I'd never played a multiplayer snowball fight. So I made one.
One giant snowy arena, one server everyone shares, and nothing between you and your friends but some boxes and a lot of snowballs. I designed it, tested it, and shipped it to the real web, and I entered it in Vibe Jam 2025.
I wanted to make a video game, and I had no idea what it should be. So I started listing games I'd never seen before, and it hit me that I had never, ever played a multiplayer snowball fight. I was like, “huh.” So I started building one, and it turned into a really fun game.
Multiplayer was never even a question. I wanted to play it with my friends, and a snowball fight on your own isn't much of a fight. Because I was building with AI, multiplayer wasn't a scary extra on top. I could decide I wanted it, design how it should work, test it, and keep going.
I was already in the middle of building when my dad told me about the Bolt Vibe Coding Game Jam 2025, a game jam run by Pieter Levels in April 2025. The one rule that mattered: at least 80% of each game's code had to be written by AI. I thought, “Okay, why not?” and entered.
Over 1,000 games were entered. I didn't win (obviously), but submitting was fun, and the game is still in the jam's directory. Search “snowball” or “sports” on vibejam.com/2025 and it comes up, under my handle, @Oreo.
I'm a vibe coder, and I'm okay saying that. AI wrote most of the code. Everything else was me: the idea, the design decisions, which features went in and how they worked, all of the testing, and getting it running on the real web.
Ask for the feature or the fix I want, in plain words.
Ask for the feature or the fix I want, in plain words.
Read what came back. What does each piece do, why is it needed, and does it actually do what I asked for?
Most of my testing was just playing the game. For multiplayer, I pulled in a friend to run into me and throw at me.
Tell it what needs fixing, what I don't like, and what I do like. Then go around again.
I built the whole thing in Cursor. I didn't know about Claude or Codex back then.
I built it up in layers, each one on top of the last. Step through them:
By the end it wasn't a demo anymore. It was a game my friends actually wanted to play.
No rooms, no rounds, no lobby. Everyone joins the same server and spawns into one giant open arena, and you just fight. Take enough hits and you respawn, then jump straight back in.
I didn't build the map by hand, but the layout is my design. Drawn from above, not to scale.
Sprint is just normal game movement. The dash is the one I was excited about: launch yourself at someone and throw at the same time. It took the most refining, because I had to decide the rules. You can dash once you've left the ground, so it works in the air. But it doesn't start recharging until you land.
Why 10? I just like the number 10. Reloading takes time, so you can't spam forever, and you have to think about when to do it.
Every kill shows up in the feed, and there's a leaderboard. It isn't permanent: it's a kill count attached to each player for the session, and it's gone when you leave.
Pausing takes your avatar off the map, so stepping away doesn't mean standing there getting pelted. Then resume, or exit to the main menu.
A little practice yard with the game's real controls. Knock over the snowmen, then try the move the arena is built around: jump, and dash over the wall while you're in the air. It's a calm, top-down, one-player version of something much louder.
The real thing is a lot more chaotic. It's first person, and everyone shares one server, so there are snowballs coming from every direction, people dashing over walls, and somebody launching off a tree. Play it at sportsduel.app
The biggest problem by far was collisions. Players needed real hitboxes so they didn't walk straight through each other, and snowballs needed to actually register when they hit someone. That took the longest and the most testing of anything in the game.
You can't really test multiplayer collisions on your own. So a lot of it was me and a friend in the game at the same time: walking into each other, throwing at each other, checking whether the hits counted, and then going back to fix whatever didn't.
The other brand-new thing was putting the game on the real web. It runs on an external server my dad bought, and I'd never worked with one before. I figured all of it out myself:
It was a little hard. It wasn't as hard as the collisions.
I wanted different kinds of snowballs. They didn't make it into the final game, but here's what they were going to be, compared to a normal snowball.
More damage, but slower. Big commitment, big reward.
Less damage, but much faster and harder to dodge.
Exactly what it sounds like.
a normal snowball
I remember when my friend and I figured out the trees were bugged. If you jumped into a tree with forward momentum, whether you were running or dashed straight into it, its triangle geometry would catch you and launch you way up into the air. We laughed so much.
From then on we'd head to one corner of the map, launch ourselves off a tree, and rain snowballs down on everyone from the heavens. We did it every time we played, even in groups of four, five, or six.
The best bug I never fixed.
The practice yard above has the same bug. Run into a tree.
I showed the game to a new friend during some free time at school. We just played and laughed, and everything was normal. We had a really good time.
sportsduel.app
If you think this is a mistake, contact your school's IT team.
The school's filter had picked it up and figured out it was a video game. My game joined Cool Math Games, Poki, and all the big game sites on the block list. I was so proud. I made a game that got enough attention to get blocked by my school, and that's still one of my favorite things about this project.
Snowball PvP showed me I can build things that make the people around me happy. I put the effort in, and I made something my friends actually enjoyed, with a lot of good memories along the way. Even after it got blocked. (“Oh damn, it was so good it got blocked.”)
Not the code, not the jam. I built something, and it brought joy to the people who played it.