November 2016 was a bracing reminder that presidential elections are won or lost through the Electoral College, a fact that Democrats may deplore but are powerless to change. The midterm election, therefore, offers some important clues about Democrats’ prospects of capturing at least 270 electoral votes in 2020.

Two years ago, Hillary Clinton prevailed in 20 states with a total of 232 electoral votes. A cursory inspection of the 2018 results reveals no evidence that Donald Trump can peel away any of these states. If anything, he has hardened the blue majority in all of them.

So where should Democrats look to win at least 38 additional electoral votes?

Iowa and Ohio, both potential swing states, do not appear very promising. In Iowa, the Republicans held the statehouse against an attractive Democratic challenger. In Ohio, Mike DeWine defeated Richard Cordray by a surprisingly large margin, confirming the pro-Republican shift in this most classic swing state. The presence of Senator Sherrod Brown, who cruised to reelection and got more votes than DeWine, on the 2020 Democratic ticket might stem this tide; it’s hard to see what else could.

Florida continues to break Democrats’ hearts. In 2014, Rick Scott beat Charlie Crist by a single percentage point—around 66,000 votes out of 5.9 million cast. In 2016, Donald Trump defeated Hillary Clinton by 1 percentage point—113,000 votes out of 9.5 million cast. In 2018, Rick DeSantis beat Andrew Gillum by about 33,700 votes—less than half a percentage point—out of 8.2 million. Compared to 2014, there was a massive voter mobilization of 2.3 million additional votes. Unfortunately for Democrats, it was almost perfectly symmetrical, leaving the balance unchanged.

Despite running a campaign that excited minorities and white liberals throughout Georgia, Stacey Abrams lost the governor’s race by 1.5 percent of the popular vote. Meanwhile, in Texas, Beto O’Rourke’s thrilling challenge to Ted Cruz fell short by more than 200,000 votes.

The results in Georgia and Texas augur a better future for Democrats in these states—someday. It might be as soon as 2020, but Democrats would be ill-advised to bet the next presidential election—and perhaps the future of U.S. constitutional democracy—on this best-case outcome.

Running on a pro-business agenda, Arizona’s Kyrsten Sinema narrowly defeated Martha McSally for the Senate seat Jeff Flake surrendered. It is doubtful that any Democrat could win the presidential nomination on such a platform. In any event, while it would be nice to have Arizona’s 11 electoral votes, they would fall well short of closing the gap.

This brings us to Pennsylvania and the upper Midwest, where Donald Trump won the presidency two years ago. This year, Democrats made a strong comeback. Pennsylvania’s Democratic governor won reelection by a wide margin, as did incumbent Senator Bob Casey Jr. Democrats regained the statehouse by a double-digit margin, and incumbent Senator Debbie Stabenow won reelection comfortably. In Wisconsin, Tony Evers defeated the two-term incumbent governor and conservative hero Scott Walker, while once-vulnerable Senator Tammy Baldwin turned back her Republican challenger by over 10 percentage points. And in Minnesota, which President Trump came close to winning in 2016, longtime Democratic Representative Tim Walz won the governorship by 11.5 points, while both Democratic senatorial candidates prevailed—incumbent Amy Klobuchar by an astounding 24 points.

Taken together, Pennsylvania’s 20 electoral votes, Michigan’s 16, and Wisconsin’s 10 represent Democrats’ best chance of regaining the White House. They are the cake; Florida is the frosting; Georgia, Texas, and Arizona are the final touches.

The bottom line is clear: As the presidential nomination contest heats up, Democrats should ask themselves a threshold question. Of the many candidates in the race, who has the best chance of prevailing in the states that are most likely to leave the Republican column?

Here are some clues. In Pennsylvania, Governor Tom Wolf was rewarded with a 17-point victory for giving his state four years of solid government. In Wisconsin, the victor over Scott Walker was a 67-year old state school superintendent who promised to focus on “solving problems, not picking fights.” In Michigan, veteran politician Gretchen Whitmer ran on a simple slogan: “Fix the damn roads.” Her opponent never had a chance.

Over the next two years, the Democratic Party will face a choice between its head and its heart.  The stakes could not be higher. The reelection of Donald Trump in 2020 may well place more pressure on our institutions than they can endure. Democrats’ task—a historic task—is to pick the candidate best positioned to prevent this outcome, and to support this candidate to the hilt, wherever he or she may be located within the party’s complex coalition.