When the selection committee sits down next March to select the field of 68 and gives them their respective seeds, they will rely on their strength of schedule, their RPI, and sadly some internal politics. But one metric that won't be used is the coaches poll (or the AP poll for that matter).
Recent NCAA selections are riddled with the bracket diverging from the polls (oddly enough the first one that springs to mind is 2002 Gonzaga, who everyone thought would get a 2 seed based on ending the year in the top 10 of the polls, but ended up with a 6).
I suppose we could expand the question to: why does any sport except FBS college football have a poll? Obviously in college football those polls are used to determine the national title game participants (which is a separate issue), but no other major college sport has this issue, as the tournament fields are determined by a committee (except for hockey which uses a BCS like system called the pairwise ratings, but the polls aren't a factor in determining the ratings).
The polls seemingly are just an exercise in telling the public who has won a lot of games. Illinois is currently ranked 10th, but beat Hawaii by one in overtime, beat Garner Webb by one at home, and struggled the whole game against Georgia Tech at home. They're undefeated, but they are far closer to the 30th best team in the country where KenPom has them than the 10th best. I don't need coaches (or, more likely, SIDs) or AP writers, most of whom probably watch fewer college basketball games than I do, to tell me who has won a lot of games. I can look win-loss records up fairly easily thanks to this new thing called the internet.
And are the polls supposed to tell us who the best team is, or just be a tool that parrots results? If the former, why is Arizona ranked ahead of Florida? Florida going to Tucson and losing by one tells me that they probably are the better team than the Wildcats, head to head record be damned. But if an AP writer were to rank the Gators ahead of the Wildcats on the heels of that loss he or she would likely be ridiculed. Again, I don't need the polls to tell me who won the basketball game. I watched it. And if I didn't watch it I could just as easily look up the score. I would hope that your average coaches or AP poll voter has the necessary brain power to realize that Florida shouldn't be punished for having to go cross country to play that game, and instead try to figure out who would win on a neutral floor.
I guess the answer (probably) is that it gives voters something to do and it gives fans a reason to debate the rankings and who is overrated and who is underrated. And maybe drunk college kids would have a tougher time remembering when it was acceptable to rush the court after a big win if there were no 1 next to their opponent's name on the scoreboard. But I bet that most college basketball fans can realize that Kansas is good, Creighton is OK and Georgia is terrible. We don't need little numbers for that.
I haven't paid attention to the polls in probably five years now. This isn't the 60s anymore: a basketball game played in California isn't a foreign concept to someone on the East Coast. Rather than dealing with the obvious flaws in the polls (eg. any kind of win is rewarded over any kind of loss. Ohio State could have won by one on a desperation three at the buzzer over Winthrop and I guarantee they would be ranked higher than Florida in this week's poll. That leads to the other obvious problem of when a team loses they are guaranteed to drop down 4-6 spots pretty much no matter what. And the team with the later loss will usually be ranked behind the team with the earlier loss).
The polls don't tell us much in terms of who the better team is, and they are of no consequence when seeding the NCAA tournament. They might give us a ballpark idea of which teams are good, but they deceive as much as they inform. So let's get rid of them. The college basketball polls are frequently inaccurate and have no real meaning attached to them. So why are we still using them?