Hierarchical Macro Strategy Model for MOBA Game AI
Authors: Bin Wu1206-1213
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We perform comprehensive evaluations on a popular 5v5 Multiplayer Online Battle Arena (MOBA) game. Our 5-AI team achieves a 48% winning rate against human player teams which are ranked top 1% in the player ranking system. In this section, we evaluate our model performance. We first describe the experimental setup, including data preparation and model setup. Then, we present qualitative results such as attention distribution under different phase. Finally, we list the statistics of matches with human player teams and evaluate improvement brought by our proposed model. |
| Researcher Affiliation | Industry | Bin Wu Tencent AI Lab benbinwu@tencent.com |
| Pseudocode | No | The paper describes the model architecture and components but does not provide any structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | To train a model, we collect around 300 thousand game replays made of King Professional League competition and training records. Finally, 250 million instances were used for training. |
| Dataset Splits | No | The paper mentions collecting game replays for training ('250 million instances were used for training') and evaluating against human players, but it does not specify explicit training/validation/test dataset splits (e.g., percentages or counts for each). |
| Hardware Specification | No | We used CAFFE (Jia et al. 2014) with eight GPU cards. |
| Software Dependencies | No | We used CAFFE (Jia et al. 2014) with eight GPU cards. |
| Experiment Setup | Yes | Batch size was set at 128. The loss weights of both phase and attention tasks are set at 1. We used ADAM as the optimizer with base learning rate at 10e-6. |