Why hello Reader,
The first time I lost to a cannon rush, it felt like my bot had got hustled out of its ladder points. One minute I am watching it go through the build I coded and then I am watching its charred husk, burned out under a pile of dumb cannons. All I could do was shake my head. Though truth be told I doubt if I had been playing I could’ve done better. That’s the thing about cheese: it looks like nonsense, but it works because it shrinks the game into something brutally simple. A tangle of thousands of choices gets reduced to a few sharp questions. For a bot, that clarity is gold. Rage-inducing for the opponent, yes, but also the perfect training wheels. Your Bot can punch way above its weight, even become a champion.
|
🔁 In Case You Missed It
The Grand Finals of 2025 Season 1 happened this past weekend and though I highly suggest to watch Zozo’s game against Negativezero as a cheese masterclass. I would be remiss if I didn’t recommend you to watch MicroMachine take out a Masters level human player in the exhibition. You’ll see how a bot can deploy surprise to beat humans.
|
So when we say cheese, we’re talking about a strategy or tactic that relies heavily on surprise, exploiting niche or match up specific weaknesses. It naturally puts you as an aggressor, so you’ll end up doing an all-in attack that could win the game very quickly but will leave your bots at a significant disadvantage if it fails. In the early stages of a new bot it cuts down the initial things you need to worry about significantly. Suddenly you don’t need to code with gas in mind. You can ignore 3/4 of the tech tree. You don’t need to think about unit compositions, or match ups. What you need to get right is
- Managing your economy
- Building a few requisite buildings
- Producing the unit or offensive structure
- Attacking
- Microing for added effectiveness
That’s it. Suddenly a very complex game can be whittled into 5 things to code for. Which is a manageable constraint. In the case of a worker rush you can take it a step further and bring it down to just focusing on attacking and microing.
You can go a long way with just cheese if you keep perfecting it. Zozo, this past season’s champion started off as a predictable, yet effective worker rush bot. From there it evolved to eventually include a macro game but it didn’t stop growing its cheesy options. It was a proxy zealot rush that won him the crown, taking out Negativezero, arguably the strongest ProBot. From cheese to champion. not bad eh?
|
🗒️ ./run Notes:
Zozo’s championship run showed how far cheese can take you. If you want to try it yourself, here’s a proxy zealot rush I took from a bot named Chance distilled down:
Build order:
- Make probes until 17
- Send one probe across the map as your proxy worker
- Drop a Pylon near their base (hidden but close enough)
- As soon as it finishes, start a Gateway (add more up to 4 if you can afford it)
- Chrono boost Zealots nonstop
- Rally them straight at the enemy, micro to keep them alive as long as possible
Pseudo-logic:
- If supply ≥ 17 → send a probe across the map
- Then build a Pylon at proxy location
- When Pylon is done → drop a Gateway
- While minerals allow → train Zealots
- Always rally units to enemy start
Key points:
- Reserve one worker as the dedicated proxy builder
- Keep your proxy location 20–25 range from their natural ramp (close enough for fast pressure, far enough not to be scouted instantly)
This example is Protoss-only, but Chance is an open-source bot packed with cheesy builds for every race — you can browse them all here
|
|
