Well, first of all, that's not the debate winner, and it's not just among decideds. It's the gun-to-your-head question ("if you
had to choose") about who they support, not who won the debate.
Don't get too excited about those numbers, either. It's a n=1503 poll, so that means the undecided/third-party sample is 120. The margin of error on 120 people in a national poll is +/-9%. Because the MoE applies to both candidate values, the margin of error on the percentage margin is +-/-18%. Trump is barely up on that question beyond the margin of error, and seems to have a lot of potential soft supporters in this poll. However, note that the race was Clinton +4 in both the pre-debate and post-debate samples. So, as good news goes, this is pretty weak.
This poll shows unenthusiastic potential for Trump, which is one of his better results lately, but I wouldn't get too excited about it...and the poll showed the same potential before the debate, so we know the baseline here might be a bit too Trump-friendly. This is what passes as good news for Trump these days, but only because the pickings are slim.