Llm game coach improves advice by tracking player and game states
TactiNavi: Serving and Revising an LLM Game Coach with Explicit Strategy Conditions
Artificial Intelligence
Summary
Giving good game advice is tricky because what works for one player or version of a game may not work for another. The authors designed TactiNavi, a system that carefully tracks when certain strategies should be used based on the player’s situation and game updates. This helps the coach give more accurate recommendations and avoid mistakes. Their tests show that this approach improves correctness from under 60% to over 80%, and they explain how to handle uncertain cases and check fixes carefully.
What this means in practice
- •For game developers: Create smarter in-game coaching systems that tailor advice to players’ specific states and game versions with explicit strategy rules.$Commercial implications: Enables commercial deployment of more accurate and maintainable AI coaching agents for open-world games from structured strategy conditions.
- •For customer support teams: Improve automated response systems by explicitly encoding contextual conditions to serve correct advice based on user state and product version.
Authors
Qi Liu, Xiaoyang Yuan, Zhuomeng Zhang, Yubin Ruan, Xinyi Mou, Wenjin Wang, Hanying Jiao, Di Wu, Jinke Li, Jingxuan Hong, Ke Feng, Zixun Sun
Abstract
A deployed LLM coaching agent answers requests whose correct response depends on player state and game version, so advice that is right for one player can be wrong for the next. We describe TactiNavi, the serving system behind the LLM coach of a commercial open-world game, which makes strategy applicability an explicit artifact shared by offline compilation, online execution, failure diagnosis, and release. An offline compiler turns versioned game evidence and expert guidance into strategy packages with explicit prerequisites, the serving plane evaluates those prerequisites as three-valued conditions over structured player state and records decision traces, and a control plane uses the resulting condition-rule dependencies to validate revisions on held-out traffic before release. In a retrospective evaluation of 1,122 logged coaching turns from three deployment periods, condition-bearing packages raise correctness from 57.9% to 82.0% over atomic facts and reduce strategy misapplication under changed player state from 18.1% to 10.4%, while trace-based claim verification reaches 86.4% precision at a median reply latency of 2.07 s. The deployment also showed that most apparent fixes are generation noise, because an unchanged-system rerun recovers 38.4% of failing turns and only one of 15 shipped revision units passes the release checks retrospectively. We distill these results into lessons on treating applicability as a first-class serving object, sizing traces for later diagnosis, and building the null model into the release path.