Before you dive in
What you’ll find in the original
- Define configurations that vary planner, constraints, and selector independently before comparing outcomes.
- Hold map, race, opponent level, seeds, realtime mode, and game-time limit fixed across selected runs.
- Count missing replays and source or service drift as evidence limitations rather than making a single-variable causal claim.
Worth knowing
The selected macro table contains 50 attempts and 49 verified replays. The author explicitly says archived source and service-timing differences prevent a controlled five-way ablation; JevMade did not rerun the games.