One’s confidence in predicting what’s likely to happen in the contest for the 2020 Democratic presidential nomination correlates neatly with how closely one is looking at it.

In the abstract, it seems like a mess, with a field of more than two dozen trimmed down only partially and a leading pack that still overstuffs debate stages. Dig a little deeper, and things seem more certain: There are really four leading candidates: former vice president Joe Biden, Sens. Bernie Sanders (I-Vt.) and Elizabeth Warren (D-Mass.), and former South Bend, Ind., mayor Pete Buttigieg. One of those four seems likely to secure the nomination, in decreasing order of likelihood.

But then you go a little deeper, and things become a mess again. National polling shows those four in the lead in three rough tiers — but in early states, things are a muddle. Look at what’s happening in Iowa, for example, and recognize what tends to happen after Iowa, and suddenly things look awfully murky once again.

With that in mind (and with voting beginning in less than a month), it seems useful to look at what we know and what we might expect.

Let’s begin at the beginning.

Iowa, Feb. 3

Iowa is an odd place to begin for a number of reasons, but, this year, it’s odd for three particular reasons.

The first is that it has caucuses, not a primary, and the winner is determined through an unusual system of horse-trading (which we’ll get into below). The second is that the results this year may not lend themselves to the sort of clear so-and-so-won narratives to which we’ve gotten accustomed. The third? It’s a state that doesn’t reflect the diversity of the Democratic electorate, muddying the extent to which one can learn from its lessons.

As noted above, polling in the state essentially shows a three-way tie between Sanders, Buttigieg and Biden.

What’s particularly important in Iowa, though, is the depth of support caucusgoers have for candidates.

On Feb. 3, when the caucuses occur, those present at a caucus site will be asked who they support. If a candidate fails to earn 15 percent of the total at the site, supporters of that candidate will be asked to choose someone else to support. That process continues until a candidate reaches the level of support necessary to earn delegates: 50 percent of the total for caucus sites awarding one delegate, 25 percent at sites choosing two delegates, 16.7 percent at sites choosing three delegates and so on.

In past years, the total level of support statewide after all the caucuses have been completed has been reported as the result. Actual delegate allocations, more important in the overall calculus for winning the nomination, has tended to be secondary. This year, though, the party will release another datapoint, the first-round levels of support for each candidate. In other words, we could see multiple candidates come out of Iowa claiming a victory: The candidate who had the most first-round support, the candidate who ended up with the most support after redistribution of underperforming candidates’ supporters and the candidate who got the most delegates. Odds are that at least two of those metrics will point to the same candidate — but there’s no guarantee that that will take place.

What’s not clear is how that might affect the usual post-Iowa bump in polling. Could Buttigieg earn the most delegates in Iowa, for example, but Sanders be able to point to the most support and Biden to stronger-than-expected first-round numbers? How does that affect Buttigieg?

It’s also not entirely clear where the support for lower-polling candidates will end up. Polling in Iowa is complicated by default, thanks to the caucus system. But it becomes even harder to get reliable polling on the second choices of lower-tier candidates simply because there are, by definition, fewer voters who support those lower-tier candidates. In the most recent polling in the state, conducted by YouGov for CBS News, Warren was supported by 16 percent of respondents, compared with 23 percent support for the other leading candidates. But 46 percent of respondents saw her as someone being considered by them, suggesting plenty of room for her to grow as a second choice for voters.

The more important point is that even the polling above is less solid than it may seem. In that YouGov-CBS poll, 31 percent of respondents said they had definitely made up their minds about whom they support. About 6 in 10 said they had probably done so — meaning their picks might change.

This, in fact, is another level at which one might look at the field itself. Since July, Biden and Sanders have been fairly static in the polls, with Sanders enjoying a bit of a bump over the last month or so. Warren and Buttigieg have been more volatile, which is possibly one of the reasons they’ve found themselves more heavily targeted in debates: Their bases of support may not be as solidly committed.

What does that consistency for Biden and Sanders mean? For one thing, in a still-crowded field, each has a reliable base of support. That’s important because of another rule in the Democratic primaries that will start becoming important in the second contest.

New Hampshire, Feb. 11

New Hampshire votes more than a week after Iowa, allowing for post-Iowa assumptions to swirl around and for post-Iowa, pre-New Hampshire polling.

The state of the race is somewhat similar to Iowa, with all four leading candidates within eight percentage points of one another. History tells us that these numbers may be subject to rapid shifts after the Iowa caucuses, so keep that in mind.

What’s interesting about the numbers above, though, is that each of the four candidates is above 15 percent support in the polling. Should any of them drop below that mark in the actual vote, he or she would be in trouble. The party’s delegate allocation rules stipulate that only those earning at least 15 percent support statewide are eligible to get delegates. If Warren drops to 14 percent on Feb. 11, in other words, the state’s 24 delegates would be split (relatively) proportionately among the other three candidates.

(This is a good point at which to plug the Green Papers website, a thorough documentation of the remarkably complicated rules surrounding delegate allocation.)

Let’s imagine, for the sake of argument, that the results in Iowa and New Hampshire match the polling in each state. All four candidates divvy up some of the vote in each state, with a virtual tie in each state, but Sanders is able to claim vote leads in each state (itself in part a function of his steady, loyal base). What assumptions should we make about where the primaries go next?

Before we can answer that clearly, we need to remember something else: New Hampshire, like Iowa, is heavily white.

Nevada, Feb. 22

In 2016, Sanders performed slightly better than the outcome presented in the above hypothetical. He tied Hillary Clinton in Iowa and crushed her in New Hampshire (in part thanks to the proximity of his home state of Vermont). But by mid-March, his candidacy was essentially over because of Clinton’s ability to run up big margins in bigger states with more diverse electorates. Clinton’s margins over Sanders correlated to the density of the nonwhite vote, and in states with heavy black voting bases, she scooped up far more delegates than Sanders got in New Hampshire.

It’s possible, then, that we might come into Nevada nearly two weeks later — which is majority nonwhite — with Sanders claiming a delegate lead and, perhaps, closing the gap in national polling, but with Biden’s advantages with nonwhite voters reinforcing the idea that he would overperform in the state. (In a late December national YouGov poll, Biden led among Hispanic Democrats by 16 points and among blacks by twice as much.)

There’s not good recent polling in Nevada to inform where things are, but this is really the heart of the question. Will Biden’s support with those groups weather his coming in second or third — or worse — in the first two states to vote? Will the perception of his electability, a powerful driver of his support, be hobbled by such a result to the extent that his base suddenly erodes? Or will they instead be seen as irrelevant speed bumps?

There’s another question, of course. After New Hampshire, we can expect a number of less-successful candidates to drop out, passing their support to those remaining in the race. Any of the four leading Democrats could conceivably be among them: What happens to the field once that happens?

In 2016, that didn’t matter, as it was essentially a two-person race. But, again, remember what happened: Clinton came into the year with a wide national lead but essentially tied in Iowa and lost New Hampshire by 23 points. It didn’t do much to slow her march to the nomination.

South Carolina, Feb. 29

What’s more, Nevada is followed by South Carolina a week later. Nearly two-thirds of voters there in the 2016 primary were nonwhite, with Clinton winning those voters by 71 points. Biden might not fare as well, but if we’re playing the narrative game, big wins from the former vice president in the two contests to follow Iowa and New Hampshire could shift things back in his direction quickly. Especially since Nevada and South Carolina allocate 90 delegates to the 65 in Iowa and New Hampshire (which will likely be divided among more candidates).

Again, though, that only works if Biden can weather Iowa and New Hampshire without seeing broad erosion in his core base of support. And in the next month, for that matter, though he has seen no such erosion, save a hiccup after the first primary debate.

Super Tuesday, March 3

After South Carolina, things explode. On March 3, a number of states — including monsters Texas and California — weigh in. It will be the most important day on the primary calendar.

What the field looks like then is hard to predict. It’s hard to know who will be in. Hard to know how Iowa will inform New Hampshire and how both will affect things moving forward. If Biden maintains his support with nonwhite Democrats past New Hampshire, he’s well-positioned regardless of the outcome of those first two contests. But even after that point, the crystal ball is cloudy.

Check back in a month — when things will be only slightly less opaque.