The poll may not be worthless. Using the Interactive Election Scorecard to compare Survey USA to a wide variety of polls, The Columbus Dispatch polls are one of the few that were more accurate in predicting actual results of an election than Survey USA. Hence, although the methodology is unusual, the Columbus Dispatch have a reasonable track record.
For comparisons see:
http://www.surveyusa.com/electiontrackrecord.html
I read through the .pdfs on the link and I'm not sure how well to trust the claims. I don't know the different comparison methodologies, but I do see that there are only 6 contests used to compare the Columbus Post Dispatch polling results. That seems like a weak statistical sample.
I also see that the University of New Hampshire fares about as well with 5 contests compared. I thought that thee methodology was also suspect, and the subject of posts here to that effect. This would lead me to suspect that the small sample of contests is a factor.
... I kinda like scientists as allies.
