An Idiot’s Guide to Winning Elections – Outlook Zen

Skip to content

RP

Uncategorized

August 15, 2026August 15, 2026

In the past week, we saw something very remarkable. In a matter of hours, Democrats saw their chances of winning the Wisconsin Governor election jump from 40% to 77%.

What changed? Did new polling numbers come out? Was there a shocking scandal? Did someone fail miserably at the debate?

No, it was none of the above. What actually happened was that the Democratic primary results came in. Francesca Hong was the favorite to win the Democratic nomination, and pundits predicted that she had a 40% chance of winning the general election. But as soon as David Crowley was announced as the nominee, the Democrats saw their odds of winning the general election double. They went from being slight underdogs, to clear frontrunners.

If you’re a Democrat, you may want to hold off on your celebrations though – we’ve seen the exact opposite happen in other elections. For instance, over the past month, Democrats have seen their odds of winning the Michigan Senate election drop from 72% to 55%. They have gone from being favorites, to what is essentially a coin-flip.

And once again, this drop in odds has been almost entirely driven by the primary election. El-Sayed started off the month with a 66% chance of winning the primary. And as the month went on, as it became increasingly clear that El-Sayed was going to win the primary, the Democrats saw their general election odds correspondingly fall.

Now, I’ll be honest and tell you that I don’t know the first thing about Hong, Crowley, El-Sayed, or Stevens. I have no idea how charismatic they are, or what policies they are endorsing. Heck, I can’t even pick out Wisconsin on a map. And yet, I don’t need to know any of these things in order to tell you that Stevens is more likely than El-Sayed, and Crowley is more likely than Hong, to win the general election if nominated. The numbers speak for themselves.

For too long, people have evaluated "electability" using extremely unreliable signals.

Personal appeal: "If I really like this candidate, general election voters will also really like them!"

General election polls: "The latest Siena poll shows that this candidate will comfortably win the general election!"

Ideology: "This candidate is an enlightened centrist. Therefore, she is most likely to win the general election, unlike all these other un-electably partisan candidates!"

Unfortunately, all of these signals are far worse than the alternative.

Personal appeal is probably the worst – just because you’re a die-hard Bernie bro or MAGA devotee does not mean the rest of the country feels the same way. We should all have the humility to not project our own feelings onto others.

Polls are generally a good signal, but only once we get much closer to the election date, and once both nominees have been chosen and subjected to the full barrage of attack ads and media scrutiny. Until then, "fresh faces" consistently overperform in polls due to the simple fact that they haven’t been bruised up yet.

Ideology is probably the most reliable of the three during the primaries, but that isn’t saying much. It is all too easy for "centrists" and "moderates" to suddenly find themselves branded as "establishment" candidates in an electorate that is tired of the status-quo. Or for someone with all the "right platform stances" to lose anyway because voters don’t think they are "authentic" or "likeable".

If this was 20 years ago, I would have thrown my hands in the air and said that’s the best we can do. Luckily for us in 2026, there is a far better way. Prediction markets, like Polymarket. Exactly the ones we had discussed at the start of this essay. To quote research on this topic:

The evidence we present in this paper shows that the markets are also accurate months in advance, and do a markedly better job than polls at these longer horizons . In making our comparisons, we compare unadjusted market prices to unadjusted polls, demonstrating that market prices aggregate data better than simple surveys where the results are interpreted using sampling theory

Prediction markets aggregate dispersed and unpublished information (i.e., a brewing scandal may be known to a few investors before the general public) … Further, prediction market stocks are based on the value of the candidates on Election Day; thus, investors are incorporating their information on how it will affect the race on Election Day, while poll-based forecasts are only able to debias the information based off of previous cycles (i.e., investors can discount a bump in the polls generated by the visit of a popular leader, but poll-based forecasts can only discount the bump if it happened regularly, at the same time, in previous cycles)

Descriptive and predictive results indicate national Polymarket data was superior to that of the polling data in predicting Trump to be the winner of the 2024 presidential election...