How a High School Student Built an AI Valorant Coach Project

BetterMind Labs
6 days ago
4 min read

Updated: 8 hours ago

Inside an AI Project by High School Students: How I Built a Valorant Coach Using Machine Learning

This AI project by high school students began as a personal frustration: being stuck in Platinum rank on Valorant. As a student in the BetterMind Labs AI program, I wanted to build something that solved a real problem—and this was it.

If you play any competitive game, you know the feeling. You’re "hard-stuck." For me, it was Platinum rank in Valorant. I’d watch pro players, study YouTube guides, and practice in the range, but in a real match, I’d make the same mistakes over and over. I knew I needed to improve, but I didn’t know what to improve.

My aim felt okay. My game sense felt… fine? I needed a coach. But not just any coach—I needed one that could watch every single second of my gameplay and give me completely objective, data-driven feedback.

So, I decided to build one myself using AI.

This AI Project by High School Students Turned Gameplay into Data-Driven Coaching

The core technology behind my AI coach is Computer Vision. In simple terms, it’s a field of AI that trains computers to "see" and interpret the visual world. My goal was to write a program that could watch a recording of my game and understand what was happening.

It needed to track key metrics:

My crosshair placement (Was it at head level?)
My utility usage (Did I waste my abilities?)
My positioning on the map (Was I standing in the open?)
My reaction time after spotting an enemy.

I started by building a basic script. The first version was clunky, but it worked. I played a few matches, fed the recordings to my AI, and eagerly awaited my first performance review.

My First Feedback Session: The AI Was Not Impressed

The AI’s first report was, to put it mildly, humbling. It presented me with a cold, hard list of my worst habits.

Crosshair Placement Score: 43% (The AI politely labeled this as ‘Requires Significant Improvement’).
Unnecessary Reloads Detected: 17. (Apparently, I reload after every single shot).
Time Spent Aiming at Feet/Torso: 58%.
Decisions Resulting in an Unfavorable Trade: 9.

It was brutal. There was no "You did your best!" or "Good try!" Just data. And the data was telling me I was not as good as I thought I was. The AI had exposed habits I didn't even know I had.

The Turning Point: A Mentor and a Machine Learning Model Changed Everything

My initial script was good at pointing out what I did wrong, but not why it was wrong in a specific situation. I was struggling to make the AI smarter, to give it real game sense. I was stuck.

Thankfully, I wasn't alone. I was working on this as part of my project for the BetterMind Labs AI/ML program, and my mentor had a game-changing idea. He said, "Instead of just programming rules, let's have the AI learn from the best."

We leveled up the project by introducing machine learning. We fed the AI hours of gameplay footage from professional Valorant players. The model learned to recognize high-level patterns: how a pro player would peek a certain corner, when they would use their utility, and how they would reposition during a chaotic fight.

My AI coach was no longer just a stat-tracker. It was now a comparison engine that could analyze my play against a "pro-level" baseline.

The Turning Point: A Mentor and a Machine Learning Model Changed Everything

The new feedback was even more detailed, and somehow, even more brutal.

"In Round 5, you dry-peeked the A-main entrance. A pro player in this situation uses a flash or recon ability 92% of the time. Result: You were eliminated."
"Your positioning on defense was 68% similar to a 'passive' playstyle, while the map control situation called for an aggressive hold. Result: The enemy team gained site control."

Did It Work? The Climb Out of "Hard-Stuck"

Armed with this hyper-specific feedback, I started to improve. Each week, I’d focus on fixing one or two things the AI pointed out. I worked on my crosshair placement. I stopped reloading unnecessarily. I thought more critically about my positioning.

Slowly but surely, it worked. The "Defeat" screens became "Victory" screens more often. My confidence grew. And a few weeks ago, I finally did it. I ranked up to Diamond.

Building my AI coach taught me more than just how to be a better gamer. It taught me that AI is one of the most powerful tools for self-improvement on the planet. It can be applied to anything you're passionate about—sports, music, art, science. It’s a way to see yourself without bias and find a clear, data-driven path to getting better. It was a brutal coach, but it was the best one I've ever had.