A strategy game maker, and an honest note about the genre

In most genres you can get something enjoyable from a good space and simple behaviour. In strategy, the space is a backdrop and the enjoyment is entirely in the interlocking rules: production, cost, counters, range, vision, timing.

Worse, those rules only reveal themselves as good or bad when played against an opponent who exploits them. A number that seemed reasonable becomes the only viable opening. A unit that seemed weak turns out to invalidate three others.

So the genre demands many iterations, and the cost of an iteration is the thing that decides whether a strategy game ever gets finished by one person.

The rules are written by Flockbay as GDScript inside your project. Adding a unit type, changing a cost curve, giving a faction a different vision radius or making artillery ignore cover are all requests rather than afternoons.

That does not make the design decisions for you. Nothing can. What it does is let you take the twentieth pass, which is where strategy games actually become good, instead of stopping at the fourth because each pass cost a weekend.

You play, you notice the opening that always wins, you say so, and it changes. Not writing the code does not mean not seeing that the cavalry rush is unanswerable.

Terrain in a strategy game is not decoration, it is the argument. A chokepoint, a river crossing, a plateau with a view, a resource that sits uncomfortably far from safety — those decide how the game is played more than any unit stat.

Which instruments you get for that depends on an answer you give before anything exists: a new project asks 2D or 3D, shown as two pictures, with no default. Pick 3D and you shape terrain directly, place water, set the weather and the time of day, and put structures where they belong. Then you walk or fly through it and see whether the map is asking an interesting question.

Pick 2D and the map is a board, which is where a great deal of strategy has always lived. You drag to say which cells are ground and the tiles, the edges and the collision are all derived from that coverage, so a chokepoint costs a few seconds of painting rather than a sculpt. Whether a tile stops a walker is declared per material, which means a road and a wall are painted with the same gesture and only one of them blocks. Buildings and units are sprites you put at an exact position, turn and size with the mouse.

Looking down at that board is not a separate mode you switch into. It falls out of gravity: declare none and the character body steers on both axes with nothing to jump away from, which is the same code that runs a side-on platformer. What a unit costs, what it counters and where it walks on its own is Flockbay's half, exactly as it is in 3D.

It runs locally. Flockbay is a desktop app for Windows with Godot bundled inside, so a map with many units on it is limited by your machine rather than by a browser tab.

Computer opponents in strategy games are famously unsatisfying, and it is not because the code is hard. It is because a genuinely optimal opponent is unpleasant to play against, and a weak one is boring, and the interesting band between them is narrow.

The productive approach is to describe behaviour rather than skill. This faction is aggressive early and cautious once it loses a battle. This one turtles and expands late. Those are legible to a player, which is what makes an opponent feel like a personality rather than a difficulty slider.

Expect several passes here as well. It is normal, and it is the part where your judgement as a player is the only useful instrument.

More on this: AI strategy generator: generating scenarios, not verdicts

A strategy game is a set of decisions, and a decision the player cannot see is not a decision. Range, vision, threat and who is winning all have to be legible at a glance, and in 3D that is harder than on a flat board because the camera can hide things.

Practically this means camera height, unit silhouette and colour do more work than any mechanic. Two factions that look similar from above will ruin an otherwise good design, and no amount of balance work will rescue it.

Test it by playing a match and pausing at a random moment. If you cannot tell within two seconds who is ahead and what is about to happen, the problem is the presentation rather than the rules.

More on this: AI deckbuilder maker: synergies you can only find by playing

The realistic advice is to start small. Two factions, six unit types, one map. A finished small strategy game teaches you more than an unfinished large one and it is roughly the difference between a project and a folder.

Building needs an internet connection, because the chat, asset generation and world data come over the network. Playing what you built does not. Your project is an ordinary Godot project on your own disk, and Godot is open source.

The platform is free. Flockbay is free on your own AI: your Claude or ChatGPT plan, your own agent, or an API key. Flockbay builds from a library of ready-made objects, or with Godot's own shapes.