What the top SC2 bot does that yours doesn’t

Why hello Reader,

I checked PiG_Bot’s ladder stats the other day and welp, I am sure you know the story. Win one, lose a couple.

Naturally if you look at the top MindMe’s negativeZero sits at number one, consistently. I see him at our community round table all the time.

So I asked him straight up: “What am I missing?”

Turns out I was optimizing the wrong thing.

Here’s what he told me.

Builds, Builds, Builds

MindMe says builds are 90% of NegativeZero’s success. Not micro. Not ML. Not even pathing.

The edge? Obsessive build timing. Probe arrives at the first pylon the instant money hits. Every Gateway, every Cybercore starts as fast as possible.

Those seconds snowball. That’s the gap between top 10 and number one.

Layered counters, not if/then spaghetti

No giant if/then chains. NegativeZero uses weighted layers.

Say the enemy has tempests.

  • Bottom: stalkers
  • Middle: phoenix + stalkers (once stargate is up)
  • Top: tempest (once fleet beacon is done)

Bot starts at the bottom, climbs as production unlocks. Each unit gets a weight tied to supply. Five tempests? Multiply by five. Everything pools together, reduces to a ratio. That’s the comp.

If you don’t know counters, that’s fine neither did Mindme. He copied his first counter values off Liquipedia. You can keep it that simple

The strategy triangle

Mindme laid out three postures: Aggressive. Defensive. Economical.

Enemy turtles? Expand. Free lead. Enemy rushes? Defend. Never attack into a defense.

How does the bot tell? Cannons go up, that’s a turtle.

Advice for new bot builders

You don’t need the shiniest tools. MindMe’s combat sim doesn’t factor in range. Still works. Get the function in place, polish later.

Don’t over-plan. Just write code. See what happens.

You can hear exactly how he thinks with my interview with him 🔽

🛠️ Workshop.log

Remember when I vibe coded a full SC2 bot? The prompt I used to keep the AI on rails, I cleaned it up with help from the community and made it yours.
It covers everything: architecture, micro, macro, scouting, data safety, even how to stop the AI from going rogue on your codebase. Fill in your race, your project name, your install path, and you’ve got a starting point that actually holds up.
This is the same structure I’m building PiG_Bot on. Adapt it, break it, make it better. Make it yours

This Week’s Community Wisdom

May the Bugs Be Ever In your Favour🪲

Email Preference:
Unsubscribe | Update your profile | 600 1st Ave, Ste 330 PMB 92768, Seattle, WA 98104-2246