# Pengo score target provenance

304000 is a **numeric research target**, not a verified comparable world-record
threshold. Stefano supplied303000 as the reference and asked to take it as
given; his subsequent instruction explicitly requires comparable leaderboard
verification before making a record claim. No source/category has been
established for that comparison.

Current experimental identity:

- MAME0.289, set `pengoja`: Pengo (Japan,315-5010 type,rev C).
- Original archive SHA-256
  c5e17040c5cdb63e1bbc7e015bd00fb1eedbe7f0bf36bdccdac65104c0c596d8.
- DSW1=176, DSW2=204;3 lives, difficulty Medium, Upright, bonus life30000,
  rack-test cheat Off. Baseline: config/pengo/baseline.yaml.
- Best verified planner trajectory9790: ACT1/ACT2 completed, ACT3 reached;
  retained tape terminates during UNKNOWN. It is not a full-game record.
- Offline privileged trajectory search,0 model calls; two fresh native-speed
  reset replays, no load/RAM write/cheat. Complete machine-state identity UNKNOWN.

A comparable leaderboard must explicitly establish game revision/set,
difficulty/DIP semantics, starting lives/bonus-life settings, terminal scoring
rules and admissibility of this assisted offline trajectory method. Until
then, even reaching304000 would satisfy the numeric target only.

Diamond mechanics and values remain excluded from the objective until a
separate current-ROM experiment verifies them. No formula is imported from
another revision, memory, or a leaderboard. Immediate research priority is
ACT3 phase diagnosis, then the separately preregistered Three Acts, Zero
Deaths milestone, then Diamond mechanics, then Scoring planner.