Donald Trump has improved his numbers for the fifth week. It rises in betting markets, in national polls and in key states. According to our average predictions, the Republican has a 56% probability of returning to the White House, compared to Kamala Harris’ 44%.
The following table shows the seven predictions that we follow in these weekly notes. The gap that I showed a week ago remains: on the one hand, the forecasters of Metaculus and the models based on surveys, which place Trump with only a slight advantage — they do not give him even 11 out of 20 options, almost a tie. However, the Republican is a favorite in the prediction markets: in Polymarket They see him as the winner with a 65% chance, double that of Harris.
(Note: I’ve said it often: the table above represents probabilities, not voting percentages.)
There are two essential messages to extract from this data. On the one hand, they tell us that the elections remain open and equal: none of these predictions translate into firm favoritism. Polymarket, which is the best for Trump, still gives Harris one in three wins. At the same time, the trend is clear: the numbers that have improved are those of the Republican.
We see this trend in multiple sources.
First sign: Trump rises in the prediction markets. In Polymarket, the injection of millions from user Fredi9999, who I told you about a week ago, has not been countered. The market has validated that bet, at least partially. Furthermore, it seems to be confirmed that he is a person. This is what the company has told The New York Times: it would be a trader French with “a lot of experiencein financial markets and services.” Trump’s progress is also seen in Metaculusa community of enthusiastic but no-money tipsters. There Trump’s chances are estimated at 53%, rising since two weeks ago.
Second sign: Trump has closed the gap in national polls. Three weeks ago, Harris had a three-point lead in the popular vote, winning 50% to 47% in estimated vote percentage. Now? The Democrat only has a one-point advantage. And remember that to win the presidency, Harris will likely have to win by a couple of points.
With this decline in the polls, why does Harris maintain between a 35% and 45% chance of winning the presidency? It is explained in the key states. Trump has also made progress in those territories, but not enough.
The situation in key states
As we already explained, the election actually depends on the result in each State. To win in November, candidates need 270 delegates, or electoral votes, adding those distributed by each place, which all go to the winning candidate of the constituency. Right now, Harris has 226 probable or certain electoral votes, which are almost the same as Trump, who has 219. The remaining 93 correspond to the so-called key states.
Trump has improved in all of these pivotal states. He has widened his lead where he already had it (Arizona, Georgia and North Carolina), he has cut it where Harris was leading (Michigan, Wisconsin and Nevada), and he threatens to turn Pennsylvania, which could be the decisive state. There the situation is a tie: the polls give Harris a few tenths, which is nothing.
That leads me to issue a final warning: we are looking closely at changes and trends that are real but small. Being the leader in Pennsylvania is a mathematical question that depends on two tenths of voting intention, the difference between 48% and 48.2%. The mistake normal of the surveys is 10 times higher. The final result could be different from a tie.
To illustrate this I have prepared one last table. It shows a series of scenarios and their probability, estimated by Nate Silver with his fantastic prediction model. In aggregate, Silver gives Trump a 53% chance and Harris a 47% chance. But I’ve sorted their 40,000 simulations into seven outcomes, based on how many key states Harris and Trump win. Well: do you know which scenario is most likely? Neither the tie nor the map scenario. The most likely scenarios are the extremes: that Trump wins all seven key states (happens 24% of the time) or that Harris wins them all (15%).
The key is to see that the States are connected. They correlate. If it turns out that the polls underestimate (or overestimate) Trump in this election, it is normal that they will do so throughout the country. That is why it is easy for all the hinge territories to end up red or blue.
It’s counterintuitive. When I say that the elections are close, I do not mean that the result necessarily will be. In reality, it is unlikely that the seven key states will allocate four for one candidate and three for the other. It only happens 20% of the time. Polls are inaccurate! They are not telling us outright that Trump and Harris will almost tie in electoral votes, but rather that they cannot see the winner. In its margin of error there is room for very different results: Harris can win by the slightest, perhaps thanks to Pennsylvania, or Trump can, but it is even easier for one of the two to surprise and take six or seven of the decisive states. .
