Chamath’s claim has two parts:

1. As of June 20, 2020, Trump has about a 25% chance to win (75% to lose).
   This is a probabilistic assessment of his personal beliefs; there’s no direct way to verify a subjective probability after the fact. However, we can compare it to contemporaneous forecasts:

   • Early and mid‑2020 quantitative models from The Economist and others generally gave Biden a large advantage, often in the 80–90% range, implying Trump around 10–20%, not 25%.(newsweek.com)
   • Some macro/market analysts around late October 2020 independently assessed Biden at ~75% and Trump at ~25%.(rbcgam.com)
     So his 25% estimate wasn’t outlandish, but it also wasn’t uniquely correct or directly testable. The directional part ("he loses") did match reality: Biden ultimately defeated Trump, 306–232 in the Electoral College.(en.wikipedia.org)

2. Prediction that the odds would move to roughly 55% Biden / 45% Trump as Election Day approached.
   This is the concrete, testable part. Looking at major prediction markets and forecasts near Election Day:

   • On June 3, 2020, before his comment, betting markets like Smarkets and PredictIt already had Biden only modestly ahead, around 50–53% vs 43–46% for Trump (roughly 55/45).(cnbc.com)
   • By late October, a macro memo summarizing prediction markets and models put Biden at ~75% in markets, and 90–96% in FiveThirtyEight and The Economist forecasts, leaving Trump at only 4–10% in those models.(rbcgam.com)
   • On November 2, 2020 (the day before the election), PredictIt still priced Biden around 63% and Trump around 40%.(forbes.com)
   • On Election Day itself, European bookmakers had Trump’s chances around 40% and Biden’s around 60%.(euronews.com)

Across both major models and widely used betting markets, Trump’s closing chances clustered roughly in the 10–40% band, not at the near‑coin‑flip 45% that Chamath predicted. The race did tighten somewhat compared with early pandemic expectations, but not to the 55/45 level he forecast; expert models in fact moved toward a larger Biden advantage as Election Day neared.(rbcgam.com)

Because the specific forward‑looking, falsifiable part of his prediction (that the odds would end up around 55/45) did not match the actual consensus odds, the overall forecast is best scored as wrong, even though his directional call that Trump was more likely than not to lose turned out to be correct.