Repeat patent plaintiffs--those who sue eight or more times on the same patents--have a disproportionate effect on the patent system. They are responsible for a sizeable fraction of all patent lawsuits. Their patents should be among the strongest, according to all economic measures of patent quality. And logic suggests that repeat patent plaintiffs should be risk averse, settling more of their cases and taking only the very best to trial to avoid having their patents invalidated. In this paper, we test those hypotheses. We find that repeat patent plaintiffs are somewhat more likely to settle their cases. But, to our surprise, we find that when they do go to trial or judgment, overwhelmingly they lose. This result seems to be driven by two parallel findings: both software patents and patents owned by non-practicing entities (so-called "patent trolls") fare extremely poorly in court. We offer some possible explanations for why a group of apparently weak patents nonetheless have so much influence over the patent system, and some preliminary thoughts about how these findings should shape the patent reform debate.