A few weeks back I took part in an experiment on the Kindleboards. D.A. Boulter volunteered to purchase up to 20 books with only one criteria: that the title hadn't sold anything on Amazon that month. I volunteered The Docks. The results are now in (full results here).
In total he picked up 15 books, and has read the first seven. In summary all were enjoyable reads, and only one was marked down for proofing and editing issues. He comments that
"‘Mountains of dreck’ may exist out there but, within this small sample of non-selling books, I didn’t find it."
It is a good point. I've heard people refer to self-publishing as "putting the slushpile online" but I have also encountered raw unedited slush. There are significant differences e.g. a surprising number of slushpile manuscripts are submitted with typos or gramatical errors (probably about 10% going by this post) because new authors assume - and sometimes state outright - that it is the job of the agent or editor to correct these errors. Most self-published authors I've encountered either self-edit or get a freelancer in to edit or proofread and acknowledge it is their responsibility. Authors who submit handwritten manuscripts to the slushpile are unlikely to e-publish.
Also, as these are ebooks, authors can withdraw and republish them if there is a major issue. This "second chance" could be seen as a trade-off for the lack of the initial editing and support of a traditional publishing house.
Now, the sample is slightly skewed: it drew from the authors on Kindleboards who are serious about their writing and presumably want to develop it enough that they have looked up a community and started discussing technique and writing with other authors. Second it asked the authors to submit their books. This narrows it down to authors who are prepared to admit to low sales, but are still confident that their book is good. On the other hand, by only taking books which had no sales that month, it was selecting against the bestsellers or higher ranked books which would be expected to be higher quality.
It may not be a purely scientific study, but it does show that there are some gems out there and that the odds of getting a decent read from a fairly random sampling are quite high (six in seven so far). User reviews are another guide: of the seven books that had reviews none had an average lower than four at the time of writing. Given that readers can view samples online it should be fairly easy to eliminate books that don't appeal.
So is it worth taking a chance on an indie book? So far it seems this would indicate yes.
I will end this by admitting my own bias here: he liked The Docks.