On 24 February 2023, the first anniversary of Russia’s full-scale invasion of Ukraine, Émile Servan-Schreiber published Hypermind’s predictions for the war’s second year. He opened with a line from the nineteenth-century French journalist Émile de Girardin, for whom governing meant predicting. He then listed forecasting failures on every side: Putin expected a quick, cheap victory, Zelensky had doubted that Russia would invade, and few people expected Ukrainians to resist as they did. He offered the crowd’s numbers “with much humility”.
A forecast is a probability, not a promise, so no single call proves a crowd right or wrong. What can be judged is a record. Hypermind’s public forecasts on the war give one, starting a month before the invasion.
Before the invasion: possible, not likely
On 20 January 2022, Hypermind’s crowd gave about a 5% chance of an invasion before February, 25% before March and 30% before April. The most likely answer was “maybe later”. The next day, as Antony Blinken and Sergei Lavrov met in Geneva, the chance of an invasion before March rose to about 52%. By 26 January, the chance of one by the end of March had fallen back to about 35%, with 21% on February.
Russia invaded on 24 February. For most of the period the crowd treated an invasion before March as a real possibility, between roughly one chance in four and one in two, but it never made February the single most likely month. That is a partial miss. It was more cautious than the warnings Western governments were issuing at the time, and less cautious than those who called an invasion unthinkable.
The early misjudgment about Zelensky
The crowd’s largest error came just after the invasion. On the question “In 2022, will president Volodymyr Zelensky lose power in Ukraine?”, the chart Servan-Schreiber posted on 6 March 2022 shows the “yes” side near 90% in the last days of February. By 6 March it had fallen to 65%. He wrote that these numbers would have seemed crazy just two weeks earlier. By 18 March, the chance that Zelensky stayed in power had risen to 66%. By 7 April, the chance that he lost power was down to 19.6%, below the 29.7% given for Putin.
Zelensky stayed in office. The crowd was badly wrong for about two weeks and then corrected itself within about six. Updating as events unfold is how this kind of forecast is meant to work, but the early numbers were still wrong.
Calls that held up in 2022
- Putin. The chance that Putin would lose power in 2022 stayed between about 25% and 30% in March and April 2022. He did not.
- Nuclear use. The chance of a nuclear weapon being used to kill in 2022 was 11% on 6 March. It reached almost 20% on 1 October, its highest since the war began. None was used.
- Kherson. On 19 October 2022, the crowd gave Ukraine an 80.2% chance of controlling Kherson on 31 December. Ukrainian troops entered the city on 11 November, after Russia withdrew.
In September 2024, Servan-Schreiber revisited the nuclear question after a Washington Post column reported that the US intelligence community had told President Biden there was a 50–50 chance Putin would use tactical nuclear weapons in October 2022. He juxtaposed that reported estimate with Hypermind’s peak near 20%. It is a rare comparison, but not a controlled one: the intelligence estimate may have used classified information, its exact event definition is not public, and the two estimates may not have been recorded at the same moment. The fact that no weapon was used does not by itself establish which probability was better calibrated.
The Year 2 forecasts, and what followed
The February 2023 article made three groups of forecasts:
- Putin was likely (80%) to stay in power throughout 2023. He was given a 68% chance of losing power before Zelensky eventually did, and the two leaders were likely (73%) to hold no talks that year.
- Front lines would stay largely frozen. Russian-held Donetsk, Luhansk and Sevastopol were likely (70–90%) to stay under Russian control, and Kharkiv, Odesa and Kyiv very likely (86–95%) to stay Ukrainian.
- Nuclear weapons would almost certainly (96%) not be used. French Leclerc tanks were unlikely (19%) to reach Ukraine in 2023. The West was expected (78%) to eventually deliver the combat jets Kyiv was asking for.
The short-term calls went the crowd’s way. Putin stayed in power through 2023, the two presidents did not hold talks, the cities named stayed under the same control, and no nuclear weapon was used. On jets, Poland pledged MiG-29s in March 2023, and the first US-made F-16s arrived in August 2024. We have not found the exact wording of the jets question or of the Leclerc question, so we do not score either. The question on who would lose power first remains open.
