We are on the precipice of history. Unless something truly odd happens–but hasn’t the whole race been odd?–Nov. 9th will bring radical change to the United States. We will wake up with either our first female president, one who limps into office under the specter of an FBI investigation, or we will welcome into the Oval Office a man with no prior political, military, or diplomatic experience, a man who threatens systemic change of all sorts.

So, what is happening in the actual election?

Real Clear Politics has the race at 203 safe electoral college votes for Clinton and 164 for Trump. They list the following states as toss-ups:

Arizona, Colorado, Florida, Georgia, Iowa, Maine, Michigan, Nevada, New Hampshire, New Mexico, North Carolina, Ohio, Pennsylvania, and Virginia.

Let’s think of these states as three separate groups–Clinton, Trump, and Toss-Up. Here are some interesting demographic differences:

Based on U.S. Census data, I looked at racial, ethnic, and educational characteristics for all states. I also looked at religious data from the Association of Religious Data Archives to determine each state’s evangelical population. Finally, I used Real Clear Politics’ definitions for Clinton, Trump, or toss-up states. So, this gives us averages for all states within these categories. The average Clinton state, then, is 74.4% white, nearly 9% Hispanic, with nearly 37% having at least a college degree, and is 8.4% evangelical.

Some things jump out quickly. Clearly, Mr. Trump’s states are much whiter, much more evangelical, but less educated and less Hispanic when compared to Clinton’s states. The toss-ups are whiter than the Clinton states, but also more Hispanic and more evangelical than typical Clinton states. In some ways, it makes sense these are toss-ups. They are demographically unusual.

What about recent electoral history (at least at the presidential level) for these toss-ups? Here is another table.

It is always dangerous to assume things. After all, it is possible we are looking at something very different in 2016, and if true, what happened previously is less important than we think. Everyone agrees 2016 has been odd and maybe Election Day will be even more odd. If the past is a reasonable guide, based on demographics and recent history, we can safely assume the following.

Trump should win AZ and GA. If he loses GA, expect it to be a long, rough night for the Trump campaign. Trump should also win Iowa. He has been leading consistently in polling, and it is very white (93.9) and 26.8% of the state population has a college degree. In short, it looks like a Trump state and, actually, so does Ohio, which has similar demographics and Trump has been ahead there for some time. These (AZ, GA, IA, and OH) bump Trump up to 215.

Mrs. Clinton should win New Mexico (which has a large Hispanic population), and she will most likely win Maine, based on recent history (I’m also assuming she will win all of Maine’s congressional districts, though there is a chance Trump will win District 2). That pushes her to 211.

From East to West, that leaves us with NH, PA, VA, NC, FL, MI, CO, and NV as the real battleground states. Together, those yield 112 electoral college votes.

My guess, based on the charts above and recent polls, is that Clinton wins Colorado, Nevada, and Virginia (that puts her at 239), while Trump likely could win Florida and North Carolina (which puts him at 259). If Trump loses Florida, there is no chance he wins the presidency. The same is probably true for North Carolina. Let’s assume he wins those. That leaves Pennsylvania, Michigan, and New Hampshire.

Republicans have not won Pennsylvania since 1988. Hillary has a narrow lead in most recent polls, though Trump has been either tied or is ahead in some polls. 1988 was also the last year Michigan went Republican. Clinton’s lead in Michigan has been more safe, but a late Republican affiliated poll shows Trump in the lead by 2 points. In New Hampshire, the polling has been all over the place, even with a spate of recent polls. George W. Bush won NH in 2004, just as his father did in 1988.

In some ways, if the election actually comes down to these three states, it will be a true test of Trumpism. The electorates are largely white and not terribly evangelical. More than anything, Michigan and Pennsylvania are precisely the kinds of states where Trump should do well. They, along with Ohio, are the typical “Rust Belt” locales, places that have never fully recovered from the loss of manufacturing jobs. New Hampshire gave Trump his first primary victory and started his momentum toward the nomination.

If Trumpism is a real phenomenon, a movement that brings in new voters from the shadows, regardless of their motivation, Trump could win all three states. If there is such a thing as a “shy” Trump voter, or if social desirability has skewed polling, Trump could win all three states. If there is a “Brexit” phenomenon that is difficult to measure because many supporters simply see polling as a rigged system, Trump could win all three states. If Hillary Clinton struggles to replicate Obama’s coalition, Trump could win all three states.

That would make Donald J. Trump President of the United States with 299 electoral college votes.

In the end, I doubt those things are true. I think the Democrats will hang onto all three states, giving Mrs. Clinton at least 279 to Mr. Trump’s 259. If things trend against Trump badly, she could win a host of other states. We shall see.

Here are poll closing times (Eastern).

7:00: GA, IN, KY, SC, VT, VA

7:30: NC, OH, WV

8:00: AL, CT, DE, FL, IL, ME, MD, MA, MS, MO, NH, NJ, OK, PA, RI, TN, DC

8:30: ARK

9:00: AZ, CO, KS, LA, MI, MN, NE, NM, NY, ND, SD, TX, WI, WY

10:00: IA, MT, NV, UT

11:00: CA, HI, ID, OR, WA

1:00 AM: AK

If things are heading in one direction, my guess is we will know it by 8:00 p.m. If the networks can call GA, VA, NC, OH, FL, NH, and PA with relative ease, because one candidate’s lead is apparently insurmountable, it might be an early night. If Trump manages to win all of those, it is safe to say he will probably win. If Mrs. Clinton wins FL, the election is probably over. If she wins PA and VA, Trump will probably have to win MI and/or CO. If the elections are tight, and exit polls are not definitive, settle in. We will be in for a long evening.