Well, yes and no. Having at least some common interests is always important to a lasting relationship. The problems I see with dating sites are:
- There is rarely a way to say "The person must meet all of the following criteria, or else don't bother showing me." For example, I would never date a smoker; even a small amount of cigarette smoke interacts very badly with my ability to sing.
Even if you can assign importance to a particular criterion, there is almost never any way for me to assign arbitrary weights to specific values for specific criteria.
For example, although I don't mind if my date drinks once in a while, I don't want to date an alcoholic. And even less extreme values are somewhat relevant. I don't personally drink, so I would probably not be a good match for someone who enjoys going out drinking. So a frequent drinker would have a negative weight, but someone who drinks occasionally or does not drink would have a zero weight, because the difference between those options is unimportant to me. And to some degree, it even matters what they drink. A girl who enjoys a glass of wine with dinner is more likely to be a good match than a girl who enjoys going down to the sports bar and having a couple of pints.
For another example, consider the religion checkboxes. Let's say you're Catholic and would prefer to marry within the Church. To do this, you would probably want to add a strong positive weight to Catholic and Orthodox Christian denominations, a lesser weight to various protestant denominations, and perhaps a negative weight to non-Christian religious backgrounds.
Many of the things I would like to include as part of my matching criteria are never available as options.
For example, Match.com allows you to specify that you like music, but such a general criterion makes no real distinction between a classical pianist and someone who likes listening to Miley Cyrus. Only one of those should be weighted positively, and I'll let you guess which.
And so on. In short, the options are too broad and nonspecific to be useful, and the weighting options are too limited, resulting in people seeming to like certain interests in one context and not in another context, not because they really do, but because certain aspects of those interests are positive and others are negative. To use your football example, a girl might like a guy who enjoys watching his favorite team at the local stadium every so often, but might not like a guy who spends every Sunday at home watching football all day. It isn't that the girl sometimes does and sometimes doesn't find guys who like football endearing, but rather that the way a guy acts on that interest and the extent to which he does so is crucial in determining whether the interest is or is not endearing to her. Without additional detail, the fact that the guy likes football is a very nearly useless data point, because it provides no actionable information.
Of course, a better solution would not use explicit weights entered by the user, but instead would ask you to rate a bunch of random people, and would then survey you to ask why you rated that person high or low, and would use your response to refine its matching criteria and weights. An ideal site would also perform detailed facial analysis and other image analysis so that "I did not find this person physically attractive" can turn into a set of weights based on how important specific physical criteria are. This is, of course, insanely complicated, which is why AFAIK nobody does it.
Oh, and I also want a really good grammar analysis tool built into a dating site. As a general rule, it should prefer to match people who have good grammar with other people who have good grammar. That said, I'd settle for a checkbox on the ratings questionnaire that lets me reject someone by saying, "This girl appears to be functionally illiterate." :-D