[eDebate] Topic side bias

A Numbers Game edebate edebate
Wed Sep 16 11:06:47 CDT 2009


Every summer there is discussion about the side bias of prospective
controversy areas and resolutions. There has also been some number
crunching and speculation about whether tournament results show any
side bias, including:

http://www.ndtceda.com/pipermail/edebate/2007-March/070410.html
http://www.ndtceda.com/pipermail/edebate/1997-October/000956.html
http://www.ndtceda.com/pipermail/edebate/2003-June/049812.html
http://www.ndtceda.com/pipermail/edebate/2003-June/049838.html

Using the data from DebateResults.com, we can see what side bias past
resolutions have exhibited:

http://code.google.com/p/anumbersgame/wiki/TopicSideBias

Under the Bradley-Terry model, the energy, China, and security
guarantee topics had small, yet highly statistically significant (p <
.001), negative bias. The Europe, courts, and agriculture topics show
no statistically significant bias either way.

Though the side bias is statistically significant for three of the six
years in the debateresults.com data, statistical significance does not
imply real world significance. The largest side bias under the BT
model in the past six years was on the energy topic. Under the BT
model for that year, in an otherwise evenly matched round, the neg had
a 52.91% chance of winning. The effect, though small, is statistically
significant because of the large number of ballots in the database.

Thanks again to John Bruschke and DebateResults.com, as well as those
who reviewed the wiki page and offered suggestions.



More information about the Mailman mailing list