I doubt it. It's more likely you would conclude that the teams are similarly evenly matched and that the 2% difference in the gap is too small to draw conclusions from.
Now, here comes the twist:
In the open game, team A committed 6 turnovers to team B's 7, and won by 2.
In the women's game, team A committed 26 turnovers to team B's 31, and won by 6.
Based on this information, you would be tempted to conclude that the open game was a tight match between equals, while the women's game was a blowout between two teams that were playing on different levels. But this is really a statistical artifact of the different turnover rates in the women's game versus the open game. Because there are fewer turnovers in open ultimate, there is less opportunity for teams to pile up a high margin of victory based on a slight edge in scoring efficiency.
Again, I'm not arguing that there is not a larger dropoff in the women's game than in the open game. What I'm arguing is that the scoring gaps alone don't suggest that. They are just a reflection of the different turnover rates in women's and open. What the above does suggest is that the results in women's are less random than open, since a small efficiency gap has more chance to snowball into a safe lead.
Some may have already figured this out, but teams A and B are the winners and losers of the 2006 UPA club finals.
# posted by Tarr : 3/27/2007 01:07:00 PM
Comments:
<< Home
Let's take a different look at the same data. Compare scoring percentages:
Open: 71% vs 65% (ratio 1.10)
Women: 37% vs 22.5% (ratio 1.63)
But then you would counter with "let's look at turnover percentage
Open: 29% vs 35% (ratio 1.225)
Women: 63% vs 77.5% (ratio 1.222)"
I think my point is that the data also does not necessarily provide a counterpoint.
Open: 71% vs 65% (ratio 1.10)
Women: 37% vs 22.5% (ratio 1.63)
But then you would counter with "let's look at turnover percentage
Open: 29% vs 35% (ratio 1.225)
Women: 63% vs 77.5% (ratio 1.222)"
I think my point is that the data also does not necessarily provide a counterpoint.
Well, my point is just that the greater scoring gap doesn't necessarily suggest a greater drop-off by itself. Showing there are many possible interpretations of the data doesn't really change that.
I think what you've shown is that rather than looking at the ratios of the turnover rates, I should be looking at the difference in scoring efficiency.
Open - 6% gap
Women - 14.5% gap
hmmm.... well, that is noticably larger. Maybe Fury really did spank Riot.
Looking back on the 2005 finals:
Open:
Furious 15/46 = 33%
Sockeye 13/45 = 29%
4% differential
Women:
Riot 15/50 = 30%
Backhoe 4/50 = 8%
22% differential
Interesting - so, even in a high turnover affair, the open final stays tight. That women's final was obviously a blowout, so the stats backing that up are unsurprising.
I guess trying to judge this stuff from a sample as small as the results of four club finals involving five total teams is really pretty hopeless.
How about College? I just randomly grabbed the 2004 college issue.
Open semi #1
CU 15/22 = 68%
Stanford 10/21 = 48%
20% differential
Open semi #2
Cal 15/28 = 54%
Brown 13/27 = 48%
5% differential
Open final
CU 15/37 = 41%
Cal 7/36 = 19%
21% differential
Women's semi #1
UC-Davis 15/43 = 35%
UCSD 13/43 = 30%
5% differential
Women's semi #2
Carleton 15/49 = 31%
Stanford 13/50 = 26%
5% differential
Women's final
UC-Davis 15/55 = 27%
Carleton 12/54 = 22%
5% differential
Other than saying "Colorado was really good that year", I have no conclusions.
I think what you've shown is that rather than looking at the ratios of the turnover rates, I should be looking at the difference in scoring efficiency.
Open - 6% gap
Women - 14.5% gap
hmmm.... well, that is noticably larger. Maybe Fury really did spank Riot.
Looking back on the 2005 finals:
Open:
Furious 15/46 = 33%
Sockeye 13/45 = 29%
4% differential
Women:
Riot 15/50 = 30%
Backhoe 4/50 = 8%
22% differential
Interesting - so, even in a high turnover affair, the open final stays tight. That women's final was obviously a blowout, so the stats backing that up are unsurprising.
I guess trying to judge this stuff from a sample as small as the results of four club finals involving five total teams is really pretty hopeless.
How about College? I just randomly grabbed the 2004 college issue.
Open semi #1
CU 15/22 = 68%
Stanford 10/21 = 48%
20% differential
Open semi #2
Cal 15/28 = 54%
Brown 13/27 = 48%
5% differential
Open final
CU 15/37 = 41%
Cal 7/36 = 19%
21% differential
Women's semi #1
UC-Davis 15/43 = 35%
UCSD 13/43 = 30%
5% differential
Women's semi #2
Carleton 15/49 = 31%
Stanford 13/50 = 26%
5% differential
Women's final
UC-Davis 15/55 = 27%
Carleton 12/54 = 22%
5% differential
Other than saying "Colorado was really good that year", I have no conclusions.
I was getting at (in a very slow way) that the answer lies in assessing how hard it would be for the losing team to improve enough to tie. The data would look different still if we looked at completion percentage or adjusted completion percentage. And there is the additional confounder that as you get closer to 100% efficiency, getting an extra point of efficiency is more difficult; thus, a 6% gap at 65% is greater than a 6% gap at 20%. But then again, a 6% gap at 20% represents a much bigger fraction of the times you score, so maybe not.
So, I don't know.
So, I don't know.
Also, I take issue with you comparing the finals as the drop off point (between elite and mid-tier teams). Fury and Riot were easily the best teams last year, and only lost to each other. Look at the semis: 15-3 and 15-2. I think the drop off becomes obvious when you look at quarters and semis over the past few years, and not just finals.
Certainly a valid point, Neva, but I don't have turnover numbers for any elite games except a few selected semis and finals. Riot/Fury is a particularly bad example, since the general consensus is that the drop-off is right after that.
If this post and the ensuing discussion proves anything, it's that it's hard to draw reliable conclusions from the scores alone. As I said at the outset, I'm not trying to argue that there is NOT a greater drop-off in the women's game. I'm just arguing that I'm not convinced there IS a greater drop-off based solely on the game scores.
Post a Comment
If this post and the ensuing discussion proves anything, it's that it's hard to draw reliable conclusions from the scores alone. As I said at the outset, I'm not trying to argue that there is NOT a greater drop-off in the women's game. I'm just arguing that I'm not convinced there IS a greater drop-off based solely on the game scores.
<< Home
