How To Read The Polls: A Guide To The 2016 Election Horserace

— -- The most difficult thing to understand while tracking the 2016 race may also be one of the most important: public opinion polling.

Like checking the scoreboard in the final minutes of a playoff basketball game, Americans are soaking up each new political poll released in this year’s topsy-turvy election cycle.

A poll isn’t a crystal ball for the future: the goal is a snapshot of the current state of the race. So how is it done? Ask a small, representative group of people for their thoughts on the presidential race.

It works best with a small sample size –- usually around 1,000 people –- as a representation of the full U.S. population. (And the likelihood of being asked is very small -- there's a better chance of being struck by lightning than being called for the next ABC/Post poll.)

Here’s what to watch while reading the polls over the next eight weeks:

1. Registered voters vs. likely voters

The distinction may only be one letter in the fine print of a poll’s methodology, but it can make a difference. Now that the election is two months away, pollsters are beginning to move away from asking any registered voters and zeroing in on only the voters who are likely to show up on Election Day.

This change tends to boost Republicans: core Democratic groups like nonwhites and young people have lower turnout rates than GOP-leaning groups. This shows in recent polling. The latest ABC/Post poll showed Hillary Clinton leading by 10 percentage points among registered voters but only 5 points among likely voters. A CNN/ORC poll from last week showed Clinton +3 among registered voters but Trump +2 among likely voters.

But not all likely voters are created equal. Because of social pressure, people are prone to say they will vote even if they probably won’t. So the best likely voter models won’t just ask if a person is certain to vote; they will also ask about how interested they are in the election and past voting behavior.

2. When a lead may not firm: Margin of error

A poll may show a candidate leading by 1 percentage point. But it's important to remember that polls are just a point estimate of the whole population. They have some wiggle room -- and that's why the margin of error is so important.

For example, the latest ABC/Post poll showed Donald Trump with 41 percent support among likely voters, plus or minus 4.5 points. This means that Trump’s true level of support among all likely voters in the whole country is somewhere between 36.5 and 45.5 percent. The same goes for Clinton: her 46 percent support could be between 41.5 and 50.5 percent.

Langer Research Associates have a nifty tool here on their website to measure whether a candidate’s lead is statistically significant or whether it could just be normal variance.

3. If one poll looks like an outlier, it probably is

Sometimes polls can fall several points away from where other polls have pegged the status of the election. This is normal and happens occasionally. The rule here is not to overreact to a single poll showing an abrupt shift.

Waiting for more polls, especially mainstream live-interviewer polls, to show the same movement helps. Another option is to monitor the averages of several polls put together. RealClearPolitics and HuffPollster have options to show movement in the race in a more measured, deliberate way.

4. National polls vs. state polls

National polls give an important overview of the state of the election; they tend to be more reliable than state polling and eliminate the complexities of various Electoral College paths to the White House. But the president is elected based on the votes of individual states –- not a the popular, or individual, vote.

Still, the winner of the popular vote almost always wins the White House (with the glaring exception of the 2000 election) and, of course, the full country is made up of states, so state polls and national polls move in tandem with each other.

But state polls tend to be more unreliable –- many use lists of registered voters to decide who to call, but many of these lists have holes of missing voters and missing telephone numbers that can skew results.

5. The partisan split: Keeping an eye on party ID

When people are trying to dismiss polls as inaccurate, they will often point to the partisan balance of the poll and say it asked too many Democrats or too many Republicans. Most major polls don’t weight based on party ID, so this will vary from poll to poll and is usually released in the poll’s methodology section.

Partisan ID is expected to be in the same ballpark as the 2008 and 2012 elections, which had Democrats +6 and Democrats +7 in exit polls, respectively. The latest ABC/Post poll had Democrats +8 among likely voters.

Some organizations will try to "unskew" the polls by re-weighting the results to their own balance of Republicans and Democrats. But weighting for party disrupts any change in voters' allegiance to their parties -- from 2004 to 2008, the partisan split went from even to Democrats +6.

6. Warning signs that some polls aren't accurate

It seems like every day there’s a new poll coming out. The most important thing to check is sampling, which basically means how people get chosen to take the poll. If it’s an Internet-only poll that anyone can take, the poll is going to miss major groups of people who may not regularly use the Internet or visit that website.

Even some opt-in Internet polls that use extensive weighting are unreliable. A recent Pew Research study found estimates among nonwhites and young adults in non-probability surveys "tend to be especially biased." Live-interviewer polls that call both landlines and cell phones have their own problems, but are the most reliable polling right now.

One way to understand the fluctuation of the main presidential voting preferences is to dive into the rest of the poll: Questions about the issues and qualities of each candidate. It’s like looking at both the statistics and the score during a football game. These numbers show the moods and attitudes that are driving the support of each candidate and can be indicators of the direction of the race.