Model debuts tomorrow. Will post in the attached thread. But early indications on the VP pick is that Shapiro will not move the needle. If anything, Trump's chances improve slightly if Shapiro is chosen.
First model forecast: Trump: 45.2% Harris: 21.1% Unallocated: 33.7% Best way to think of this is as the maximum price I'd be willing to pay (on 0.00 to 1.00 scale) to bet on either candidate. To explain further: the Unallocated amount reflects inherent uncertainty in the
If forced to give a full probability forecast: Trump: 61% Harris: 39% Looking back through the past few months, this hasn't changed much at all from when Harris was just a hypothetical candidate. Only the perceptions have changed: 1. Harris the candidate vs. hypothetical
Quick update Trump: 48.0% Harris: 28.5% Unallocated: 23.5% Still no VP pick. Signs of Harris starting to plateau. Plus a relatively good week for Trump. Full forecast: Trump: 61% Harris: 39% No change in the plot. Other forecasts are over reacting to short term noise.
August 10 update: Trump: 42.5% Harris: 41.5% Some people don't like the "unallocated" %, so I'll leave that out. It can be calculated by 100 - 42.5 - 41.5 = 16.0%. If this confuses you, convert to a minimum price / odds at which it's worth considering a bet: Trump 100/42.5 =
Model update: Trump 52.5% Harris 40.3% Still some uncertainty because of RFK decision and time until the election. Significant market moves in the past week. Likely a combination of: 1. Harris releasing economic policies (unpopular) 2. More Harris media exposure. 3. RFK
Labor Day model update: Trump: 61% Harris: 34% RFK: 0.1% Trump back to pre-dropout/coup numbers v. Harris. Model Uncertainty down to 5% and should continue to decrease with time. RFK with a non zero chance of winning, must be accounted for.
A bit late on the move @NateSilver538 https://x.com/PpollingNumbers/...
@HarryDCrane What would be the reason behind Shapiro not moving the needle? Pennsylvania looks to be by far the most important state this cycle (due to number of electoral votes and closeness in polls).
@daschnauzer90 All I can really say is that that's what the data is telling me. Some theories on why: 1. VP rarely matters 2. Shapiro helps with PA voters who are already in Harris's pocket 3. The small gain Shapiro generates in PA is eliminated by losses in MI, OH, WI, where he is a liability
@HarryDCrane Fraud
@HarryDCrane Excited for this
@HarryDCrane training on poll data, weighted and aggregated? from multiple elections? proportion of predictive power from novel variables, engineering techniques, model + tuning . . .? many questions, excited to see