SirStone/SirRoboGarage
1
0
Fork 0
Code Issues 28 Pull Requests Actions Projects Releases Wiki Activity
392 Commits 5 Branches 0 Tags
7516ca47c0c81edb8a80ba2a66b4bd0b674f65cb
Commit Graph

392 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
SirStone 7516ca47c0 docs: quick-recap README for bitbrain gun (inputs/outputs, knobs, verdict) 2026-09-26 14:47:42 +02:00
SirStone 8f44140783 j136 add TR_MODULE_* on/off switches + single [modules] boot inventory 2026-09-26 14:46:29 +02:00
SirStone fd34e4ae4a j135 draw-only geometry overlay (TR_GEO_DEBUG) + global debug-draw switch (TR_DEBUG_DRAW) 2026-09-26 14:29:46 +02:00
SirStone 3903ec71b4 j134 ledger: note phantom_meteor is dead code (not selectable), out of scope 2026-09-26 13:01:36 +02:00
SirStone 3237b65f40 j134 ledger: propagated fire fix to all 5 movers via one shared fire_tracker (TR_FIRE_FIX), per-mover catch table 98.888->100, live tick alignment measured (correction belongs on event_tick+2) 2026-09-26 12:59:35 +02:00
SirStone 882733819b j134 fix: tracker must take the mover's switch per-call (instance flag was silently off on the object-literal construction path) + apply the correction on the reading that carries the server's turn-N+1 energy change (two-slot buffer); live-verified alignment 2026-09-26 12:56:26 +02:00
SirStone 6ad5d99922 j134 fire fix: share ONE fire_tracker across tfil/ring/strafe/learned/surf (TR_FIRE_FIX, default on); env-gated TR_FIRE_DIAG alignment trace 2026-09-26 12:40:19 +02:00
SirStone bd57db6d63 j133 strafe: correct the comment's measured fire count (67065, catch 98.888%->100.000%) 2026-09-26 12:25:06 +02:00
SirStone 3bee4353b2 j133 ledger: appended 'Missed fires + the label question' (catch 98.888->100%, 746/67065 shots were blind: 456 by the server +3*power bonus, 290 by our own same-tick damage; exact label still -0.230, state-conditional model -0.347 -> the observable state is the constraint) + report fixtures 2026-09-26 12:22:31 +02:00
SirStone 799438bbe9 j133 strafe fire detection: correct the enemy energy delta for the server's +3*power hit bonus and our own damage, never drop a too-large drop (TR_STRAFE_FIRE_FIX, default on); catch 98.89%->100% of enemy fires on the 70-battle corpus 2026-09-26 12:18:21 +02:00
SirStone 0566264b06 j132 strafe display check: also assert the chosen tile's ramp border and value plate reach the SVG 2026-09-26 11:54:14 +02:00
SirStone cc332138b3 j131 learned movement: real bullet-endpoint resolution (TR_LEARNED_REAL_EVENTS, default off) + exact-geometry Gate A/B (inversion NOT fixed; state still the constraint) 2026-09-26 11:54:11 +02:00
SirStone d5047598b5 j132 strafe display check: also assert the chosen tile's heat number reaches the SVG with the heat grid ON and OFF 2026-09-26 11:53:10 +02:00
SirStone 4db46f02dd j132 strafe pick quality: chosen tile's heat in the [strafe] log line (heat=h/10.0 alt=best-other ok=0|1) and ON the tile (green/yellow/red ramp + value), legible with TR_STRAFE_HEAT_GRID=0; deterministic display check vs pathMaxHeat 2026-09-26 11:51:33 +02:00
SirStone 8dd9b3b3b5 j130 learned movement outcome label: live panel results (no arm beats strafe; label swap is a dead heat; gap is information) + Gate A report 2026-09-26 11:35:47 +02:00
SirStone 61def1c3e9 j130 learned movement: outcome label (P(hit|state,g)) mode + Gate A; pre-registered outcome arms 2026-09-26 11:27:03 +02:00
SirStone 4c934a67c1 j129: add entry-point check for the virtual-bullet overlay draw proc 2026-09-26 11:03:14 +02:00
SirStone 752d3a3829 j129: virtual-bullet debug overlay (TR_VBULLET_DEBUG, default off) - travelled path, predicted aim ring, hit/miss vector; shared turret colour table 2026-09-26 11:02:39 +02:00
SirStone a5c893ace6 learned movement: record the pre-registered prediction as partly wrong (right on wins, wrong on the hit-rate mechanism) 2026-09-26 10:56:02 +02:00
SirStone 4f332dcb95 learned movement (SBC): live panel results and the direct answer (module is a wash vs strafe on wins; state conditioning is a CI-separated hit-rate gain vs the same mover without it; the label is the wrong quantity) 2026-09-26 10:55:14 +02:00
SirStone ee98827a62 learned movement: danger-map alignment diagnostic (corr(P(arrival bin), P(hit|bin)) = -0.342) in the gate + ledger 2026-09-26 10:45:26 +02:00
SirStone b4e54fc3cd j127: allocation batch results - forced shares do not beat pattern; allocation not a lever (closed axis) 2026-09-26 10:45:10 +02:00
SirStone a436e9f16a j128 learned movement: state-conditional counted-SBC wave danger (TR_MOVEMENT=learned, default-off), offline gate + pre-registered panel arms 2026-09-26 10:42:10 +02:00
SirStone 2a3a62a0ab j127: fix TR_RACK_SHARE round-boundary dwell lock (bot.tick resets, tracker currentSince persists) - share path only 2026-09-26 10:29:45 +02:00
SirStone 6fd5fe6328 j127: forced-share allocator (TR_RACK_SHARE, default-off) + allocation batch pre-registration 2026-09-26 10:13:22 +02:00
SirStone 2e336452e6 HeadOn gun colour: red -> pure white (#FFFFFF) 2026-09-26 09:50:08 +02:00
SirStone e25a00175a Pattern gun colour: orange -> green (#00CC44 / #66FF99) so it is not confusable with BitBrain's pink 2026-09-26 09:40:22 +02:00
SirStone 2d8b7d7875 j126: read bot settings from a .env file (default ./.env, --env-file flag, TR_ENV_FILE); file wins over shell leftovers, boot report labels (source: .env) 2026-09-26 09:08:39 +02:00
SirStone 8e109bae0b closing j125: verify movement ship end-to-end (clean build SuccessX, full 16-test guard suite all counts match, real default+revert battles) and restructure gun/movement ledgers outcome-first 2026-09-26 05:06:49 +02:00
SirStone cf77d0d647 j124: spinner + fair-melee results doc; analyzer convergence tests 2026-09-26 05:05:09 +02:00
SirStone 5fe28574ab gun j123 Batch 2 outcome + phase-2 final answer: len6 does not replicate (wash on 33 opponents); the shipped Pattern is already tuned on its match-length/history parameters; no default changed 2026-09-26 04:54:40 +02:00
SirStone 1b59b65810 gun j123 Batch 1 outcome + Batch 2 prereg: len6 is the only arm to beat pattern on wins (15 opp), but all 5 arms are wins-positive incl. the bearing-invariant rad_offset; pre-register the 33-opponent confirmation with a structural control 2026-09-26 04:22:49 +02:00
SirStone 6c21dc4c58 j124 fair melee: add strong-field melee harness + driver (legacy champion field) 2026-09-26 04:08:44 +02:00
SirStone 06da7abb0e j124 spinner claim: add ConstantSpinner adversary + spinner A/B arms, panel, analyzer 2026-09-26 04:05:36 +02:00
SirStone 2a98aba91b gun j123 Task A+B prereg: expose Pattern's TR_PATTERN_LEN/TR_PATTERN_DEPTH (default parity) and pre-register the 6-arm match-parameter sweep 2026-09-26 04:01:51 +02:00
SirStone 467e07a6a4 gun ledger: fix the arm count in the final answer (six configurations, not seven) 2026-09-26 03:55:32 +02:00
SirStone 2a5ea5c17b gun j121 Batch 2 outcome: the wider panel kills the damage hint (bitbrain wash, tmhorizon WORSE); onlyPattern confirmed; phase closed 2026-09-26 03:54:52 +02:00
SirStone c343c00aaf gun j121 Batch 2: correct the SpinBot path in the wider panel before any battle (it is not under /tmp/tr_bots) 2026-09-26 03:35:11 +02:00
SirStone 3fd6db97e7 movement ship: gate v2 fresh-data primary passed (sign-flip p=0.045, CI [+0.02,+0.58]); default TR_MOVEMENT flipped to strafe 2026-09-26 03:19:33 +02:00
SirStone 087e26955f gun j121 Batch 1 outcome: onlyPattern CONFIRMED across the 15-opponent panel; pre-register Batch 2 wider-panel calibration 2026-09-26 02:58:41 +02:00
SirStone 514674886d movement gate v2: pre-register the sign-flip primary test on fresh 300-battle data (before any battle) 2026-09-26 02:44:22 +02:00
SirStone feefc1912e movement final: confirmation result and ship decision (gate failed on the sign-test leg; default NOT flipped) 2026-09-26 02:41:08 +02:00
SirStone 1d8143a15e gun j121 Batch 1: pre-register the 6-arm Pattern-panel tournament and the decision rules 2026-09-26 02:22:38 +02:00
SirStone ff03e81591 movement final: pre-register the ship criterion before the confirmation battles 2026-09-26 02:21:09 +02:00
SirStone 47ce244740 movement j119 Batch 3+4 results: the reversal/dwell and heat-field axes are a clean negative; nothing beats strafe 2026-09-26 02:18:22 +02:00
SirStone 1256357f07 movement j119 Batch 3+4: pre-register the reversal/dwell and heat-field arms 2026-09-26 01:40:03 +02:00
SirStone 7311aaef5a movement j119 Task A: make the shipped TFIL heat shape env-overridable (default path byte-identical) 2026-09-26 01:38:51 +02:00
SirStone c07e6d9599 movement ledger: the post-hoc structure of the strafe win (60 points, both batches) 
56/60 opponent-level win deltas are non-negative and no opponent family
regresses reproducibly (the one WallAvoider loss reverses in Batch 2).
corr(dwins, d_hit_rate) = -0.09, corr(dwins, d_damage) = +0.38: the aggregate
win is a survival effect, but a per-opponent hit-rate gain does not predict a
per-opponent win gain - so hit rate stays an explanation, never a proxy.
 2026-09-26 01:33:24 +02:00
SirStone 44e2d191e3 movement ledger: fix two claims in the Batch-1/2 prose 
- tfil is 4th of five on round wins, not last (ring is nominally 0.04 lower,
  ns) - the Batch-1 commit message overstates one word; the correction is
  recorded in the ledger rather than rewritten.
- the Batch-2 direct answer quoted three of four CIs excluding 0; it is four of
  four ([+0.04,+0.63], [+0.16,+0.60], [+0.27,+0.89], [+0.22,+0.72]).
- added the cleanest aggression isolation of Batch 1 (ring - ring_notemp, same
  engine and heat field, range weighting alone): +41.9 dmg/run, -0.11 wins/run,
  +12.8 pp incoming hit rate at 236 vs 395 px.
 2026-09-26 01:32:28 +02:00
SirStone 7d3645da5a movement Batch 2: the strafe win over shipped tfil replicates; the range target decides nothing 
Same frozen panel, same 3x3 design, new session on commit 8efa627 (no source file
changed since 1984a78, so the same code), 225 battles, 0 invalid runs. Arms:
tfil, strafe_notilt, strafe_325 + the tilt re-armed at 600px and 250px.

Paired vs tfil: strafe_325 +0.58 wins/run [CI +0.27,+0.89] 11/12 p=0.0063;
strafe_notilt +0.47 [+0.22,+0.72] 10/11 p=0.0117; tilt_600 +0.40 [+0.04,+0.76]
(sign test 8/11 p=0.23, sign-flip p=0.049); tilt_250 +0.38 [+0.07,+0.69] 10/12
p=0.039. Incoming hit rate -5.2..-6.9 pp with 0/15 opponents favouring tfil.

The range TARGET is not the lever: re-arming the tilt moved the achieved
distance from 459px (no steering) to 478px and 415px, and none of the three is
separable on wins. This overturns Batch 1's reading that the tilt costs wins -
the honest statement is that the tilt's win effect is below this design's
resolution. tfil reproduced to within 1.4 pp (40.7% -> 39.3% of rounds), so the
baseline itself is stable across sessions.

Ledger: Batch 2 section, the verbatim analyzer report, a data-driven
what-to-try-next, and the session log.
 2026-09-26 01:31:08 +02:00