Before the election many of us were curious as to how close the polling/predicting might turn out to be. I was following 538 due to them being one of the most accurate predictors of the 2016 election (538 was off in 2016, but not as much as most others).
They were pretty close this time around.
Background: 538 uses three different models that tend to put out slightly different results. They show their results as probabilities, their "predictions" are the results they show as being the most probable.
So
for the House, they showed the most probable results as being that the Dems would pick up 38, 39, or 36 seats (lite, classic, deluxe probability models), with an 85% probability that the Dems would pick up enough seats to gain control of the house. So the Dems in the house did better than what 538 considered to be most probable, but it was still pretty close.
In the Senate, 538 showed the most probable outlooks as the Republicans picking up 0.7, 0.5, and 0.5 seats - somewhat problematic in that there is no such thing as half a Senate seat. At any rate, the GOP pick up of two seats outperformed 538's most likely probability model by one or one and a half seats. 538 gave the Republican party an 82% chance of retaining control of the Senate. Smaller sample size, for whatever that is worth.
538 Predicted that among governorships, the Dems would most likely end up with 24 to 25 governorships, depending on the model. That would leave about 199 million people living in States with Dem governors, and 130 million in states with Republican governors. The results there were 27 states with Republican governors and 23 with Democratic governors. So the Republicans outperformed and the Dems under-performed relative to 538's most like forecast, with 538 being off by one or two states depending on the model.
So the polling was imperfect as always, but still fairly close. Certainly accurate as far as the overall trend of Dem's flipping Republican held seats.