# Which rest pose does the RUNNING GAME agree with? GP_TITLE entry 7 (title_jp).
# 2026-08-30. Oracle: docs/re/captures/title-builds/live-title-jp-at-rest.png
#
# The two candidates are ptlogo_eff3.t32's resting pose under the two keyframe
# eras -- the ONE element that differs between them on this screen:
#   stale era (origin/main)              rest (108,72)
#   fixed era (formats-pin-2026-08-30)   rest  (98,42)
# Scored over the 388x423 bbox where the two renders differ (74 934 px), so
# the measurement is not diluted by the ~92 % of the frame that is identical.
#
# CONTROL 1 -- vertical alignment found by sweep, not assumed:
#   offset  0  RMSE 87.29 | 25  83.39 | 40  56.37 | 45  32.41 | 50  53.08 | 60  72.84
#   sharp minimum at 45, which is the known game-surface offset in a 1280x720 frame.
# CONTROL 2 -- the bbox discriminates: the same box scored against a DIFFERENT
#   screen's capture (the EN title) gives 98-103, against 40-58 here.
# CONTROL 3 -- --black changes nothing (58.412/41.690 either way): every pixel in
#   this bbox is covered by an element, so the canvas never shows through it.

## The answer
  stale era  rest (108,72)    RMSE 58.412
  fixed era  rest  (98,42)    RMSE 41.690   <-- the game agrees with the FIXED era
  fixed era  --settle t=213   RMSE 40.210

## The noise scale: sweep the screen's own timeline with --at
  t=  0  RMSE  77.974
  t= 15  RMSE  78.737
  t= 30  RMSE  78.942
  t= 45  RMSE  63.053
  t= 60  RMSE  71.123
  t= 75  RMSE  66.222
  t= 90  RMSE  58.504
  t=105  RMSE  54.741
  t=120  RMSE  45.486
  t=135  RMSE  41.276
  t=150  RMSE  40.586
  t=165  RMSE  40.197
  t=180  RMSE  40.336
  t=195  RMSE  40.252
  t=210  RMSE  40.230
  t=225  RMSE  40.141
  t=240  RMSE  40.068
  t=255  RMSE  58.878
  t=270  RMSE  78.414
  t=285  RMSE  78.423

# The capture sits on a broad PLATEAU, t=135..240, flat to 1.2 RMSE across
# 105 units, with sharp rises outside it (78 at t=0, 78 at t=270).
#
# So: the stale-vs-fixed margin (16.7) is ~14x the plateau's own flatness and
# is decisive. The settle-vs-rest margin (1.5) is INSIDE that flatness and is
# not. This capture cannot separate settle from rest; it separates the eras.
