tibia

Tibia Project Overview

Tibia is a robust typed Tibia-rules simulator developed in Spiral, aimed at generating code for C, Rust, and WGSL (used for GPU applications) from a single source. The simulator is utilized to train agents that are evaluated against the top open-source Tibia bots using a standardized and reproducible benchmark known as the OT Agent Arena.

Core Features

The simulator encompasses a variety of modules and functionalities, including:

• Simulation Modules: Comprising 27 .spi modules located in src, along with standards checked by eoie standards.
• Application Components: Containing 15 files in apps/arena, which includes 9 Lua probes and corresponding shared client modules, as well as a server-map tool found in apps/map.
• Rich Rule Set: Governs mechanics such as geometry, vitality, coins, random number generation (rng), cooldowns, damage, combat interactions, creature behaviors, experience gain, skills, blessings, and wiring functionalities.
• Typed Domain: Features aspects like quantity (phantom units), readiness (GADT witnesses), vocations (indexed spells by vocation), catalog (existential spawns), views (HKT mapped and counted), and terrain layouts, particularly highlighting the scanned arena spot.
• Decision-Making Framework: Includes policies for agent behavior, neural policymaking based on integer networks, and evolutionary search strategies.
• Diverse Simulation Methods: The simulator offers a variety of entry points such as hunt simulations, multi-creature simulations, and a live Lua entry for OTClient.

Recent Developments

As of October 21, 2026, numerous findings have emerged from the simulator's performance evaluations, particularly with respect to live clear times and potion usage. The simulator has demonstrated impressive predictive capabilities:

• It accurately predicts clear times during combat and estimates potion usage effectively, with adjustments made based on live testing feedback.
• In various testing scenarios, the simulator’s calculations provided substantial insights into decision-making processes, revealing that tuning interventions generally match or exceed neural policies.

Fidelity and Live Measurements

The simulator's fidelity is bolstered by extensive live measurements and comparisons across different mechanics. Key findings regarding mechanical accuracy include:

• Knight and creature engagements align closely with real-time game data, demonstrating that the simulation effectively mirrors live game behaviors.
• Upgrades to the simulator have enhanced performance metrics, with a notable improvement in potion consumption rates and enemy engagement behaviors.

Results and Evaluation Protocols

Simulation results are based on strict performance protocols and detailed analytics comparing simulated outcomes against live event data. Key insights include:

• Targeting improvements highlight that focusing on current targets enhances kill counts significantly.
• Evaluations of healing strategies and terrain navigations have yielded optimal approaches for different combat scenarios.

Conclusion

The Tibia simulator stands as a critical tool for developing enhanced artificial agents within the Tibia universe. By leveraging sophisticated algorithms and real-game dynamics, it not only fosters innovation in bot development but also enriches the overall gaming experience in the world of Tibia. This unique blend of advanced simulation, real-time data accuracy, and comprehensive performance evaluation positions Tibia as an essential project for both researchers and developers in the gaming industry.