In my field making predictions is considered risky. We tend to predict under the influence of our emotions (fear and desire). This is especially the case when we’re predicting things that are uncertain.

One of the ways to get around that is to run a simulation.

I ran a few simulations of the remaining primary contests between Hillary Clinton and Barack Obama.

For those of you that hate math skip the next paragraph and go straight to the results. I’m going to explain what I did using as little formal math as possible.

I began by putting in each primary (there are no more caucuses) in the order they occur. A few occur on the same day. I put in how many delegates each state has. I put in what I thought the turnout would be (using mostly Michael Barone’s numbers). I put in the average vote share of Hillary Clinton in the last three polls. Then I calculated a “variance” term for how far off the polls could be. If a state had only one poll the “variance” would be higher. If a state’s polls were very different from each other the “variance” would be higher. The later a state’s primary was the higher the “variance”. Next I used a random number generator to predict how well Hillary would do in each state within the range of the “variance”. Finally I put in a mechanism for dealing with momentum. The Hillary’s performance in the preceding primary would be factored in to how well she did in the next primaries. All numbers are for Hillary. Here’s what all of that looked like for the two simulations I ran.

In the first scenario I had Hillary win 50.1% of the Pennsylvania vote. This is the simulation least favorable to Hillary.

In the second scenario I had Hillary win 55.1% of the Pennsylvania vote. As you can see that’s higher but still realistic (and a bit low).

From then on I let randomness and momentum take over.

Here are the results I got from the Hillary unfriendly scenario:

In this scenario Hillary narrows Obama’s delegate lead by 22. She narrows his popular vote lead by 525,000 votes. She is still behind in both delegates and the popular vote.

Here are the results from the second scenario:

In this scenario Hillary narrows Obama’s delegate lead by 42. She takes the lead in the popular vote by 82,000 votes.

Now we shouldn’t take either scenario as anymore than a plausible prediction constrained by fact and divorced from emotion.

I haven’t run a Monte Carlo simulation but I suspect if I did, the odds of Hillary winning the popular vote would be greater than 35%. Perhaps she has a better shot at the nomination than we assume. If she enters the convention with the most votes I don’t think the superdelegates will let her lose.

I can address any technical questions related to the simulations in the comments.