I Built an AI That Picks Your Perfect Sport And You Can Try It Right Now

There are over 8,000 sports in the world. Most of us will only ever try three or four in our entire lives. What if the one sport that would make you happiest, keep you fit, and perfectly match your personality is something you have never even heard of? That question kept me up at night. So I built something about it.

1. The Problem Nobody Talks About

Think about how you picked the sport you play today. For most people, it was not really a choice. Maybe your school had a cricket team and that was the only option. Maybe your friends were into football, so you joined them. Maybe your parents signed you up for swimming classes because that was what was available nearby.

The truth is, most people end up in sports by accident, not by match. And here is the surprising part. Research in sports psychology consistently shows that people who play sports that genuinely match their personality and interests are far more likely to stick with it, enjoy it, and stay active for life. It is not about being the fastest or the strongest. It is about finding something that feels like it was made for you.

I have friends who tried five different sports before finding one they actually loved. I also have friends who quit sports entirely because they never found the right fit. They thought they just were not “sporty” people. But that is rarely true. More often, they just never met their match.

So I asked a simple question. Can we build a system that recommends the right sport for the right person?

2. How It Actually Works

The system asks you four simple things, and it takes about two minutes total.

1. Your interests — Do you prefer team or solo activities? Indoor or outdoor? Competitive or casual? Do you enjoy strategy, or do you just want to move? (12 quick sliders)

2. Your strengths — How do you rate your own endurance, speed, flexibility, coordination, and reaction time? No tests needed, just honest self-assessment. (8 sliders)

3. Your physical profile — Age, height, weight, and optionally things like your sprint time or jump height if you know them.

4. Sports you have tried before — This is crucial. It tells the system what to avoid in the discovery section.

Behind the scenes, a machine learning model processes these 29 inputs and produces three ranked lists: sports to play, sports to watch, and sports to discover (ones you have never tried but would probably enjoy).

Here is what surprised me most. Your interests matter more than your body measurements. When I analyzed how the model makes decisions, personal interests account for 35.4% of the prediction, while physical metrics like height and weight account for only 26.6%. This completely flips the old way of thinking. Whether you would enjoy rock climbing depends more on how much you love adventure and challenge than on how many pull ups you can do. Whether you would love chess boxing depends more on your appetite for strategy than on your current fitness level.

3 . The Discovery Feature — Finding Your Hidden Match

This is the most exciting part, and honestly, what makes this project different from every other sports recommendation tool out there. Most systems just suggest more of what you already know. If you like football, they suggest more football. If you run, they suggest more running.That is not discovery. That is just repetition.

SportRec goes further. It computes a “profile” for each sport based on the interests of people who actually play it. Think of it like a personality signature for each sport. Then it compares your interest profile against every sport’s signature. If there is a strong match with a sport you
have never tried, it flags it as a discovery recommendation.

The result: 85.7% of discovery recommendations are sports the user has never tried before. That is not a bug. That is the whole point. It is not about confirming what you already know. It is about opening doors you did not know existed.

Imagine being a football player your whole life, and the system suggests you try Ultimate Frisbee because you love team dynamics, outdoor activity, and strategy. Or being a runner and discovering that rock climbing matches your endurance mindset perfectly. That is the magic moment we are after.

4. Does It Actually Work?

I tested the model against several baselines to see if this approach actually holds up. The table below shows the key results.

The Stacking Ensemble, which combines multiple classical machine learning models, outperformed every other approach on ranking quality. When we add the Discovery mechanism, we intentionally trade some ranking accuracy for a dramatically higher novelty rate. That is the
trade-off. We want to show you sports you have never considered, even if the model is slightly less certain about them.

The live system uses both approaches: the Stacking Ensemble for “Play” and “Watch” recommendations where accuracy matters most, and the Discovery variant for “Discover” recommendations where novelty is the priority.

5. What I Learned Building This

This project taught me a few things that go beyond just machine learning. First, simple models often beat complex ones. I tried neural networks. I tried deep learning. In the end, a well-tuned ensemble of classical models performed best. Sometimes the fancy solution is not the right solution.

Second, deployment is harder than training. Anyone can train a model in a Jupyter notebook. Getting it to run fast, handle real users, and stay online is the real challenge. Building the API, optimizing inference time, and making the interface feel instant was easily half the work.

Third, and most importantly, personalization matters. Every person who has tried this tool has discovered at least one sport they had never seriously considered. Watching that happen never gets old.

6. Try It Yourself

The model is not just sitting in a research paper. It is deployed and live at gmora.dev website.

It is free. It requires no signup. It works on your phone, tablet, or laptop. It gives you results in under a second. And it supports Sinhala and Tamil, so it is accessible to students across Sri Lanka.

Whether you are looking for a new hobby, trying to stay fit, or just curious about what sport might secretly be perfect for you, give it a shot. The worst that happens is you confirm what you already suspected. The best that happens is you find your new passion.

This work has been accepted at the ICDSIAI-26 international conference. If you try it out and find a sport you love, I would genuinely love to hear about it. That would make this whole journey worth it.

This article was also published on my personal blog on Medium.