Explicit Optimization of Global Strategy (Fixing a Bug in UDT1) — LessWrong