On Friday morning, Matt Drudge was feeling celebratory on Twitter.

(Drudge, no dummy at the Internet, purges his tweets regularly, which is why we've included a screenshot.)

The daily Los Angeles Times-USC Dornsife tracking poll, Drudge pointed out, showed Donald Trump up 5.6 points on Sep. 30, an increase in the narrower 3.9-point lead he enjoyed the day of the debate. What's more, the poll shows Trump gaining a point and Clinton losing ground since then. Good news!

We regularly warn people against taking one poll as representative of the entire race, and the LA Times-USC poll is a good example why. The poll is consistently more favorable to Trump than are other polls, as you can see by comparing the margin between the candidate in the RealClearPolitics polling average with the daily results from the Times-USC.

That doesn't mean the Times poll is wrong, and as you can see it tends to rise and fall in concert with the overall average. But as we explained last month, it uses an unusual methodology which may play a role in why its different. Survey director Jill Darling explained to The Post by email earlier this month that the results are also weighted to include shifts in the extent to which people say they're likely to vote. "It is not a percentage decline in the number of people saying they will vote for a candidate, as you have in a traditional poll," she wrote, "it is a percentage decline in everyone’s average likelihood of voting for that candidate, as expressed over the past 7 days."

We got other polls on Friday, too: Traditional polls from battleground states. And those suggest a different shift in the race.

New polls from Suffolk University in Nevada, Mason-Dixon in Florida, the Detroit News in Michigan and WBUR in New Hampshire all show Hillary Clinton improving relative to the RealClearPolitics polling average on the day of the debate. (All the polls were conducted after the debate.) The margins of that improvement differ, but in each state the result moved somewhat in her favor.

Clinton plus-six in Nevada, up over eight points from the average. Up four on Trump in Florida, up 3.5 over the average. Leading Trump by seven in Michigan and New Hampshire, increases of about two points over the average in each.

The most important result of those four polls was the one in Nevada. There, Suffolk gave Clinton her biggest lead in the state of any survey so far this cycle. What's more, we can compare this new survey with one Suffolk conducted last month, when Clinton was faring much better nationally. Over the past month, even after she's seen support dip overall, Clinton gained four points in a head-to-head contest against Trump.

Among demographic groups, the biggest shift in the margin toward Clinton was with older voters, Republicans and women. The Republican shift was largely because Trump went from getting 89 percent of the vote from members of his own party to getting 78 percent. (A dip in Republican support is also what drove his bad August.) The biggest shift toward Trump was among those aged 35 to 49.

Again, though: Be cautious reading too much into one poll. A six-point Clinton lead in Nevada is an outlier at the moment, just as the Times-USC poll is an outlier nationally. What we'd expect to see with a dominant debate win (as Clinton enjoyed on Monday) is a shift in the national polls toward the winner and, as state polls come in, a shift in those to the winner as well. We'd expect to see, in other words, something like what we see in these four state polls and not what we see in the Times-USC poll.

Whether or not Clinton changes her strategy based on the latter, as Mr. Drudge recommends, remains to be seen. Whether or not Trump changes his based on those state polls seems at this point somehow less likely.