(CNN)**First things first:** The theme song of the week is Bob James' Angela, the theme song from the television show Taxi.

**Poll of the week:**A Marist College poll gives Democrat Tony Evers a 54% to 41% lead over Republican Gov. Scott Walker in the Wisconsin gubernatorial election.

This poll is more optimistic for Evers, the most likely Democratic nominee, than the only previous nonpartisan poll of the race. Walker led Evers by a 48% to 44% margin in a June Marquette University poll. The average of the two puts Evers ahead by 4.5 percentage points.

**What's the point:**Regardless of which Wisconsin poll you're more likely to believe, Walker's try for a third term is in trouble. At no point during his two previous reelection campaigns did Walker trail by anything close to 5 points.

So just how much trouble is Walker in?

To find out, I collected every governor poll I could find for campaigns dating back to 2006. In the more than 120 races I collected polls from, two trends emerged. The first is that large leads early in the campaign tend to revert towards the mean.

The second is that polls taken from January of the election year to mid-summer have tended to overestimate how well the party that controls the White House eventually does. This is the only time a Republican has controlled the White House since Walker first successfully ran for governor in 2010.

Looking at just the average poll, Walker's 4.5-point deficit would be forecasted to result in a 7-point loss in November.

Fortunately for Walker, it's not that simple. Early governor polls aren't always telling. There is a wide margin of error in forecasting.

Gov. Charlie Baker faced a double-digit deficit in early polls in the 2014 Massachusetts governor's race before going on to win. For Walker, this means that although he's projected to lose by 7 points, only looking at the polls suggests he still has a 25% chance of winning. That's as simple as flipping a coin twice and having it land on heads both times. Some might remember that now-President Donald Trump faced similar odds in 2016.

Moreover, race raters from the Cook Political Report and Inside Elections still rate the Wisconsin gubernatorial race as lean Republican despite Walker's poll numbers.

Now, race raters such as Cook and Inside Elections aren't forecasting what will happen in November. They are only rating where they see the races right now.

Even so, ratings at this point are usually quite telling of the November result. Looking at the governor elections since 2006, I compared where the race raters had each individual race at this point and what actually happened. Like with the polling, the White House party's candidates are overestimated in early race ratings. Despite this fade in poll standing for the president's party, races that are solid for one party and even races that are only likely for candidate at this point are almost always won by that candidate. Candidates such as Walker, whose race "leans" in his direction, are favored to win about 70% based upon how previous races like his broke down.

If you were to average the two probabilities (the polling and the race ratings), you would end up with Walker having a 48% chance of winning reelection.

This split between the polling and the race raters is not a Wisconsin specific phenomenon. We can look at all 30 of the 2018 gubernatorial races for which there is at least one poll not taken by one of the campaigns. Then we can see how often candidates in these positions are expected to win based upon the trends of campaigns since 2006. We can do the same for the race ratings in these 30 races (i.e. compare race ratings from prior campaigns at this point to the eventual result and forecast an outcome for this year's races).

I wouldn't read too much into the exact probabilities of each race (e.g. 98% vs. 95% or 70% vs. 75%) because we're only dealing with a few election cycles to calibrate the models. And there is at least one race (Connecticut), which I think is closer than either the public polling or race ratings at this point forecast for November. Still, the broad trends are clear.