[eDebate] Does mutuality matter?

[NEIL BERCH]

I've begun to examine preference data from our recent Novice/JV Nationals.  
I'll report on some of the more standard measures later on, but I decided to 
first take a look at the effects of mutuality.

We used a 9 category system, with equal numbers of judges in each category 
(about 11).  Throughout preliminary rounds, we tried to maintain a maximum 
difference between teams in a debate of one preference category.  Indeed, 
about 60% of our rounds had perfect mutuality.  Only about 3-4% had a 
difference of more than one category (some in later rounds with teams 
eliminated, some in presets for which a team did not file a preference sheet 
in time for round 1; indeed, one preset was a 1/9 debate, where the 
unfortunate team did not file its sheet until after preset pairings were 
out).

Here's the interesting part:  there were 167 debates where one team 
preferred a judge more than the other team did.  The team that preferred the 
judge more won just 83 of those debates (49.7%).  If we exclude the 14 
debates where there was a difference of more than one category (8 of which 
were won by the team that preferred the judge more), then the team that 
preferred the judge slightly more won 75 out of 153 debates (49.0%).

What does this say?  If this is not a fluke result, it suggests that we 
overvalue mutuality.  Indeed, to me that would imply that we should program 
to prefer a 2/3 or 3/2 debate to a 3/3.  It might even imply that we should 
prefer a 1/3 or 3/1 to a 2/3 or 3/2.

I'd be interesting in hearing:
a)  whether others have looked at similar data for other tournaments, and, 
if so, what results they've gotten.
b)  what people make of the data.

--Neil Berch
West Virginia University