I review and analyze a substantial range of the literature on the relative and absolute accuracy of juries and judges in their trial verdicts in a post at the Law Stack Exchange forum.
It is clear that judges and juries agree on the verdict in roughly 78% of criminal jury trials, that judges are much less likely to acquit defendants than juries do, (3% v. 19%+), and that there are a small but significant number of wrongful convictions by juries. (3.3% to 5% in death qualified capital murder-rape cases, and probably somewhat less in ordinary felony cases).
It is also clear that certain fact patterns and circumstances (many of which involve misconduct or error by law enforcement or prosecutors or defense counsel rather than judges or juries) are more likely to give rise to wrongful convictions than others.
It is not nearly so clear whether juries or judges are more accurate, although there is strong circumstantial evidence to suggests that judges wrongfully convict defendants more often than juries do, and that juries acquit people who are factually guilty more often than judges do (something that is sometimes legitimate if one believes that "jury nullification" is a legitimate practice).
I also briefly discuss arguments for and against juries that are unrelated to their accuracy.