I'm outsourcing this prediction to Kevin Drum:
My sense—though I'd prefer actual data if anyone has collected it—is that secession votes usually follow a pattern: the leavers get an upward bump a few weeks before voting day, but stayers get a bump in the few days before voting day. A fair number of people flirt with the idea of leaving, but then get scared at the last minute and decide to vote for the status quo instead. Basically, in any secession referendum, I figure that Leave needs to be polling at 55 percent or higher to have a realistic chance of winning.
As of today, the polls are still tied, so my guess is that Brexit will fail on Thursday. If I'm right that about 5 percent of the leavers will get cold feet and change their minds, the final tally will be something like 53-47 percent in favor of remaining in the EU. We'll find out in a couple of days.