Will the Democrats retake the House in November? It’s too soon to tell, but the only poll on which I would rely is the generic congressional ballot at Rasmussen Reports (see this, for example). The polling data, which are behind a paywall, span April 2007 to May 2015 (when the poll was discontinued), and January 2018 (when the poll was resumed) to the present.
This graph compares the polling results to date with the actual nationwide vote shares compiled by House candidates in the general elections of 2008, 2010, 2010, and 2014:
Rasmussen advertises a 2-percentage-point margin of error, which is borne out by the results for the four elections. In fact, the generic congressional ballot was spot-on in 2010 and 2012, while the GOP under-performed slightly in 2008 (the year of the financial crisis) and over-performed slightly in 2014 (a mid-term referendum on Obama).
Let us say, for the sake of argument, that this year’s polls are spot-on. The latest poll gives the GOP 50 percent of the 2-party vote. How does that translate into House seats? Recent history is probably the best guide:
A 50-percent share of the vote would give the GOP about 51.5 percent of House seats; that is, the GOP would hold the House. In fact, two years ago the GOP won more than 55 percent of House seats with 50.5 percent of the two-party vote.