polling

538 is on top of that. In order to control for favoritism affecting the results, they compared polls before and after to determine who actually improved as a result of the debate and who didn't. There's a whole bunch of information there and all sorts of interesting graphs, but the basic upshot...