538 Model Megathread

[muon2]

Discussions like these are why I wish Nate would actually release the 538 model (or just the results of the simulations) for us to play around with.  We could then actually get a sense of whether the firewall really exists within his model.

The results of the EC simulation are on the main page at the bottom. The chart doesn't have much detail, but one can see the peaks for the most likely combinations out of the simulation. The highest peak generally corresponds to the combination if all the states follow their current most likely candidate. The way I read it today, one can pull out that the most likely occurrence is a Clinton win with 272 EV. That is the current Clinton states on his map plus ME-2.

[muon2]

The 538 model now has the 272 freiwal for Clinton turned into a 266 freiwal for Trump.

Trump has no states in the now-cast closer than R+2.6. Clinton has 4 states closer than D+2.0: CO, MI, NH, PA. A Clinton win with 272 is still the most probable single outcome, but an additional flip from D to R is more likely than a flip from R to D. So the now-cast has a number of Trump-winning outcomes with nearly the same likelihood as Clinton 272. That just shifts the total number of simulated Trump wins ahead of Clinton wins.

In the polls-only forecast each has two states within 2%: CO D+1.3, FL R+1.6, NV R+1.5, NH D+1.7. That makes it roughly equally likely that states will flip to either candidate from the 272 freiwal scenario. With equal flip possibilities each way, Clinton maintains her position as the favorite in this forecast.

[muon2]

If you use HuffPo 3-way and limit the polls to nonpartisan pollsters and likely voters the current numbers from their trend lines are Clinton 45.1%, Trump 40.7%, Johnson 5.1%. For the 2-way it gives 45.8% to 42.4% (moderate smoothing)

[muon2]

If you use HuffPo 3-way and limit the polls to nonpartisan pollsters and likely voters the current numbers from their trend lines are Clinton 45.1%, Trump 40.7%, Johnson 5.1%. For the 2-way it gives 45.8% to 42.4% (moderate smoothing)

Why would you do any of this?

This in response to posts on both sides complaining about inclusion of partisan pollsters. HuffPo lets me filter them out so I did this as a point of comparison.

By further comparison that same exercise today has Clinton 45.2%, Trump 42.5%, Johnson 4.9% in the 3-way and Clinton 45.4% to 42.8%  in the 2-way. Clinton's share in both is essentially unchanged from my post 4 days ago.

[muon2]

What is the uncertainty of the model, given the # of simulations they run?  That is, each time they enter new polls, they apparently run 10,000 simulations based on the latest #s, and that produces (among other things) an overall projected vote margin and win probability.  But let's say they then ran *another* 10,000 simulations with the same input #s but a different random seed?  How different would the results be?  Because if they'd be different, then in theory you could put in a favorable poll for one candidate and it would end up "helping" the other candidate, just because of simulation noise.  But I'm assuming that 10,000 is enough for the simulation noise to be small?

This is purely a statistical noise effect; given Clinton's win percentage of 70% or so, we shouldn't be surprised by jumps of 0.6% or so in Clinton's win percentage just from rerunning the simulations (1 sigma).  If you start looking at individual battlegrounds, we definitely shouldn't be surprised if one of them jumps a percent or two just from statistical noise.

I wouldn't have expected changes that significant with 10,000 trials, although I will say my technical experience in statistics is fairly limited.

Statistical noise scales as 1 / sqrt(number of trials), so it takes a lot of trials to get your error below the percent level.  To decrease your statistical error by a factor of 10, you need to run your simulation 100 times longer.

This does mean that 538 really shouldn't be listing the percentage points on their probabilities, unless they feel like running a million simulations each time.

With 10K trials and a 50% expected value the estimated standard error is 0.5%, for a 20% (or 80%) expected value the estimated standard error is 0.4%. For a technical paper they should include the decimal on their percentage results, but also quote the standard error as 0.5% (or less for percentages away from 50%). However, if the intend the user to interpret the error as a margin of error then that is 1.98 times the standard error. In that case they should not use the decimal and report the value as a whole number percentage. Since polling and journalists tend to assume a margin of error then whole number percentages would be preferred.

[muon2]

If you use HuffPo 3-way and limit the polls to nonpartisan pollsters and likely voters the current numbers from their trend lines are Clinton 45.1%, Trump 40.7%, Johnson 5.1%. For the 2-way it gives 45.8% to 42.4% (moderate smoothing)

Why would you do any of this?

This in response to posts on both sides complaining about inclusion of partisan pollsters. HuffPo lets me filter them out so I did this as a point of comparison.

By further comparison that same exercise today has Clinton 45.2%, Trump 42.5%, Johnson 4.9% in the 3-way and Clinton 45.4% to 42.8%  in the 2-way. Clinton's share in both is essentially unchanged from my post 4 days ago.

The Trump line in my exercise is still rising but more slowly than it did last week.

3-way: Clinton 44.5%, Trump 43.6%, Johnson 4.4%
2-way: Clinton 45.3%, Trump 43.6%

The Clinton data is still very steady as it has been over the last month. Note that when Johnson is not included Hillary gets a quarter of his vote, but the rest goes to undecided. To eliminate recent noise and movement between undecided/third party I can put on more smoothing on the two-way results and she is at 46.0%. However, over time Trump continues to gain at the expense of Johnson, but the trend lines don't show him catching her by Tuesday.

[muon2]

If you use HuffPo 3-way and limit the polls to nonpartisan pollsters and likely voters the current numbers from their trend lines are Clinton 45.1%, Trump 40.7%, Johnson 5.1%. For the 2-way it gives 45.8% to 42.4% (moderate smoothing)

Why would you do any of this?

This in response to posts on both sides complaining about inclusion of partisan pollsters. HuffPo lets me filter them out so I did this as a point of comparison.

By further comparison that same exercise today has Clinton 45.2%, Trump 42.5%, Johnson 4.9% in the 3-way and Clinton 45.4% to 42.8%  in the 2-way. Clinton's share in both is essentially unchanged from my post 4 days ago.

The Trump line in my exercise is still rising but more slowly than it did last week.

3-way: Clinton 44.5%, Trump 43.6%, Johnson 4.4%
2-way: Clinton 45.3%, Trump 43.6%

The Clinton data is still very steady as it has been over the last month. Note that when Johnson is not included Hillary gets a quarter of his vote, but the rest goes to undecided. To eliminate recent noise and movement between undecided/third party I can put on more smoothing on the two-way results and she is at 46.0%. However, over time Trump continues to gain at the expense of Johnson, but the trend lines don't show him catching her by Tuesday.

I think 538's model is reacting like the trendlines I'm seeing with HuffPo's tool. The Trump poll numbers have stopped moving up so his trendline comes down and hers moves up. Here's what it gives me this afternoon.

3-way: Clinton 45.5%, Trump 42.9%, Johnson 4.5%
2-way: Clinton 46.3%, Trump 43.3%

[muon2]

Is it just me, or have you noticed that in just one or two days, trump's chances went from 35%-ish to what is now about 28%-ish.
That is a huge drop in such a small amount of time.

It's not so much a huge drop as it is a slight change in the slope of the projecting line. Think about a laser pointer aimed at a spot across a room 40 feet away. As the pointer is walked towards the wall a slight shift in the hand holding it can move the spot by a number of feet. Towards the end the point won't move as much because the pointer is much closer to the wall.

The trendline last week was very favorable for Trump but there was a slight shift back away over the weekend. That moves the projection point back away from last week's projection, just like a shift of the light on the wall. It's too close to the election to shift all the way back to where Clinton was a three weeks ago, but it resets the final point back about a week.