Developer builds a mahjong replay AI that estimates opponents' waits and tenpai
bdsqlsz · x · 2026-11-13
The author built BAKKAI, a mahjong replay lab powered by a defensive analysis model. Users import a full four-player game (with red fives) and review opponents' tenpai, waits and hidden-hand distributions from their own seat.
Features include post-game review with 3D/2D table replay, step-by-step model analysis at each discard, a match overview (round results, riichi, calls, starting scores), toggling opponents' hands, and keyboard shortcuts for stepping, jumping to key events, autoplay and switching between opponents' estimates. The page labels the analysis as experimental and uncalibrated.
More from Apps
- XERJ Reader-1 launches in preview: verifiable document Q&A at $0.01 per 1M input tokens — omarsar0 · 2026-11-14
- Haru: open-source realtime voice chat app inspired by the movie HER — cocktailpeanut · 2026-11-14
- Run a private AI fully offline on your phone with PocketPal AI in 5 steps — alex_verem · 2026-11-14
- Rand Fishkin hammers the point: AI citations are not recommendations — sanderssays · 2026-11-14
- Building an interactive 3D physics book on how planes fly with AI — techartist_ · 2026-11-14
- Nous Research ships Hermes Agent mobile beta build 44 with streamlined navigation and faster chat — NousResearch · 2026-11-14