For the second time ever, Chris Cillizza had a good tweet on Wednesday.

Granted, Cillizza just took a photo of something that our colleague Elise Viebeck put together, which explains it. But it's interesting!

Viebeck took the Associated Press's list of when it declared winners in states in the 2012 presidential election and compared that to when the polls had actually closed. It gives us a concrete sense of something we already knew: Trying to figure out when election night actually ends is a fool's errand.

(That said: Here's our interactive look at how long it is until polls close in every state.)

Of course there's a correlation between when AP calls a state and how close it is. I took Viebeck's data and compared how long it took for the AP to make its call (that is, the number of minutes between when polls closed and the organization declared a winner) and the eventual margin of victory in the state.

Two races don't fit neatly on the scale there: Florida, which was very close and took a few days to resolve, and D.C., which was not close at all. Otherwise, there's a general pattern. Once the margin in a race is 25 points or more, it's called instantly (thanks to exit polling, statistical modeling and initial returns.) Slightly closer races can be called quickly — but sometimes take a little bit. Fairly close races take an hour — or, in the case of very close races, multiple hours.

The curved, dotted line on that graph approximates the trend. Notice that many of the races don't fall on that line. The ones to the left of it were called sooner than expected; those to the right, later. It's a rough guide.

But we took that rough guide and made a little tool that will estimate when a race will be called depending on the margin in the state — or figure out the likely margin, depending on when the race is called. This is based on one cycle of data and we know that not all of the 2012 results fit the curve perfectly, so don't expect pinpoint accuracy. Unless it happens to be pinpoint-accurate, in which case we meant to do that all along.

An additional bit of happy news: Chris Cillizza can now tweet this tool out and, for the second time in one week have a tweet worth looking at. Impressive.