Scott Alexander said his biggest failure was item 10 of his 2018 predictions, where he gave Roe v. Wade only a 1% chance of being overturned.

A factual claim about the 2018 predictions post. True: the post (item 10 or nearby) gives a 1% probability that Roe v. Wade would be overturned within the 2018-2023 window. Mostly true: it gives 1% but the item number or the exact wording of the horizon differs slightly. Mostly false: it gives Roe's overturning a low probability other than 1% (e.g. 5%), or phrases the item as a related proposition. False: the post has no such prediction, or gives a high probability.

Set before the research began.

1. E1 Predictions For 2018 Slate Star Codex · 6 February 2018 Own record

   “10. No major health-care reform legislation gets passed: 95%”

2. E2 2018 Predictions: Calibration Results Slate Star Codex · 22 January 2019 Own record

   “8. [This was missing in original]”

3. E3 Mantic Monday 5/9/22 Astral Codex Ten · 10 May 2022 Own record

   “By the way, in 2018, I got this horribly wrong - I said only a 1% chance of Roe v. Wade getting repealed in the next 5 years.”

4. E4 Links 7/18: URL Mountains Slate Star Codex (Scott Alexander) · 30 July 2018 Own record

   “I think my previous 99% certainty that it wouldn’t was inexcusably far too high, but still expect the court to avoid doing it openly.”

5. E5 Brett Kavanaugh likely gives the Supreme Court the votes to overturn Roe. Here’s how they’d do it. Vox · 10 July 2018

   “there is likely a 5-4 majority on the Court to overturn Roe v. Wade”

The quoted item 10 from the 2018 predictions post (E1) is about health-care reform, not Roe v. Wade. No quoted line from that post shows a Roe v. Wade prediction or its probability. The only support for the 1% figure is Scott Alexander's own later blog posts: E5 mentions his "previous 99% certainty" that Roe would not be overturned, and E3 says he gave it "only a 1% chance" over 5 years. These are ordinary statements by the subject, not the original post's text, and no independent source confirms the figure (E6 only discusses the Court's votes).