It looks like an action game. You control a character called MegaMan.EXE on a 3×6 grid of panels, three rows deep, split down the middle — your side, their side. Viruses move and shoot in real time. You dodge along rows and fire back.
But you're not playing an action game. You're playing a deck.
Before the fight you build a folder of Battle Chips. During the fight, every few seconds, a Custom gauge fills and the game stops dead to deal you a random hand from that folder — five chips, and you can only select the ones that share a letter code or a name. Pick your combination, unpause, and now you have a sword, a cannon and nothing else until the gauge fills again.
That tension is the whole design. Real-time positioning on a grid, turn-based resource management layered on top, and a deck you assembled hours ago deciding what options you get right now.
I've been rebuilding that system for phones. It's out now, it's free, and it's called Mega Man: Advanced PET. This is a post about the design problems, because they turned out to be more interesting than the code.
The input problem
The GBA had a d-pad. This matters more than it sounds.
Battle Network's combat assumes you can move exactly one panel, in a chosen direction, at a moment of your choosing, while a projectile is already travelling toward you. The d-pad gives you that for free. Four discrete directions, tactile, no ambiguity, no visual occlusion.
A touchscreen gives you none of it. Every option is a compromise:
A virtual d-pad puts a fake control under a thumb that has no way to feel where it is. Players drift off it constantly, and they only notice when an input doesn't register — which is exactly the moment they were dodging something.
Swipe gestures read cleanly when you swipe deliberately. They read terribly when you're panicking. Fast repeated swipes get merged, misinterpreted, or eaten by the gesture recogniser's timing window.
Tap-the-destination-panel is precise and unambiguous. It also puts your thumb directly on top of the six panels you most need to see.
There isn't a clean answer here. Every touch control scheme for a real-time grid game is choosing which failure mode you can live with. That's worth saying plainly, because a lot of mobile ports pretend the problem doesn't exist and ship a virtual d-pad without comment.
The Custom screen is the part that survives translation
Here's the good news: the half of Battle Network that's a card game ports beautifully.
The Custom screen is already a menu. It already pauses the action. It already expects deliberate, considered input rather than reflexes. Touch is better at this than a d-pad — you tap the chips you want instead of cursoring across a grid of them.
Advanced PET has 273 chips, and they fuse the way they should: by matching code letters or matching names, chaining into Program Advances. The full chip library is browsable on the web, searchable by element, power and rarity, if you want to look at the numbers without installing anything.
What makes chip fusion interesting as a design object is that it makes folder-building a constraint puzzle rather than a power calculation. A folder of thirty individually excellent chips with thirty different codes is worse than a folder of mediocre chips that all share the letter B, because the second one actually combines when it's dealt to you. You're not optimising for chip strength. You're optimising for the probability that your random hand contains a legal combination.
That's a genuinely good system, and it survives the platform change completely intact.
Making a fan game feel occupied
Battle Network had hub areas — cyberspace plazas where other NetNavis stood around. In the originals they were NPCs with two lines of dialogue each.
Advanced PET has The Square, which is the same idea with actual people in it: a shared real-time space where players walk around as their own Navis, with name plates, live chat, a trade bazaar, and community goals. It runs on Supabase.
The design constraint that shaped this most was that event viruses spawn deterministically — every client computes the same spawns at the same time from shared state, rather than a server telling each client what appeared. That's a meaningful architectural choice for a project without dedicated game servers, and it's what makes a shared world possible at all when the thing coordinating it is a database with a realtime channel rather than an authoritative simulation.
The trade-off is that determinism has to hold under bad mobile connectivity, which is a harder guarantee than it sounds.
Co-op raids sit on top of the same social layer. You invite people from your BrotherBand friend list, the boss scales to how many of you turn up, and rewards are shared across everyone who joined. There's also local Bluetooth NetBattling, which exists purely because sitting next to someone and linking up was a real part of what the GBA games were.
Some smaller decisions worth defending
Your step count charges the Cyber-Meter. Real-world walking fuels battles. This could easily be a gimmick, and in a monetised game it would be — but here it's the thing that gets you to open the app when you're not already playing it.
No pay-to-win, no ads. The game is free and there's no monetisation at all. This isn't generosity, it's the only defensible position for a fan project built on someone else's IP.
Colour-blind-friendly UI and adjustable text. Battle Network leans heavily on elemental colour matching — fire beats wood, wood beats electric — and a game that encodes core mechanics in colour alone is unplayable for a meaningful chunk of players. Haptics and text scaling are in for the same reason.
The website doesn't load YouTube until you press play. No Google content, no cookies, until you actively ask for the trailer. Small thing, but it's the difference between claiming to care about privacy and implementing it.
On the legal question, because someone always asks
It's a fan game. It's free, there's no monetisation of any kind, and it isn't affiliated with or endorsed by Capcom. Mega Man is their trademark.
Fan projects on active IP live at the owner's discretion, and everyone building one knows it. The only honest position is to not monetise, not pretend it's official, and accept that the calculus belongs to Capcom rather than to me.
Where it is
Out now on the App Store and Google Play, free, in English, German and French.
Sixty seconds of actual gameplay:
If you played the originals and have opinions about what I got wrong — particularly on the input scheme, which is the part I'm least settled on — the Discord is where that conversation happens.
I'd genuinely rather hear that the controls don't feel right than not hear it.
Top comments (0)