• Around the time of the podcast (immediately after the first debate and Trump’s COVID diagnosis), major betting and prediction markets generally implied something in the ballpark of a 30–40% chance for Trump. For example:

  • A Reuters report on October 1, 2020 noted that on Smarkets, Biden was around a 63% favorite after the first debate, implying Trump in roughly the high‑30s/low‑40s. (timesofindia.indiatimes.com)
  • A mid‑October Guardian summary of online bookmakers put Biden around 69% and Trump about 32%. (theguardian.com)
  • Crypto and exchange markets like Polymarket and FTX generally priced Trump in the low‑30s to high‑30s percent range in the weeks leading up to the election. (decrypt.co)

• Expert statistical models at the time disagreed strongly with those market odds:

  • FiveThirtyEight’s model, for instance, gave Trump only about a 12% chance by October 22, 2020. (fivethirtyeight.com)
  • The Economist’s model around late October put Trump as low as 4–14%. (decrypt.co)

• After the fact, several analyses argued that 2020 election prediction markets showed systematic irrationalities and arbitrage opportunities, particularly in the way they priced Trump even after it was clear Biden had won key states. (asteriskmag.com) However, these critiques focus on inefficiencies and bias; they do not pin down a consensus, ex‑post “true” pre‑election win probability for Trump.

• Because:

  1. The “true probability” of a one‑off event (Trump winning in 2020) is inherently unobservable;
  2. Credible models at the time spanned a wide range (roughly 5–40%) for Trump’s chances; and
  3. Post‑election research has shown that both polls and markets had notable flaws without definitively establishing that Trump’s real odds must have been above or below the ~30% level,

  we cannot rigorously determine whether Sacks was correct that Trump’s true winning probability was “somewhat higher” than the ~30% implied by markets.

Conclusion: The claim is about an unobservable quantity (true probability) in a context where contemporary and retrospective analyses diverge. Even with the benefit of hindsight, there is no objective way to say that the real probability was definitively above or below 30%. Therefore the prediction’s accuracy is ambiguous rather than clearly right or wrong.