I was texted election night by a few friends asking how I thought the night went for 2xParked.
Well, the Mets lost again that night…so that was not great.
Other than that, I am not in the prediction-making business. My role is to create a process to help you make the best predictions using a probabilistic mindset. My job is to translate the possibilities into probabilities by acknowledging that some scenarios are more likely than others, but all are possible. I always attempt to translate probabilities into tangible things, like coin flips or dice rolls, to provide a sense of what they mean to the not-statistical mind.
So how did I think the night went for 2xParked? Except for the Mets, I thought my analyses leading up were pretty solid and I am proud of them.
July 2024
In July 2024, I asked 5 key questions to kick off the mayor’s race.
Who will get endorsed by the Times? The answer ended up being no one, but it was critical to the race nonetheless. Not only did the NY Times not give the likely boost Brad Lander might have needed, an official endorsement would have also brought greater visibility in their anti-endorsement of Mamdani. It could have been a real obstacle to Mamdani’s momentum.
Who will be the “progressive” candidate? I mentioned at the time in 2024 that the difference between Bill de Blasio and the other progressives running in NYC (e.g. Cynthia Nixon, Maya Wiley) was (1) the masterful campaign and (2) the united progressive front. In the end, this was also Mamdani’s keys to success, although I think he could have gotten there sooner had Adrienne Adams and Brad Lander dropped out earlier when it was clear they were not the choice.
What lessons were learned on Ranked Choice Voting? TBD.
How will candidates get out on the streets? In the first fully post-Covid citywide election, the ground game of Mamdani and the mobilization of his followers was the same recipe as the pre-2021 success for candidates. No other candidate did that to the same magnitude.
What should we consider with polling? Well… I said it in 2024 and I will say it now….
The Dust Settle Principle
I am officially coining this as “The Dust Settle Principle.”
Most poll followers are familiar with presidential polls. Even in the most battleground of states, the undecided vote in the poll is <5%. When pollsters or poll modelers model out the undecided vote, they often ignore it or make bad assumptions. This is how the commissioned polls did their own RCV simulations in this last cycle. (This is also why I mostly ignored this part of the poll data).
This method’s risk is minimal when a poll has little undecided votes, but in our mayoral primaries, the undecided was quite high. This can be a reflection of (1) undecided of which candidate to choose and (2) undecided of whether to vote. This adds more randomness and variance to the prediction.
I mentioned many times that no candidate should feel safe with a high undecided vote in the polls. In 2013, when the dust settled and the undecided vote finally lowered to <10%, Bill de Blasio found himself going from underdog to leading candidate. In 2021, when the dust settled and the undecided vote finally lowered to <10%, Kathryn Garcia emerged as a clear top 3 choice. In 2025, we got one decent poll when the dust settled and the undecided vote finally lowered to <10%, and it was the one with Mamdani on top.
The uncertainty in your mental model when interpreting a poll should be proportional to the undecided vote: The Dust Settle Principle.
A Potential Oversight
Given where we are today, roughly 5% of my simulations had Mamdani doing this well or better in the first round. This not a bad job of my MayorModel, but definitely on the high end. (Cuomo actually performed right around the median simulation).
What I do not know is the reason why Mamdani performed on the upper end of my projections, and I don’t think we will have a sense until the RCV are counted. Many of the well-known other candidates (e.g. Adrienne Adams, Scott Stringer, Brad Lander) greatly under-performed, and I wonder if there is a correlation I did not account for. Was this under-performance due to (1) Mamdani bringing in more votes than expected, (2) supporters of both Mamdani and these candidates not taking chances and switching Mamdani to 1st choice instead of 2+, or (3) a true under-performance due to too high expectation?
In likelihood, it is a combination of (1), (2), and (3), but the RCV will help guide that answer. (Note: Given where we are in vote tallying as of Wednesday night, there is a non-zero chance if it is (2) that Cuomo could still make this a very tight race).
My One Prediction
I made one actual prediction in my newsletters which was not probabilistic, but was based in probabilities. I predicted that if a candidate received at least 43% of the vote and a significant chunk of the vote was counted, the winner would be determined on election night. In my model there was a 99% chance of that candidate winning. Mamdani received 43.5%, and Cuomo conceded.
This is Just the Beginning…
We still have a general election this November. Mamdani might be a favorite to win today, but don’t count out Eric Adams (or Andrew Cuomo if he runs). Adams is at 23% in Polymarket and rising. 4+ months is a long time to make significant political moves. 2xParked will keep you informed as new general election polls begin to be released.
Like your posts!