Overview
- Researchers advise asking questions that split the remaining suspects as close to 50/50 as possible, such as 8–8 from 16 or 7–8 from 15.
- For certain totals the ideal split shifts, with examples including an 8–10 split from 18 and a recommended 1–3 question when four suspects remain in some situations.
- The team warns against early questions about rare features like glasses because they remove too small a subset to be efficient.
- They say three-part “tripartite” questions can improve outcomes, though the constructions are cognitively awkward and risk confusing players.
- The analysis is presented in a brief arXiv preprint titled “Optimal play in Guess Who,” and the authors have released a legally distinct online game to practice the strategy.