MSNBC’s Lawrence O’Donnell makes a good point about the extremely low probability that Donald Trump will get a fair trial in Washington, D.C. He doesn’t use the term “fair trial” and he doesn’t seem to mind, but that is the point he’s making.

But he overstates dramatically. After pointing out that of 700,000 people in D.C., only 676 voters voted for Donald Trump, O’Donnell states:

The odds of one of the 676 Republicans who did vote for Donald Trump in Washington, D.C. yesterday ending up on his jury there are worse than your odds with any lottery ticket you could buy anywhere.

Do you see the problem? There are a few. The big one, of course, is that 676 out of 700,000 is approximately 1 in 1,000. Let’s make it exact. It’s 1 in 1,036. There are many, many lotteries with worse odds. In fact, all of them have worse odds, if we’re talking about the lotteries with prizes greater than, to pick a number not at random, $1,000 in return for a $1 ticket.

There are other more-minor problems. I’ll mention two.

First, there are just shy of 700,000 people living in D.C. Probably, no more than 500,000 of them, and probably even fewer, are eligible to serve on a jury. So that makes Trump’s odds better than O’Donnell says, although still grim. Let’s pick the 500,000 number. So Trump’s odds are 1 in 740.

Second, I don’t think O’Donnell understands the idea of people acting on principle. I can certainly imagine a Trump voter voting to convict Trump once the judge has laid out the instructions to the jury. That would make Trump’s odds worse than 1 in 740, although, of course, it doesn’t rescue O’Donnell’s absurd claim. But I can also imagine that there are principled people among Democratic jurors who would, following a judge’s instructions to the jury, vote to find Trump not guilty. That would, of course, make Trump’s odds somewhat better.

The bottom line: Trump cannot get a fair trial, but O’Donnell badly exaggerated how unfair it would be. It’s too bad O’Donnell didn’t consult his colleague Rachel Maddow, who I hear is smart.