No. 19-46 · argued May 4, 2020
United States Patent and Trademark Office, et al., Petitioners v. Booking.com B.V.
Post-argument lean
From the argument alone, before the decision
76%
chance the petitioner prevails, from how the bench divided its words
Respondent won Adjudged to be AFFIRMED. Ginsburg, J., delivered the opinion of the Court, in which Rob...
The Justices’ words, by the side they were spoken to
47%
53%
To the petitioner · 1,587 words, 48 turns48 turns, 1,791 words · to the respondent
The bench spoke 1.1× as many words to the respondent as to the other side. Since OT2017 the side that draws more of the bench’s words has tended to lose, so this argument reads above the usual 70% petitioner win rate. Here the lean pointed the wrong way.
Each Justice’s questioning
Lean beside the vote each Justice cast
| Justice | Words to the petitioner ◂ │ ▸ to the respondent | Lean to petitioner | Voted for |
|---|---|---|---|
| Roberts, C.J. | 72%+3 | Respondent ✗ | |
| Thomas | 63%+3 | Respondent ✗ | |
| Ginsburg | 61%+3 | Respondent ✗ | |
| Breyer | 71%+7 | Petitioner ✓ | |
| Alito | 70%+9 | Respondent ✗ | |
| Sotomayor | 68%+9 | Respondent ✗ | |
| Kagan | 62%+2 | Respondent ✗ | |
| Gorsuch | 70%+4 | Respondent ✗ | |
| Kavanaugh | 74%+5 | Respondent ✗ |
A Justice’s lean is the chance they vote for the petitioner, from their own word split and the whole bench’s; the small figure is the change from their usual petitioner rate. Click a row to pick out that Justice in the transcript.
Listen and read
176 turns
Press play, or click any line of the transcript to start there
The Court’s recording · each line timed to the audio
Loading the transcript…
From the Court’s official transcript.
October Term 2019, lean against result
Each dot an argument; click one to open it
27 of 40
leaned 75% or more to the petitioner, and the petitioner won
0 of 0
leaned toward the respondent, and the respondent won
35 of 55
decided cases where the lean pointed the right way (always picking the petitioner: 35)
About this page. Built from the Court’s own transcript and recording.
- The lean. Backtested leave-one-Term-out on 511 decided arguments (OT2017–OT2025): 72% of outcomes called right against 70.5% for “the petitioner always wins”, with forecast error (Brier score) 0.197 against 0.209. It refines a strong base rate more than it flips calls. Each Justice’s lean: 67.5% of 4,408 votes against 63.4% for their usual rate.
- Words, not questions. Every Justice turn in an advocate’s time counts toward that advocate’s side; amicus time counts for neither. Counting turns adds nothing once words are counted, and the raw rule “more questions loses” does worse than the base rate, because petitioners argue first and have rebuttal.
- Timing. Each line is timed to the recording: speech recognition matched the Court’s audio to its transcript, and its own text is used for nothing but the clock. Click a line to hear it.
- Method and findings: docs/argument-transcripts.md.