Experiment: SAIDA and Iron as Terran benchmark-only SOTA opponents

[minsing-jin]

Goal

Add SAIDA and Iron to the benchmark experiment plan without assuming runtime commandability.

Rationale

SAIDA and Iron are strong Terran bots, but strength does not imply they can safely accept runtime intent injection. They should start as benchmark-only or launch-config targets until source/build access proves otherwise.

Tasks

• Locate authoritative source/binary availability and license constraints.
• Record BWAPI versions, map/runtime assumptions, and launch requirements.
• Define how they can be used in commanded-vs-baseline matches as opponents.
• Identify whether any launch-time strategy configuration exists.
• Explicitly classify as Level 0 benchmark-only or Level 1 launch-config unless source-level hooks are proven.

Acceptance Criteria

• SAIDA and Iron entries exist in the benchmark matrix.
• Runtime command injection is not claimed without source hook evidence.
• Match-plan support requirements are documented.
• Replay/log artifacts needed for verifier comparison are documented.