Virginia’s upcoming gubernatorial election has attracted a lot of attention for two primary reasons.

The first is there isn’t a whole lot else going on.

The second is more important. The race serves as a temperature reading on the national political mood 10 months into the era of President Trump. It’s a state the Democrats should win: The sitting governor is a Democrat, and Hillary Clinton won it last November. The Democrats have so far been unable to get a victory in a contested statewide or federal race against a Republican in 2017, often because the Republicans had a home-field advantage. In Virginia, the advantage is the Democrats’.

So a new poll from Hampton University released on Wednesday was a shocker. It showed the Republican candidate, Ed Gillespie, leading Democrat Ralph Northam by 8 points.

For perspective, that’s only the second time Gillespie’s led in a poll in recent weeks, and it’s the only result that’s outside the margin of error to Gillespie’s benefit.

In other words, reading it quickly: Very bad news for Northam.

One should not read polls quickly. In this case, there’s a critical caveat. Instead of asking respondents who didn’t indicate a choice between Northam and Gillespie who they preferred, those respondents were simply listed as “don’t know.” The result is “don’t know” ended up getting more than a quarter of the vote.

Why does that matter? Because it means a quarter of the possible electorate which will weigh in on the race — and would probably weigh in on behalf of the candidate who they’d tell the pollsters they were leaning toward — isn’t counted. Maybe Gillespie’s got a core of support but Northam’s got a ton of soft support, and 25 of those percentage points would go to him. Without asking, we don’t know. So we have instead this weird three-way race.

Hampton’s done this before, and its polling has been similarly questionable as a result.

Before the 2016 presidential election, Hampton polled twice in a week. The first time, Trump led by 3 points — literally the only poll that had him leading in the state at any point last year. The second poll, conducted right before the vote, had Clinton up 4. She won by about 5.

Maybe the vote shifted 7 points in a week and Hampton was the only pollster to notice it. Or maybe not.

Before the party primaries in the state earlier this year, Hampton had Gillespie winning by a mile and Northam losing. Gillespie ended up winning in a squeaker; Northam won easily. Particularly in an off-year primary featuring candidates with much lower name recognition than those running for president, all of those “don’t know”s would have had an effect.

Generally, we advocate for looking at polling averages because outlier polls — like this one — are considered in the broader context of the race. (The RealClearPolitics average has Northam leading by 4 points, even including this poll.) Outliers may be right! It’s generally wiser to consider them as outliers, particularly when recent history hasn’t been terribly kind to the pollster.

Scott Clement contributed to this report.