Deep reinforcement learning for real world problems
Dota 2 is a popular Multiplayer Online Battle Arena (MOBA) video game. As an Esport, Dota 2 has a prize pool of over USD$40 million in 2021 for its annual flagship competition. Strategy plays a vital role in determining the outcome of games, and teams are constantly looking for means to gain a compe...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/163306 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | Dota 2 is a popular Multiplayer Online Battle Arena (MOBA) video game. As an Esport, Dota 2 has a prize pool of over USD$40 million in 2021 for its annual flagship competition. Strategy plays a vital role in determining the outcome of games, and teams are constantly looking for means to gain a competitive edge.
This work attempts to explore prediction models based solely on the team compositions at the start of a game. In essence, it attempts to predict which team is favoured before actual gameplay begins. Thereafter, we attempt to train and evaluate an AI agent to play the drafting game using Monte Carlo Tree Search. We use data from real matches obtained from the STRATZ API endpoint. |
|---|