Joined by @chad and @mindme we do a practical bot dev roundtable, focused on movement and scouting reliability, and how to iterate faster with local testing. DogTato shares custom pathing work and a Docker plus SQLite regression test dashboard, while MindMe digs into rush intel edge cases and target prioritization. Main themes are pathing vs influence maps, early game intel bugs, and building tooling that makes failures easy to find.
Key takeaways
-
Prefer A* style pathing for air scouts when influence maps create jitter or tiny step moves.
-
Combine distance and threat cost, like expected DPS, then tune a threshold so a scout will take some risk.
-
Avoid micro stepping, pick farther waypoints so units keep momentum through turns.
-
Treat scout info as high value, a hallucinated Phoenix can be worth the sacrifice.
-
Fix intel false positives fast, bad ramp block detection can cancel expansions and derail the build.
-
Add target ignore and priority rules, ignore Adept shades for siege and burrow logic, prioritize carriers and brood lords over interceptors.
-
Batch testing pays off, 15 local games can surface rare crashes and show matchup trends quickly.