One of the questions any political reporter is asked most often is a simple one: Who’s winning the presidential race?

The answer to that question, though, can be complicated. Do you mean who has got more national support? Are you curious about the breakdown of the electoral college? How worried are you about niceties like margins of error or the number of polls being conducted?

Then there’s an additional level of context. How the race stands at the moment is one thing, but a lead for former vice president Joe Biden in, say, Georgia, is exceptional in part because it’s an aberration from recent history. That Biden is even close there or in Texas marks a shift from past elections that is itself instructive about 2020.

It can be tricky to compare current data to historical data for the simple reason that polls aren’t always comprehensive or equivalent. The Washington Post has polled repeatedly for decades, but even our polling has gaps: periods in which public sentiment isn’t measured; and relatively sparse polls at the state level.

There is, however, a resource that can offer context for where things stand both at the state level and relative to past elections. The analysts at FiveThirtyEight are known for compiling an average of national presidential polls (which can be seen here). But they also have averages both nationally and at the state level stretching back more than 40 years, where available, data that they make public.

So we took that data and compiled the average margin between the Democratic and Republican candidates in each year and each state (where available) to depict the flow of each contest over time and how the 2020 election matches up. When a line dips below the horizontal indicator in the middle of the graph, the Democrat was leading. When it’s above, the Republican was.

There are gaps in the data, and of course polling isn’t always precisely on the mark. (For past elections, we’ve indicated the actual result with an arrow at the far right of each graph.) But the interactive below gives some sense of where the contest stands.

Show average, including

Your browser cannot display this graph.

You can see from the default view that the 2020 election has been unusually stable. (We’ve reduced the default set of contests to just those in the past 20 years. Feel free to add more, recognizing that things get messy quickly.) Compare it with 2016 or to 2012, and the picture that emerges is a race that hasn’t really changed very much.

That’s true at the state level, too. While in places such as Georgia the lead moves back and forth between Biden and President Trump, the actual shift in the margin has been fairly modest. In states that were important in 2016, such as Michigan, the race has been consistently stable to Biden’s advantage.

One thing worth noting in the contest in Michigan is how the polls erred. If you look only at the past three contests, you will see that where the polling averages were off the mark as the races each year concluded varies. In 2008, the actual margin for the Democrat in Michigan was wider than the polling average; hence, the arrow marking the 2008 results is lower than the final average. That happened in 2012, too — but in 2016, the actual result was more friendly to the Republican than the polls suggested. (As Trump will happily remind you.)

That actual results will deviate from polling is normal. What’s interesting here is how they deviated. In each year, the polls failed to capture the extent of support for the winning candidate. In 2012 and 2016, that was in part a function of a failure to predict who would turn out to vote. It’s worth noting that this could happen again, in one direction or the other. It’s also worth noting that Biden’s lead in Michigan is currently larger than the miss in 2016.

This interactive will update with new national and state polling each day. (It refreshes every few hours as new polls are added to FiveThirtyEight′s averages.) None of this should be considered definitive, any more than you should consider any poll taken more than a month before an election precisely predictive of the results. But it should, at least, answer that ever-so-common question about where things stand.