Abstract
We consider the performance of the foremost academic House of Representatives forecasting models in the 2018 elections. In creating open-source implementations of these models, we outline key underlying assumptions. We find that although the results were unsurprising, they indicate a weakening of many traditional forecasting indicators.