I was surprised that the betting markets had the Democratic Party with a 19% chance of retaking Congress. To do so, they would have to flip nearly 50 seats!

In the modern era, the Republicans have pulled this off twice (1994, 2010), while the Democrats have only reached this number in 1974 (Watergate scandal). That’s because wave elections of the magnitude needed usually occur only in midterm years when the sitting president has an ebb in popularity and the Republican party is stronger voting patterns during midterms.

Perhaps an interesting way to look at the current situation is: right now the Republicans have their largest House majority since just before the stock market crash kicked off the Great Depression.

The challenge of flipping that seems insurmountable in an electorate as polarized as this one.

The press has mostly been focused on Donald Trump dragging down the ticket and Paul Ryan’s attempt to dance with him. So the articles focusing on the chances of the House flipping have actually decreased from previous years because the conventional wisdom is the only incumbents who actually are at risk are Republican U.S. senators in blue or battleground states like (Illinois, Wisconsin, New Hampshire), and (Pennsylvania, Florida, Ohio).

But the betting markets are capturing something the press isn’t: a wave is a wave is a wave, and this year is looking to be a Democratic wave due to raw demographics, being a presidential election year, and Donald Trump. This is compounded by the fact that strategically the Trump campaign is planning on funding and running its logistics by piggybacking off the Republican National Committee. This exacerbates the problem because: 1) it siphons funds and resources from congressional elections to the general; 2) battleground states are ones with large electoral vote counts and have no correlation with contested districts; and 3) the Republicans have increasingly depended financially on large donors via Citizen’s United and outsourced most of the voter outreach and research to third party organizations, so neither the money, nor the knowledge for the republicans is actually in the RNC.

By giving 5-to-1 odds, the markets are saying, “Yeah, it’s going to be a wave, but that’s not enough. But hey, we could always be wrong!” My instincts say it should be closer to 25% than 19% — akin to saying, “It’s possible the wave will be big enough, but it’s still far more likely it’ll take another election year for this wave to crest.”

That’s because I think betting markets have not (yet? ever?) correctly priced in the non-linear impact of gerrymandering. Right now everyone is saying that gerrymandering favors the Republicans. This is certainly true: through gerrymandering, Republican state houses in 2010 has acted as a deep red firewall against a Democratic demographic wave and districts are more partisan than ever because of it. Evidence certainly backs the power of gerrymandering up as, nationally, there were more votes for Democratic congressional candidates than Republican ones in 2012 and 2014, and yet the Republican party gained seats.

However, read that again, nationally there were more votes for Democratic congressional candidates than Republican ones in 2012 and 2014. While this shows success of gerrymandering, that success is a recipe for long term failure, and that gerrymandering had to have been larger to hold back that failure.

States like California are forced by law to district in a non-partisan manner. In California, the Democratic Party has every state-wide office and a supermajority in the state senate and congress. That’s what gerrymandering is holding back.

The traditional way of gerrymandering would be to pack heavily Democratic districts together. This is especially easy to do because, in the modern era, their strength has been around cities which are already geographically compact. However, that trick was maxed out after 2002 when suburbs/exurbs became both population-dominant and purple. The 2012 approach to gerrymandering became much more sophisticated because the overall voter demographics haven’t been favoring the Republicans since the 1980s.

To understand the new approach used by the Republicans, known as REDMAP, consider a state like Texas which gained four seats in Congress for 2012. Because Latinos and cities account for all of that growth and it was a presidential election year, one would expect a congressional wave for the Democratic party that year — about seven seats gained by them. After a hard-fought legal battle, the districts were gerrymandered to net the Republicans only three and the Democrats only one. Not only that, the districts are so partisan that there was only a single congressional election within 10 points in the state in 2012 (TX-23). The markets are “pricing in” the power of REDMAP to gain Republican seats in the face of adding a huge number of Democratic-leaning voters to the state’s makeup.

However, in order to do have this sort of success, REDMAP couldn’t just create TX-35 (connects San Antonio to Austin) or TX-33 (connects liberal parts of Ft. Worth to Dallas) to build ultra-concentrated Democratic districts, it also would have to transform TX-2 which wraps around TX-18 and TX-29. The latter two concentrate majority-minority districts, but the first was gerrymandered in 2002 to create a safe Republican seat out of a Democratic one and gerrymandered again in 2012 to absorb number of Democratic-leaning people so that the state could create TX-36 as a safe Republican district, while maintaining TX-2 as a safe Republican one.

How? Basically the census increase must mean that REDMAP would have to weaken Republican districts by adding Democratic-leaning populations to them. However, it ensures it is “safe” by using advanced voting models that add large local cohorts who would vote Democratic but don’t vote to these Republican districts. REDMAP’s modeling “knows” these people don’t vote, but demographically it also knows they’d vote for the Democratic Party heavily if they did. Think of it like an advanced version of building a prison in a red district and counting non-voting prisoners from cities as “residents” to pad your district’s population numbers.

Now consider this election year. The last Democratic congressional wave was 2006 mid-term election. This implies the shelf life of gerrymandering in the United States is about six years after the last census—after that the models start breaking down because people move or whatever. We are six years after the last census! Furthermore, REDMAP’s Democratic “non-voters” tend to be Hispanics, and Trump has activated that very same cohort after he called them rapists and decided to build a wall with Mexico.

What happens when the statistical software is modeling based on bad assumptions and blind belief in them have caused people to tax it past its breaking point?

I’m not willing to say that TX-2 will flip — a 37 point swing is near impossible — but there is a huge Democratic wave sitting behind that Great Red(MAP) Congressional Firewall, and the top of the ticket is actively ripping out parts near the cracks of its very weak foundation.