🔮 Trying to Kill (and Convict) Trump Made Him Stronger

And 7 other lessons from the election

Prediction markets are fundamentally about the future. And while we try to keep The Oracle forward looking, it’s helpful to look back after big events and see what we can add to the discussion on prediction markets. After reviewing the data, here are eight observations that jump out about the race:

1.  Trump was Favored Most of the Year.

The Polymarket general election market launched in January 2024. Between then and the election, Donald Trump held the lead (over 50% odds) on 194 days, and the Democrat (Biden or Haris) was ahead on 111. Viewers of Polymarket were getting useful information about the state of the race for much of the election year, i.e. that it was a close race with a Trump edge. In contrast, probabilistic forecast models such as ABC News’ FiveThirtyEight launched late in the summer, providing their formal forecasts for around three months leading up to the election.

2. Trying to Kill (and Convict) Trump Made Him Stronger.

The start of Trump’s federal documents case, the New York hush money verdict, and the Butler PA assassination attempt are all correlated with lifts in Trump’s odds. The assassination attempt stands out as the sharpest spike in the data, although it was soon overshadowed by Biden’s withdrawal from the race and the “Brat Summer” honeymoon for Harris that followed. Looking back at the race as a whole, it’s easy to tell a story about how the legal attacks on Trump, combined with the attempts on his life played into his core campaign narrative of unjust persecution.

3.  Harris Didn’t Outrun Biden

In election post-mortems, the conventional wisdom among Democrats seems to be that Biden would have done much worse than Harris had he stayed on the ticket. But there’s not much evidence in the data to back this up. Both Biden and Harris spent more than 60% of the time trailing Trump in the odds (64% of days for Biden vs 62% of days for Harris), and their average odds were within a point of each other (48% average odds for Biden vs 47% for Harris).

The picture looks worse for Biden if you isolate the period after the June 27 debate with Trump, in which Trump’s average odds are 63.3%. But for the full period that Harris was at the top of the ticket, her odds ran about a point lower than Biden on average, even with his debate debacle.  

4. Vance Proved Himself at the VP Debate 

Trump’s odds took a sharp hit on July 15, the day he announced JD Vance as his running mate, and there was certainly a lot of negative press coverage surrounding his rollout. However, it seems possible that Vance’s debate performance against Tim Walz redeemed his image and boosted the ticket overall. The October 1 VP debate kicked off a strong final month for the Republican ticket in the Polymarket odds. 

5. Some Swing States Were Harder to Predict.

The graph above shows the accuracy plotted using the log loss function (lower value implies higher accuracy). The dip and then sharp rise around mid-July corresponds to the crash in Biden’s odds following his debate with Trump. The accuracy seems to have been consistently higher in North Carolina, Georgia, and Arizona, while Michigan was the least accurate.

6. Markets Missed the Trump Popular Vote Win. 

Except for a 5-day stretch during the Biden dropout saga, Trump’s odds to win the popular vote never were above 50%, averaging just 32% throughout the whole life of the market. Given this major miss, it may be worth a deeper dive into polling data and demographic shifts to understand what happened here.

7. Polymarket Odds Were Election Night “Signal.”

If you were watching Polymarket on election night, you would have seen Trump pass 95% to win before midnight EST, nine hours before the AP called the race. To be fair, 95% odds (1 in 20 chance of being wrong) is a lower level of certainty than what the AP uses to formally “call” a race. But for an average news consumer looking to understand a fast-developing situation, Polymarket proved itself to be a faster-updating source of “signal” than traditional outlets.   

The most ‘wrong’ (farthest from the final result) any election market ever got was Kamala Harris reaching 67% odds to win Michigan in late September and 77.6% odds for her to win the popular vote in mid-August. 

8.  Polymarket vs. Other Models

Comparing Polymarket’s accuracy in the general election to ABC News’ FiveThirtyEight model shows that Polymarket was consistently closer to the result than the model. This makes sense because formal models are limited to pre-selected data sets, while market-based forecasts can consider any piece of information that any profit-seeking market participant can dream up.

What’s Next? 

The 2024 election has been a test case for marrying news and prediction markets. After the election, we will be broadening our coverage to include: 

• Understanding the Trump Administration. For example, a dedicated page on policies and personnel we’re likely to see. 

• Geopolitics and World News. The new global elections page gives you a calendar of upcoming world elections at a glance. 

• Market-Driven Opinion. As we’ve seen with our interview series, prediction market odds are a great starting point for opinion and debate. Keep an eye on your inbox for more interviews with people who sit on opposing sides of key markets. 

• New Verticals. We will be expanding our market depth in sports, economics, and pop culture, and more, so expect to see more coverage in all these areas. 

Write for The Oracle

To do this we need your help.

We have only begun to scratch the surface of what is possible with prediction market-driven journalism, and plan to expand The Oracle going forward. 

If you’re interested in writing for The Oracle, email oracle@polymarket.com with some info about yourself, what topics you are interested in covering and an article idea or two. 

We’re especially interested to hear from contributors with a track record of publishing in sports, politics, international affairs, economics, crypto, and data journalism.

* Special thanks to Ben Bell for help with the data analysis for this article. 

---

Disclaimer

Nothing in The Oracle is financial, investment, legal or any other type of professional advice. Anything provided in any newsletter is for informational purposes only and is not meant to be an endorsement of any type of activity or any particular market or product. Terms of Service on polymarket.com prohibit US persons and persons from certain other jurisdictions from using Polymarket to trade, although data and information is viewable globally.