# ─────────────────────────────────────────────────────────────────────────────
# arms_movement_b2.txt — BATCH 2: the range axis ON THE WINNING ENGINE.
#
# Batch 1 (docs/movement_campaign.md) found that pure perpendicular strafe with
# the range tilt DISABLED (`strafe_notilt`) beats the shipped `tfil` on round
# wins across the frozen panel: Δwins/run +0.38 (9/9 decisive opponents,
# p=0.0039) with no detectable damage cost, and it does so while fighting
# FARTHER away (456 px vs tfil's 382). The shipped `tfil` came last of five arms
# on wins. This batch asks the obvious next question:
#
#   does the range the mover ends up at drive the win, or is the tilt MECHANISM
#   itself the cost?
#
# `strafe_notilt` has no range control at all and drifts to ~456 px. Re-arming
# the tilt with a target FAR (600) and NEAR (250) around that emergent 456
# separates "range value" from "tilt present".
#
# Format: name | ENV=value ENV=value | label
# ─────────────────────────────────────────────────────────────────────────────

tfil | | shipped baseline, re-measured in this session (replication)
strafe_notilt | TR_MOVEMENT=strafe TR_STRAFE_RANGE_TOL=999999 | Batch-1 winner, replication
strafe_325 | TR_MOVEMENT=strafe | strafe default (range 325), replication
tilt_600 | TR_MOVEMENT=strafe TR_STRAFE_RANGE=600 | tilt ON, target 600 (farther than the emergent 456)
tilt_250 | TR_MOVEMENT=strafe TR_STRAFE_RANGE=250 | tilt ON, target 250 (much nearer than the emergent 456)