Observed arrival · 2026-11-14
Clocktower Bench puts AI models in the village
A benchmark pits AI models against each other in the hidden-role game Blood on the Clocktower and reports their results.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People comparing AI models on social deduction
- Worth noticing
- Ratings use mirrored role deals, with 90% intervals estimated through whole-pair bootstrap resampling.
Field notes
The public methodology says ratings are fitted over mirrored games, with 90% intervals generated from whole-pair bootstrap resamples; overlapping intervals are treated as ties. It separates performance as the town from performance on the hidden team, and reports setup effects such as a 12-player deal (+44 Elo) and a Poisoner in play (−37 Elo) against a ten-player reference deal. The latest replay is also presented with player roles, duration, and token count.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
⊠LoginAccess appeared gated
$PaidCommerce or pricing visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
ƒJavaScriptBrowser-side code central
One card from the complete issue