Hello everyone,
I’m interested in developing an AI to assist with competitive and technical analysis of fighting games. I’ve noticed that video analysis/interpretation is still a very advanced field at the moment. The feedback it provides on how to improve is often based on a data source without much context, resulting in generic and context-free advice.
To address this, I’ve coded a mod that can transform Street Fighter 6 matches into a script. I’m wondering if it’s possible to teach the AI to “read” the script, or if an API or more advanced code is necessary.
My specific questions are:
- Is it possible to train an AI to analyze fighting game scripts without an API?
- What are the challenges involved in teaching an AI to interpret fighting game scripts?
- What resources (e.g., tutorials, libraries) are available to help me with this project?
It seems there has been some confusion regarding the AI’s function.
The goal is not to create a bot that plays the game, but rather to provide me with data in a faster and more consistent manner.
This is why I am looking for ways for the AI to “read” the game, either through video analysis or by extracting specific information from the game itself.
Here are some examples of the data I am interested in collecting:
- How many times did I lose neutral?
- How many times did I fail to punish something I could have punished?
- How many normal pokes did I miss and hit?
- How many times did I win neutral against the opponent?
- Frame data for each move
- Hitboxes and hurtboxes for each character
- Startup and recovery frames for each move
- Information on character states (e.g., crouching, jumping, etc.)
By collecting this data, I hope to:
- Gain a better understanding of the game’s mechanics
- Identify patterns and trends in my own gameplay
- Develop strategies to improve my gameplay
I believe that data analysis can be a valuable tool for improving fighting game skills. I hope that by sharing my work, I can help others to improve their gameplay as well.
Any insights or advice would be greatly appreciated. Thank you in advance for your help!