Our simulations utilize the projections made in the House Forecast, specifically the projected margin of victory in every race, and the chance both parties have of winning. Each sim uses randomly generated numbers to create a different outcome. The difference between the projected result and the simulation is also affected by how much uncertainty we have in each election. For example, uncertainty is higher when we are further away from election day, and when there have been few polls published.

We don’t treat each race as a separate entity. Democrats or Republicans are likely to overperform expectations nationwide, and we also run through their performance with different demographic groups, including, for example, race and education. If a party exceeds projections in one district n Southern California, there's a good chance they will do the same in the neighboring district as well. Like in real life, sometimes a surprise scandal happens or a candidate runs a stunningly strong campaign and greatly outperforms expectations.

Keep in mind that sometimes the results will be a little bit wacky - especially when we have a long time until election day. This isn’t a sign of a broken forecast, as these outcomes are within the range of possibilities, even though they are unlikely.