Since we ran the Michigan numbers early Monday morning, three new polls are out that make the state look more like a true toss-up and less like one that favors Mr. Romney.
Two of the surveys, from Mitchell Research and American Research Group, in fact give Rick Santorum a nominal lead in Michigan, by 2 and 1 percentage points respectively. The third, from Rasmussen Reports, gives Mr. Romney a 2-point advantage.
We also added a hard-to-track down survey from Baydoun Consulting, which gave Mr. Romney an 8-point advantage. However, it is less recent than the others, having been conducted on Thursday night rather than over the weekend.
Among the five polls that were conducted over the weekend — including those that had been included with the previous update — three give Mr. Romney a small lead while two show an edge for Mr. Santorum.
Mr. Romney still has the advantage in the FiveThirtyEight forecast, but it is more tenuous than the one we released overnight. The model gives him a 64 percent chance of winning the state, down from 77 percent in the previous forecast.

What’s less clear is whether there’s actually been a tick of momentum back toward Mr. Santorum. The Rasmussen Reports and Mitchell Research polls did show Mr. Santorum’s position improving slightly from surveys they had conducted on Thursday night, and there were hints in those polls that some of Mr. Santorum’s conservative base voters may be returning to him after flirting with Mr. Romney.
But the broader trend has been toward Mr. Romney, both in Michigan and in polls of other states like Arizona. Our model, which has a statistical technique to calculate momentum, considers it something of a draw and does not give much of a bonus to either candidate based upon it.
At least one polling firm, Public Policy Polling, will be in the field again on Monday night. It may also be worth watching to see whether the campaigns are raising or hedging expectations, which sometimes reveals their confidence about what they are seeing on the ground and in their internal polls.