You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Commit 88c68f6
Browse filesBrowse the repository at this point in the historyBrowse files
Publish the measured balance, on a build where a seed reproduces a run
Fourth 24-seed run, and the first whose numbers are worth writing down: the
tiers are finally monotone.
CADET 15/24 boss damage 4.7 29% untouched 52.8s
PILOT 10/24 boss damage 4.9 14% untouched 47.2s
ACE 5/24 boss damage 5.6 2% untouched 46.3s
Against the baseline this whole series started from: 0.6 hull a boss fight and
57% of them finished without a scratch.
Per boss at PILOT, which is the table that did the work -- an average over five
bosses hides exactly the thing worth knowing, whether the first one anybody
meets is a fight or a formality:
S1 SENTINEL 2.3 hull 30s 8% untouched
S2 HIVE MOTHER 7.5 hull 66s 25%
S3 LANCE 2.3 hull 42s 18%
S4 BULWARK 3.4 hull 42s 7%
S5 WARDEN 9.7 hull 56s 7%
HIVE MOTHER, the sector-2 boss, measured 1.0 hull and half its fights untouched
before this series.
Stopping the tuning here. The ladder is monotone and the shape is right; going
further would be fitting the autopilot rather than the game, and the autopilot
has perfect reaction and no pattern memory, which is not what a person has.
Docs carry all of it now: the measured tables, the two economy leaks, the tier
knob that paid the player back, and the determinism bug -- including that its
first test passed with the bug deliberately reintroduced, because it traced only
the one boss that has no walls.
Screenshots regenerated. Full suite green.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AKLqS4PgnCXrzKHDH8Cnzy
0 commit comments