WinBolo

The 1987 tank game that refused to die is back, and it's GPL

Stuart Cheshire wrote Bolo in 1987. By the mid-90s it was the multiplayer game on Mac, the one running over AppleTalk in computer labs while someone pretended to study. John Morrison built a Windows port in 1998. It was the only one Cheshire ever officially blessed, and a community grew around it that ran tournaments and traded maps for the better part of a decade. Now it's on Steam, free, cross-platform, with the source headed to GitHub under GPL v2 shortly after launch.

That last bit matters more than the Steam listing. This isn't a remaster studio strip-mining nostalgia. It's the original Windows port author rebuilding his own thing for modern hardware and putting the source in public.

What you're actually playing

WinBolo is a top-down tank game for up to 16 players on a single island. You capture bases, which are the only places you can refuel, rearm, and repair. You capture pillboxes, which will murder anyone caught crossing open ground. You harvest trees. You build roads to move faster. You wall off chokepoints. You lay mines.

Then somebody detonates one of those mines. It leaves a crater. The crater fills with water. The water turns into a river. And the map you started fighting on is genuinely not the map you're fighting on an hour later. Trees regrow, slowly. Burn through them too fast and you can't repair your pills. That's the loop.

Two modes. Freeform, where alliances form and dissolve mid-match and stabbing your ally in the back is just Tuesday. Or tournament, teams locked before the match starts, no diplomacy at all, two sides grinding until one holds every base.

The 2.0 rebuild, by the numbers

The technical sheet is where this gets interesting, because Morrison clearly took the rebuild seriously instead of slapping a Steam wrapper on the 2008 codebase. According to the project's own docs, the netcode is now authoritative-server with lag compensation done via position-history rewind, client-side prediction, and an adaptive jitter buffer. Shells, in other words, should land where you aimed them. Not something the original WinBolo could always promise over a 2002 DSL line.

Self-hosting uses UPnP and NAT-PMP to punch ports automatically, with UDP hole punching to reach servers tucked behind home routers. One codebase across Windows, macOS, Linux, Steam Deck, iOS, Android, and the web, all on shared servers. Seventeen languages, including in-game chat.

The AI work surprised me. There's a new brain called GoalHunter (the site calls it NewAutopilot, same thing) that runs on goal priorities, threat trajectory simulation, and territory influence maps. Brains are written in Lua, with a visualisation debugger. For anyone who wants to go deeper, the simulation is exposed as a steppable C loop with Python bindings, Stable-Baselines 3 and PyTorch support, and ONNX export, so you can train a model and ship the result as an in-game opponent. That's a real ML harness, not a marketing line.

The map editor is similarly overbuilt in the good way: flood fill, magic wand, four seeded procedural generators (Tournament, Natural, Maze, Fractal), text-as-terrain, image import, symmetry scoring, 100-deep undo, stamp library. Logs and replays live inside the client with WinBolo.net integration so you can scrub through other people's matches without leaving the game.

The "free, open source" question

The business model is unusual, so worth being clear. The game is genuinely free on Steam, the App Store, and Google Play. There's an optional Supporter Pack for people who want to fund development. The full source lands on GitHub under GPL v2 after launch, continuing the lineage from when the original WinBolo opened up in 2008. Morrison even pitches the Supporter Pack at, among other people, the players who never quite mailed in their shareware fee back in 2001. Fair enough.

GPL v2 has implications worth thinking about. Anyone can fork it. Anyone can run a server. The community can outlive the developer, which for a game this niche is the only realistic survival path.

What the old community does with it

The existing video footage, going back to the WinBolo.net days, is mostly sped-up tournament replays and pillwar demos, narrated by handles like JollyRoger explaining how to take a pillbox without dying. That's the audience this is built for, and the audience it has to re-find. Not streamers chasing trends. People who learned pill micro twenty years ago and still remember the maps.

Where this fits, and where it doesn't

The comparisons write themselves badly. It's not Worms (no turns), and it's not World of Tanks (no progression grind, no monetization). The closest living thing is probably Atomic Tanks, or maybe the indie revival scene around games like Liero, but Bolo's persistent-world base capture plus dynamic terrain puts it somewhere genuinely on its own. The base-and-pill economy, the way the map physically rewrites itself through play, the diplomacy in freeform, none of it has a clean modern analog. Which is both the appeal and the marketing problem.

The reach problem nobody talks about

A free, open-source, top-down 2D tank game with a 1987 lineage is not going to win a Steam algorithm fight. The launch firehose will bury it inside hours, because Steam's discovery system rewards wishlists and concurrent players and aggressive trailers, none of which a niche revival starts with. A generic "indie multiplayer" push would burn whatever marketing budget exists on impressions from people who'll bounce in fifteen seconds, because the game looks, from a thumbnail, like a free flash game from 2004.

The audience that will actually stick is narrow and specific. Original Bolo veterans from the Mac and early Windows era. The Lua and open-source modding crowd who care about GPL v2 and the ML harness. Retro multiplayer YouTubers who cover oddities like Subspace/Continuum and Netrek. Map-editor power-users who migrate between games like this and Warzone 2100. The small but real Steam Deck couch-multiplayer scene. Reaching those people one by one is the only sane play, and CreatorFetch is the kind of infrastructure that lets a one-person studio actually execute that kind of targeting at launch instead of praying the algorithm notices.

Worth your time?

If you played Bolo on a Quadra in 1994, yes, obviously, go download it. If you didn't, the honest answer is it depends on whether the loop above sounds interesting in itself. The graphics are modernized SVG/PNG, but the game still looks like what it is, a top-down 2D tactical tank game. No campaign. No progression carrot. Servers are the game, and a multiplayer revival lives or dies on whether servers stay populated past the first month.

What it has going for it: a real technical rebuild, a developer who's been with this codebase since 1998, GPL source, mod tooling that's actually serious, and a design nobody has successfully copied in thirty-eight years. That's a reasonable hand to play.