TASK 2 - power selection, a clear win. bestPower used an ABSOLUTE
MinHitRate = 0.40 bar. Measured per-bin virtual rates (rolling-100 fraction)
show no bin ever clears 40%, so 11 of 14 guns were stuck at bin 0 (power 1.0)
even where higher bins were comparable:
Linear p1.0 44% p1.5 39% p2.0 30% p3.0 29% old bin 0 -> new bin 3
Accel p1.0 44% p1.5 40% p2.0 26% p3.0 29% old bin 1 -> new bin 3
Pattern p1.0 50% p1.5 40% p2.0 27% p3.0 12% old bin 1 -> new bin 2
Replaced with a scale-aware PowerBarFrac = 0.50 (a dimensionless FRACTION of
the gun's own best bin rate). 13 of 14 selections now pick heavier bullets.
Real effect vs DrussGT (8 rounds x 3 runs): hit rate unchanged (7.56% ->
7.47%) but damage dealt +52% (157 -> 239 per run) and rounds end faster.
Same accuracy, half the shots, half again more damage.
TASK 1 - the TM pattern-classifier gun does NOT earn its slot. It was built as
a mixture of experts with a corrected-Granmo TM as a multi-class gate over
HeadOn/Linear/Circular/WallBounce/Accel, labelled by which expert's prediction
was closest to the actual enemy position (an exact, supervised, per-shot
label - no delayed credit). Offline it loses to the best of its OWN experts on
essentially every fixture, and against DrussGT it cost real performance:
baseline (path+relative) 7.56% real hit rate, damage 157
+ power fix 7.47%, damage 239
+ power fix + TM gun 5.59%, damage 133
The gun was selected on 806 ticks and fired 24 real shots at 4.2%.
So the tree ships with EnableTmSelector = false: code and wiring kept intact
for re-enabling, but it is not in the active rack.
Worth recording from the clause dump: the gate DOES latch onto meaningful
structure. On energy-threshold-turner, HeadOn's clauses key on the energy bits
(the rule's own driving variable) while Circular keys on distance/velocity. So
the TM is learning something real and interpretable - it simply cannot beat
'always pick the best expert'. Root cause (INFERRED): the closest-expert label
is noisy because several experts are near-tied, and under the path metric the
winner varies by power bin while the gate sees one shared per-tick input, so a
one-vs-rest gate over a saturated 870-bit clause space has no margin to exploit.
(Zero-padding the 2-frame window was tried first and saturated every clause at
256-755 included literals; alternating the two real frames fixed that.)
Also factors the corrected feedback into an exported tmLearnDir and exports the
encoding/TM primitives; the Tsetlin tests still reproduce the documented
mean=13.8 included literals, so the refactor is behaviour-preserving.
Verified: 33/33 guard checks, tsetlin tests green, metric checks green, new
power-selection guard green (13/14 selections change; relative bar still picks
bin 1 and not bin 3 for a [30,25,12,5]% profile), 12/12 offline==online
acceptance under the shipped default.
The selection thresholds were calibrated for a rate scale that does not exist.
MEASURED on an exact offline replay of a fogged live WorldState vs DrussGT
(1397 selection ticks), the 0.10 absolute floor fires on 53.0% of point-metric
ticks and forces HeadOn, which has a REAL hit rate of 2.0-4.4% - worst or
near-worst of 13 guns. HeadOn's selection share: 69.1% (abs+point) -> 43.5%
(rel+point). My earlier claim that the floor fires ALWAYS is REFUTED - it is
53%, because bestRate is a max over gun x bin and a >=50-sample bin
occasionally clears 10%. The mechanism is confirmed; the literal statement was
not.
Scale-aware mode (GUN_SELECTOR_MODE, absolute|relative, default relative):
RelTieMargin = 0.20 dimensionless FRACTION of bestRate, replacing the
fixed 2pp band so the band scales with the metric
FloorPeakFrac = 0.25 the floor fires iff bestRate < 0.25 * peakRateRef,
SelectorWindow = 256 where peakRateRef is the field-best rate over the last
256 selection ticks - keeping the original 'don't trust
a collapsed field' purpose but only when the field is
bad RELATIVE TO ITS OWN RECENT BEST, and counting only
guns with >= MinObsBeforeCompete samples so cold-start
100% spikes cannot pin HeadOn
also pools the rate over power bins instead of taking the max over bins, so
one lucky bin no longer wins
absolute mode is preserved byte-for-byte for rollback.
A/B vs DrussGT, real server hit rate, 3 runs x 10 rounds per config, one frozen
binary:
absolute+point 3.66 / 2.45 / 5.01 pooled 3.76%
absolute+path 7.55 / 8.21 / 6.83 pooled 7.57%
relative+point 7.66 / 6.18 / 5.79 pooled 6.59%
relative+path 7.15 / 7.55 / 6.90 pooled 7.21%
absolute+point is SEPARATED from all three (p < 0.0001); the other three
OVERLAP each other (p = 0.18-0.64). So the METRIC is the dominant lever and
under path the two threshold models are statistically tied.
DEFAULT SET: metric = path, thresholds = relative. absolute+path was nominally
0.35pp higher but indistinguishable (p = 0.64); relative is the principled
scale-aware fix, is the only model that works under BOTH metrics, and prevents
the point-metric catastrophe if anyone switches back. Shipping absolute would
ship the accidental side-effect this work exists to remove.
STILL NOT SOLVED: the selector remains only a moderate ranker.
Spearman(virtual rank, real rank) is 0.52 for the winning config, 0.36 pooled
for path and 0.04 for point - and it is INCONSISTENT across run sets. The
metric switch won by de-selecting HeadOn, not by ranking guns better. That is
the next problem.
TASK B, report only: do NOT drive selection from raw real hit rates yet.
Only the selected gun fires, so unselected guns get near-zero real shots
(GuessFactor 20, Linear 24 vs HeadOn 733); noise is fatal (n=470 at p=10% gives
+/-2.8pp, most guns n<200 gives +/-5pp+ across a 3-15% spread); and real rate is
conditional on when the gun was selected. A blended signal with forced
exploration and shrinkage is defensible in principle but needs thousands of
shots per gun across many battles. Real rate is best used OFFLINE as the
evaluation metric - which is exactly what this A/B did.
RELATED BUG FLAGGED, not fixed: MinHitRate = 0.40 in bestPower is on the same
wrong scale - no bin ever clears 40%, so once every bin has data, power
selection falls back to bin 0 (power 1.0) late in a round.
Verified: 33/33 guard checks, 11/11 metric checks, tsetlin green, 12/12
offline==online acceptance under the shipped default, run_range rc=0 over 20
fixtures. Adds analyze_selector.nim to measure floor/tie/bestRate/HeadOn-share
per config on any fixture.
Adds GUN_VBULLET_METRIC (point|path, default point = unchanged behaviour) so
the virtual-bullet hit model can be selected at runtime with no rebuild. Both
the live tracker and the offline replay read the same value, so the 12/12
offline==online acceptance holds under EITHER setting (verified for both).
A/B AGAINST THE LIVE BOSS, real server-side hit rate as ground truth, 5
battles x 12 rounds per metric on one frozen binary:
point 4660 shots / 219 hits = 4.70% (per-run 3.16-5.53)
path 4834 shots / 359 hits = 7.43% (per-run 6.55-8.24)
The distributions DO NOT OVERLAP: path's worst run beats point's best run.
+2.73pp, +58% relative, z = 5.56, p < 0.0001. Range distributions were
identical (~460-478 px), so this is not a range confound.
MECHANISM - and this is the important part. The gain is SELECTION, not better
gun learning. Under the point model every gun's virtual rate is compressed
into 0.6-4.4%, so HeadOn sits inside the 2pp tie margin and takes 72.6% of
selection ticks / 76.9% of shots - while HeadOn is 11th of 13 by REAL hit rate
(2.3%). The path model widens the band to 4.7-13.7% and ranks HeadOn 10th, so
its shot share falls to 35.9% and Pattern/Accel/WallBounce get picked instead.
Counterfactual: applying the point model's per-gun real rates to the path
model's shot mix yields 7.65%, i.e. essentially the whole observed gain.
So the selector, not the guns, is where the win lives.
PER-GUN REAL HIT RATE vs DrussGT (path mix, the answer to 'which guns are
worth keeping'): WallBounce 10.8, Pattern 10.5, Accel 10.0, Displace 9.3,
Circular 9.2, AvgLead 8.5, KNN 5.7, StopShot 5.2, GuessFactor 3.7,
Tsetlin 2.9. Per-gun N is small (hundreds of shots) so single-gun ordering is
indicative, not definitive.
TWO CAVEATS, recorded because they undercut a naive reading:
1. One adversary. DrussGT is a wave surfer and HeadOn is genuinely bad against
surfers, so part of this may be matchup-specific.
2. The path model is NOT a better general ranker. Spearman(virtual rank, real
rank) is 0.52 under point vs -0.04 under path. It wins by accidentally
fixing HeadOn's mis-rank, not by ranking guns better. A more durable fix is
to address the selection logic directly - which is the next job.
Also adds a focused guard test (test_vbullet_metric) covering parsing/default,
a receding-target point-miss/path-hit, a perpendicular-target path-miss, and
replay determinism.
Verified: 33 guard checks, 12/12 acceptance under both metrics, tsetlin tests
green, range 34.3% (point, unchanged) / 50.8% (path).
The previous fix (learn the residual against a constant-velocity base) was
structurally right but cost us on real wave-surfing movement: GF 108 -> 55,
KNN 101 -> 74 on the classic DrussGT captures. Root cause: the linear base is
a poor model for a surfer, so the residual histogram is noisier than the old
total-lead histogram.
FIX: blend the RANGE between a radial-only forecast and the geometric one by
radialFrac (the fraction of recent per-tick motion that is radial), keeping
the constant-velocity bearing. dist = radialDist + rf*(linearDist - radialDist).
New VelocityTracker in common_libs/guns/lead_forecast.nim; the window default
is 32 and results were identical at 16 and 40, so it is not tightly tuned.
Nine candidate bases were measured and rejected WITH NUMBERS rather than by
argument, which is why I trust the winner:
velocity scaling 0.8 recovers DrussGT but destroys wall-bounce 241 -> 20
radial-only range excellent DrussGT, wall-bounce 241 -> 140
short-window averaged vel worse than both bases outright
hard reversal/speed gates help DrussGT, lose nothing, but weaker than blend
radial-fraction blend best on BOTH <- shipped
Result (hits per 2000; classic-5 = classic DrussGT captures, tr-5 = the new
closed-loop TR captures, synth-10 = the rest):
base classic-5 GF/DGF tr-5 GF/DGF synth-10 GF/DGF
current(prefix) 108 / 108 41 / 39 1302 / 1302
linear(postfix) 55 / 76 9 / 4 2702 / 2692
BLEND 171 / 100 86 / 87 2717 / 2703
Strictly better than both on classic-5 GF and on every synthetic bucket. The
one figure below the old base is classic-5 DecayGF (108 -> 100, -8/2000,
within noise) and that is stated plainly rather than hidden.
TASK B - enemy energy in learners. KNN gains an 8th feature, enemyEnergy/100,
on a FIXED [0,1] scale (not min-max) because threshold behaviour keys off
absolute energy. Honest result: it is NEUTRAL on the target fixture (77 vs 77)
and roughly neutral in aggregate. The base change, not the feature, moved that
fixture. Tsetlin already encoded enemyEnergy and now scores 88/400 on
energy-threshold-turner against Linear's 43/400 - a 2x margin, which is the
'can a TM learn a high-level pattern' question answered in gun form.
TASK C - is the virtual-bullet metric itself faithful? Quantified: scoring the
bullet's PATH against BotRadius instead of the single point at aim distance
raises every gun by +31% (GF) to +86% (HeadOn), so the current model is
PESSIMISTIC, and it RE-RANKS materially: Linear 9th -> 6th, AvgLead 7th -> 3rd,
GuessFactor 4th -> 9th, DecayGF 6th -> 12th. The 12/12 offline==online
acceptance still holds under the path model (verified with a temporary env
hook driving both sides), so no red flag. VERDICT: do NOT switch. The point
model is the standard virtual-bullet PREDICTION-ACCURACY fitness - the bullet
must arrive at the predicted point at the right time - while the path model
measures hypothetical hit chance against a target that never dodges, and in
open-loop fixtures it over-credits directional guns (HeadOn 35% on DrussGT,
100% on constant-velocity) for exactly that reason. The models differ
materially but the current one is not shown to be unfaithful FOR ITS PURPOSE.
Because the metric drives gun SELECTION, this is now being A/B'd against real
hit rate versus the live DrussGT boss, which is the only ground truth we have.
Verified: 20 fixtures 35636/104000 (34.3%); 33 guard checks; 12/12 acceptance;
tsetlin tests green; live gauntlet 5/5.
The classic captures were OPEN-LOOP: replayed DrussGT never dodged OUR
bullets. These come from real TR battles through the working Java bridge, so
the recording contains genuine reactions to ModularBot's live fire. The
open-loop caveat is gone (perfect-information remains).
PRIMARY RESULT - the boss beats us badly. DrussGT 1447 - ModularBot 300 over
15 rounds, ModularBot winning only round 5 (DrussGT died at tick 1893). Rounds
are long, not truncated: mean 1335 ticks, ModularBot got off 1134 shots.
ModularBot 1134 shots / 60 hits = 5.3% real hit rate
DrussGT 1400 shots / 169 hits = 12.1% real hit rate
So DrussGT's gun is ~2.3x more accurate than our entire rack, on top of far
better movement. That is the number to move.
Also captured: shield-on variant (DrussGT 939-287, 9/10 - ModularBot takes
round 1 to the known shield warm-up), and vs SpinBot 1175-0, Crazy 1080-1,
Corners 1659-0. 20,026 + 12,629 + 10,824 + 11,507 + 2,575 ticks.
Movement statistics match the classic set within ~0.04 on the perpendicular
and radial fractions, so this is the same wave surfer in TR physics:
TR vs modularbot: perp 0.967, radial 0.001, 52.8% at full speed,
reversing 46.3%, median range 464 px.
CLOSED LOOP PROVEN, not asserted. ModularBot's fire is a heat-limited near
metronome (median interval 14 ticks), which gives a usable exogenous clock:
- event-locked |delta heading| oscillates 0.96 -> 2.69 deg about a 1.47 deg
mean with the fire period, almost every lag outside the 95% band of a
400-iteration phase-shuffled null;
- cross-correlation of |delta heading| against the fire impulse peaks at
r = +0.111, lag 12 ticks, permutation p = 0.005 (null peak mean +0.016);
- OWN-FIRE CONTROL is flat, so the oscillation is enemy-driven rather than
an internal cadence;
- range response is weak (~3 px over 30 ticks, near noise) and is therefore
NOT claimed, and per-bullet dodging is not claimed either because the
bullet detector (shield) is off.
CAVEATS: still perfect-information (observer gives true positions every tick,
unlike the live bot's stale between-scan WorldState) so these remain optimistic
vs live play; and they are open-loop AT REPLAY TIME - 'closed_loop' describes
the capture, not a later replay. TR conversion residual is ~1.5 deg mean
because the TR server moves along the pre-turn heading, vs 0.000 deg for the
classic captures.
Adds analyze_closed_loop.py (PSTH event-locking, phase-shuffle permutation
null, cross-correlation, own-fire control) and per-round result sidecars.
The entire GuessFactor family scored 0% on clean circular and wall-bounce
trajectories. Two hypotheses were on the table and BOTH were wrong:
- MEA range too narrow / edge clamping: REFUTED. Measured 0 clamped shots
out of 837/849/957, required offsets peak at ~33 deg against MEA
28.1-46.7 deg, and the 8 in arcsin(8/bulletSpeed) is correct (it is the max
robot SPEED, not the hit radius). Changing it to BotRadius=18 would have
coarsened resolution for nothing.
- Peak selection: REFUTED. A sweep of every constant GF value showed the
ORACLE-BEST constant offset on the original gun was only 6% circular,
4% wall-bounce, 7.5% random-walk. No peak choice could have done better.
The learning path was fine too: ~850-960 observations per fixture, 0
starved waves, well-populated histograms.
REAL CAUSE: the GF family aimed at the FIRE-TIME distance. The virtual-bullet
metric resolves a bullet at the AIM-POINT distance and scores that single
point against the enemy's position on that tick, so with any radial target
motion the bullet stops at the wrong radius and misses even with a perfect
angle. Angle-only prediction is structurally unscoreable under this metric.
FIX: give the GF family a self-consistent constant-velocity forecast as its
base reference (new common_libs/guns/lead_forecast.nim, which iterates the
flight time to the same fixed point circular.nim uses), so the histogram
learns the RESIDUAL against that forecast and the aim point lands at the
right radius. Applied to guess_factor, decay_gf and knn_gun.
Same defect fixed in Linear: it did a one-shot dist/bulletSpeed extrapolation
and never iterated its flight time.
The oracle sweep proves the structural fix, independently of tuning: the best
achievable constant GF moved 6% -> 20% (circular), 4% -> 57% (wall-bounce),
7.5% -> 49% (random-walk).
MEASURED, all 15 fixtures: total 39.0% -> 44.4% (30399 -> 34654 hits).
circular GF 6 -> 23, DecayGF 6 -> 21
wall-bounce GF 0 -> 60.2, DecayGF 0 -> 60.2
constant-vel GF 26 -> 100, DecayGF 26 -> 100, KNN 26 -> 100, Linear 87 -> 100
random-walk GF 0 -> 53, DecayGF 0 -> 52, Linear 24 -> 53
StraightLine GF 8 -> 77, DecayGF 8 -> 77
Non-regression: 33 guard checks pass, the range's 12/12 offline==online
acceptance still PASSES, tsetlin tests green, live gauntlet 5/5.
HONEST TRADE-OFF, recorded rather than hidden: on the 5 real DrussGT
wave-surfing captures the GF family REGRESSES - GuessFactor 108 -> 55,
DecayGF 108 -> 76, KNN 101 -> 74 hits per 2000. The linear base is a poor
model for a surfer, so the residual histogram is noisier than the old
total-lead histogram. Linear itself improved there (95 -> 105). The synthetic
range and the live gauntlet both improved, and the structural bug is provably
fixed, so this was judged worth the cost - but recovering the DrussGT
regression is the next job, not something to wave away.
The unmodified DrussGT.jar connects, wave-surfs, fires and beats every
adversary we have. 5 rounds each, all rounds won:
SpinBot 542-16, Corners 823-4, Crazy 604-0, RamFire 900-0,
ModularBot 506-72.
It is genuinely surfing, not drifting or stalling. Movement statistics
against the classic captures, same metrics, same analyzer:
opp perp TR/classic reversing TR/cl median range TR/cl
SpinBot 0.967 / 0.962 0.466 / 0.464 462 / 402
Corners 0.894 / 0.920 0.462 / 0.419 520 / 504
Crazy 0.841 / 0.829 0.527 / 0.439 370 / 378
RamFire 0.762 / 0.651 0.533 / 0.417 326 / 283
ModularBot 0.956 / - 0.466 / - 453 / -
Every round starts moving within 3-11 ticks and runs at 75-82% full speed.
Implemented: the classic remaining-quantity motion model (delegated to the
TR Bot's own Nat-Pavasant model - identical constants: accel 1, decel -2,
max 8, body turn 10-0.75|v|, gun 20, radar 45 - so getDistanceRemaining and
getTurnRemaining are exactly self-consistent with what is emulated); event
synthesis with classic ordering; bullet identity via object identity; gun
heat; rounds; radar cadence; firing translation; and the ThreadManager
landmine is killed by installing a no-op IThreadManagerBase in
ContainerBase.instance (verified: without it 'RobotException: ThreadManager
cannot be null!' kills the bot thread; with it the write succeeds).
Also adds TrBattleCapture, an observer that dumps per-tick state in the SAME
JSONL fixture format as the classic capture, so legacy bots can be captured
from Tank Royale battles too.
HONEST DIVERGENCES (README section 5.9): the TR server moves along the
PRE-turn heading and then turns, while classic aligns displacement with the
POST-turn heading, so the analyzer's conversion error is ~1.5 deg rather than
0.000 deg; distanceRemaining decrements by target speed rather than actual
distance; collision clamping differs; BulletMissedEvent can fire less often
because age-expiry has no TR event; bulletId is a local temp id;
StatusEvent/onPaint/SkippedTurnEvent are never delivered. Physics fidelity
diverges by construction - expect to retune.
EnergyDomeWorker (the bullet shield) is off by default: its precise
bullet-detection warm-up makes round 1 up to 79% stationary vs 35% in
classic. Rounds 2+ match classic closely with it on, but the pure surfer is
the consistent path. DRUSSGT_SHIELD=1 re-enables it.
Jars remain out of git.
Task 1 - the acceptance proof was unrunnable because RecordWorldState was a
compile-time const set to false. It is now a RUNTIME switch
(let RecordWorldState* = existsEnv("TR_RECORD_WORLDSTATE")), default OFF, so
ordinary runs write no fixture, and acceptance_offline_vs_online.nim enables
it for the battle it spawns and clears it afterwards. Restored and run twice:
12/12 deterministic guns match exactly (128-tick and 546-tick battles), with
Tsetlin reported separately as stochastic. Both nimble build variants clean.
Task 2 - does a compact encoding turn the TM's 99.35% into a READABLE rule?
Measured across window sizes (fixed seed, no tuning):
frames TEST acc eff.lits/clause firing clauses counterfactual low/high/mean
10 99.35% 152.8 37 100/24/62.4%
3 95.94% 54.9 35 96/20/58.6%
2 99.48% 39.6 38 95/25/60.7%
1 98.30% 19.2 45 100/24/62.3%
So 2 frames is strictly better than 10 on BOTH axes: +0.13 accuracy for 4x
smaller clauses. The 3-frame dip is non-monotonic and left unexplained rather
than smoothed over.
A readable rule WAS partially recovered. Five clauses carry the exact Gray
form !g10 ^ !g9 ^ !g8; g10 is inert in this data, so the effective rule is the
2-literal proposition !g9 ^ !g8, i.e. energy < 25.6. That is a genuine
threshold in readable propositional form - but at 25.6, NOT the labelled 30,
because 256 is a power-of-two Gray boundary expressible in two literals while
300 needs a longer conjunction. The TM found the nearest SIMPLE threshold.
The honest caveat: that threshold is not the ensemble's decision mechanism.
The counterfactual follow rate (high 24%, mean 62.3%) is statistically
identical at 1, 2 and 10 frames, so compactness did not make the model read
energy - its vote is carried by co-occurring bearing/velocity/heading/wall
literals. Also identified: clauses containing all 11 Gray energy bits are
satisfied at exactly one raw value (50, the dataset floor), so they are
'energy has hit the floor' detectors, not thresholds.
Methodological fix worth keeping: the earlier single-frame counterfactual
wrote energy into all 10 frame slots including the zeroed ones, reviving dead
clauses and producing a spurious 2% high-follow rate. setEnergyFrames now
rewrites only the exposed frames; the corrected figure is 24%.
The gun has never contributed anything: Tsetlin.vHits was byte-for-byte
equal to Linear.vHits in every measured round of every run, because its
learned correction was always exactly 0.
Six diagnosed defects fixed, plus one that was required to make the first
one work:
1. Type I now conditions on the clause output. It previously rewarded
included true literals unconditionally, omitting Granmo's (c=0, lk=1)
-> toward Exclude counter-force, so true literals ratcheted toward
Include forever. This was the root cause of the saturation.
2. Type II was unreachable dead code: its guard required cOut==1 AND
lits[lit]==0 AND st>0 (included), but cOut==1 guarantees every included
literal is 1. Its direction was wrong too - it should increment EXCLUDED
false literals when the clause fires.
3. Resource allocation restored: Granmo's (T - clip(v,-T,T))/(2T) target
replaces |error|/(2*RESID_MAX); TM_T was only an output normaliser.
4. Label baseline fixed - the factor-2 shrink. predX = linearX + cx, so the
label was delta - cx while the learner's output IS cx, giving
error = delta - 2cx and a fixed point of cx = delta/2: HALF the needed
correction even with perfect feedback. TmTrace now stores linearX/linearY
and training uses delta.
5. Hits no longer zero their label (a hit means |miss| < 18px, not 0).
6. The enemy-energy feature was duplicated - tmEncodeFrame passed
state.selfEnergy with a stale comment claiming enemyEnergy was absent,
while WorldState.enemyEnergy exists. Enemy-energy rules were literally
unrepresentable.
7. REQUIRED EXTRA: tmEvalClause now implements Granmo Eq. 6 - an all-Exclude
clause outputs 1 during learning and 0 during classification. Without it,
fix#1 deadlocks every clause at empty.
MEASURED EFFECT (energy-threshold-turner fixture, seed 1):
mean included literals per active clause 714.0 -> 13.8
active clauses 100/100 -> 53/100
nonzero corrections 8/764 -> 708/764
Tsetlin virtual hits (Linear = 27/400) 27/400 -> 69/400
Divergence achieved: offline on 7/8 fixtures, and in a live gauntlet
(RandomMover: Tsetlin 199/1200 vs Linear 288/1200, vDropped=vStarved=0).
Tsetlin now LEARNS but is not yet competitive with Linear - the regression
head is untuned, flagged as follow-up rather than claimed as a win.
Also ignores compiled test harnesses that have no file extension, which the
existing '**/tests/test_*' rule misses.
The question was whether we can fight a genuine legacy leader bot in Tank
Royale. Answer: the API side is now PROVEN, not estimated.
KEY FINDING: the classic robocode.* API is a thin delegation layer over a
public seam. javap -c shows AdvancedRobot forwarding every call to
_RobotBase.peer (IBasicRobotPeer/IAdvancedRobotPeer), and _RobotBase.setPeer
is public final. So we do NOT need to reimplement the API: we reuse the
genuine robocode.jar and implement only the 75-method peer interface.
Consequences:
- DrussGT's 22 sources compile against the real API with ZERO unresolved
symbols. (The literal 22-file javac fails only on two PRE-EXISTING
duplicate classes - GFRange and Indice are declared both inline in
DrussGunDC.java and as standalone files - and the 6 classes that ship
without source. The jar supplies all of them, so a shim never cares.)
- Runtime smoke test PASSES: the unmodified DrussGT.jar is loaded through a
child URLClassLoader and driven for 200 synthetic ticks, emitting movement
intents every tick and 169 fire requests, with its thread surviving.
This is the cheap path to the 'final boss', and it also unlocks the whole
roborumble archive rather than one bot.
Traps found by measurement:
- ScannedRobotEvent's constructor order is (name, energy, bearing, distance,
heading, velocity) - NOT heading-before-bearing. The wrong order silently
yields distance=0, an immediate KD-tree insert and an NPE in
EnemyMoves.predict; it hung the first smoke run.
- RobocodeFileOutputStream has a hard engine dependency (resolves
IThreadManagerBase via ContainerBase) and throws 'ThreadManager cannot be
null!' outside the engine, killing the bot thread. Reached from DrussGT's
own contain() error logging, so it must be stubbed.
- robocode.RobotDeathEvent is required and was NOT in the predicted API list.
- Bullet.equals() is genuinely called for bullet identity, not just getters.
REMAINING WORK (README section 5): coordinate rotation DONE, execute()->tick
bridge DONE and proven, ThreadManager fix scoped. The main open item is the
classic motion model (setAhead distance semantics vs TR speed), estimated
1-3 days, plus event synthesis/ordering, gun heat, bullet identity, round
and radar cadence, and firing translation. NO API UNKNOWNS REMAIN.
Physics fidelity will still diverge from classic - expect to retune.
Jars stay out of git (blocked by tools/robocode_shim/.gitignore); the
genuine robocode.jar and DrussGT.jar are referenced from /tmp via
ROBOCODE_JAR / DRUSSGT_JAR.
Measured hazard counts for a Java->Nim port, with the correction that only
22 of 28 top-level classes ship source (6 do not: 5 in the gun package plus
dMove/Scan), so a full port would need a decompiler while a shim would not
care at all.
Real traps: 193 float / 35 casts / 147 literals / 60 float[] in the movement
closure (the danger histogram is float[171] -- porting 32-bit Java floats to
Nim's default float64 diverges silently); ~68 non-final statics; and 4 Java
single-& sites with side effects, which break under Nim's short-circuiting
'and'. Non-issues, correcting earlier assumptions: 0 sites of %-on-negative
(angle normalisation is floor-based) and FastTrig has no lookup tables, it is
7 coefficient-exact polynomials.
Movement scoping: 5007 LOC across 14 files, ~4.8-6.1k Nim LOC, 4-8 focused
agent-days to first-compiles. It can be ported WITHOUT the gun (data flows
movement->gun only), but the harness never routes HitByBullet into movement
modules, so the danger bins would never train -- that plumbing is the real
blocker, not the translation.
There are no genuinely competitive adversaries for Tank Royale, and the
in-repo ones were broken until recently. Classic Robocode 1.9.5.5 is
obtainable (SourceForge, 20.4 MB) and its programmatic control API
(robocode.control.RobocodeEngine + BattleAdaptor.onTurnEnded) can run real
battles headless and expose per-turn robot state. So a legacy leader bot's
MOVEMENT can be captured and used as a gun-testing fixture with no port.
Captured unmodified DrussGT 3.1.4159 vs spinbot/ramfire/crazy/corners and a
mirror match: 28,797 ticks, plus two trivial-bot contrasts. Conversion to
the Tank Royale convention is validated to 0.000-0.001 deg by recomputing
the direction implied by (heading, speed) and comparing it against the
recorded per-tick displacement -- i.e. the data is proven to be genuine
recorded motion rather than a mangled export. (A first attempt treated the
snapshot API's headings as degrees; they are radians, ~95 deg off.)
The statistics confirm it is really a wave surfer: perpendicular to the
opponent 65-96% of ticks, radial ~0.001, 41-72% of ticks at full speed,
reversing on 42-46% of ticks, holding range at a 283-526 px median. The
straight-line contrast is radial-dominant (0.75) with ZERO reversals.
Discovery: DrussGT detects predictable guns and switches to a bullet-shield
stand-still mode, so captures against sample.Walls/TrackFire had to be
rejected as non-movement.
CAVEATS, recorded in DRUSSGT_FIXTURES.md: these are open-loop (replayed
DrussGT never dodges OUR bullets) and perfect-information (the observer
gives true positions every tick, unlike our stale live WorldState). Both
make our guns look better than in live play, so use them for RELATIVE gun
ranking, not absolute hit rates.
Jars stay out of git; capture tooling is reproducible via capture.sh.
Gun evaluation previously required a full end-to-end battle (Java server +
battle runner + websocket IPC to 2 bot processes, 50 rounds, ~3.4 min) and
yielded only ~300-900 REAL shots across 13 guns -- far too few to rank
guns, which is why tuning needed many repetitions.
VirtualTracker is already a pure function of (WorldState stream, gun list);
the only reason it needed Java was where WorldState came from. So the range
replays a seq[WorldState] through the SAME tracker: offline and online
scores are the same metric by construction, not an approximation.
ACCEPTANCE TEST (the point of the whole thing): record one live round, replay
it offline, compare per-gun virtual hit rates. 12/12 deterministic guns match
EXACTLY, reproduced twice. Tsetlin is compared separately because tmLearnOne
calls rand(). Getting to 12/12 exposed two real ordering quirks in the live
loop: run() calls go() before the aim/fire block, so tickBullets resolves
against the NEXT tick's scan while the prediction used the previous one; and
if the target dies during that go() the final tick's spawn+resolution is
skipped entirely. The recorder emits an end marker for the second case.
The 5th (selected-gun) predict call was verified to be a no-op.
Measured cost: 8 fixtures (1770 ticks, ~92k virtual bullets, 13 guns) replay
in 2.9 s, ~32k virtual bullets/s -- roughly 70x faster and 100x more samples
than a live gauntlet.
Also adds a per-tick WorldState recorder behind const RecordWorldState
(default off, mirrors the ShotLog idiom) which records the state the bot
ACTUALLY builds, staleness included, rather than true positions -- recording
the latter would hand the guns perfect information and produce flattering
scores.
9 new guard checks (33 total, all passing), including fixture round-trip,
replay determinism, stationary->HeadOn 100%, constant-velocity->Linear>HeadOn,
and the energy-threshold turner crossing at t=41.
Measured, not assumed. With the gate temporarily opened to 20 deg, every
real shot was logged (tick, angle error at fire time, distance, power,
hit) across 3 gauntlets: 2611 shots, 57.3% aggregate. Findings:
- The geometric cone atan(BotRadius/d) is directionally confirmed but a
WEAK lever: even at 0.0-0.1 deg error the hit rate at 400-600px is only
~53-57%, because PREDICTION error dominates alignment error.
- Real effect of tightening the gate: 57.9% -> 68.0% aggregate hit rate
(fixed 0.1 deg), not the 76.9% previously reported -- that was a
high-variance draw (per-rep 62.8/66.4/77.2%).
- The shipped range-aware gate (SafetyFactor 0.6) does NOT beat the fixed
2.0 deg gate on hit rate (55.8% vs 57.9%, ~1.5 sigma, inside noise). It
fires 22-28% more shots and therefore lands more total hits (~509 vs
~434 per rep). No per-adversary score delta exceeded the 300-point
run-to-run noise band, so no config is demonstrably better on score.
Shipped anyway because it is strictly more expressive (a fixed threshold is
the special case), tunable from one const, and physically motivated, but
the honest verdict is recorded in-code: the gate is not the bottleneck.
AimThresholdDeg is removed; shouldFire now takes distPx. Degenerate or NaN
distance falls back to the ceiling rather than dividing by zero.
Also adds a per-shot logger to ModularBot behind 'const ShotLog' so the
measurement above is reproducible, and 10 new guard checks (24 total, all
passing) covering monotonicity, clamping, formula, perfect alignment,
gross misalignment and degenerate distance.
Cross-checked against the server source: the gun fires BEFORE the turn is
applied, so the logged angle error is the true departure error, and
fireAssist auto-aim is off (unset by the Nim API and forced false by
setAdjustRadarForGunTurn).
tm-learning-tracks.md covers three things, all marked [FACT]/[INFERENCE]/
[UNKNOWN]:
- Section A: the Tsetlin gun's label is measured against the wrong baseline.
predX = linearX + cx, so rx = actual - predX = delta - cx, and inside
tmLearnOne error = residual - predicted = (delta - cx) - cx = delta - 2cx.
The fixed point is cx = delta/2 -- HALF the correction needed, even with
perfect Granmo feedback. Fix: store linearX/linearY in TmTrace and train
on delta. Also: hits zero the label instead of carrying their true
residual, and the per-clause step is magnitude-blind.
- Section B: what a TM is actually good at (AND-clauses over binary
literals, readable output) and why this repo suits it -- the gun already
builds an 83-bit x 10-frame Gray-coded window (870 bits). Includes a
falsifiable known-rule benchmark proposal.
- Section C: delayed-reward learning belongs to the MOVEMENT layer, not the
gun. The gun's outcome is delayed but exactly pairable via
(fireTick, powerBin), so its effective lambda is 1 and discounting would
only destroy information.
Also records that docs/papers/tm-deb-paper.pdf was deleted by the user as
AI-generated and unverifiable, while the Granmo-based feedback diff in
tm-deb-assessment.md stands on its own.
Verdict: the document (docs/papers/tm-deb-paper.pdf, 'Generated by
Gemini Notebook') solves temporal credit assignment under delayed
reward, which is not our problem. Our gun's failure is clause
saturation (~131 of 1740 literals included per clause -> conjunction
fires with probability ~2^-131 -> correction identically 0), and
TM-DEB only scales update FREQUENCY by gamma^dt, so it would leave
the fixed point untouched and additionally delete the long-range
feedback we need: gamma^90 = 4.4e-7 at the paper's own gamma=0.85,
and those are the shots whose lead matters most.
Credibility signals recorded in the doc: reference [2] misattributes
authors/venue/year; Algorithm Spec 2 steps 9-12 are literal '%'
placeholders so the automata update is simply absent; Table 1 is
titled 'Expected' and reports never-measured accuracies; Eq. 12 is
not a faithful copy of Granmo's Lemma 2.
The audit also produced the actionable result: a line-by-line diff of
Granmo Table 2/3 feedback against tmLearnOne, identifying why the
automata saturate - Type I never conditions on the clause output so it
omits Granmo's c=0, lk=1 -> -1 w.p. 1/s counter-force (true literals
ratchet toward Include), Type II's guard cOut==1 AND lits[lit]==0 is
unsatisfiable and therefore dead code, and the (T - clip(v,-T,T))/(2T)
resource allocation is missing entirely.
Four guns cached a whole prediction per tick while predict() is called once
per power bin, so every bin after the first (and the real fired shot, which
shares lastState) reused the power-1.0 lead. Fixed by caching only the
speed-INDEPENDENT derived state and recomputing the lead per requested speed:
- stop_shot: also fixes prevSpeed being written before it was read, which
made abs(speed) < abs(prev) permanently false and the entire
stop-prediction branch unreachable (it was just Linear).
- displacement: the cache key included bulletSpeed, so the guard missed on
all four bins and the 15-tick window advanced ~4x/tick, making the
inferred velocity ~4x too small.
- averaged_lead: tick cache removed outright. pattern_matcher: split into
speed-independent match+path and per-call lead.
FeedbackEvent gains fireTick/powerBin (additive; only virtual_bullets
constructs one) so guns can pair feedback to the exact shot instead of
guessing by coordinates. tsetlin uses it: traces are now keyed exactly by
(fireTick, powerBin) with a 1024-slot ring, and the 10-frame window shifts
at most once per tick (it was shifting ~4-5x/tick, so isWarmedUp tripped
after ~2 ticks).
KNOWN INCOMPLETE: tsetlin still does not diverge from Linear in battle. The
two named bugs are fixed (a 600-tick sim shows trainedShots=2141,
traceMisses=0, and a fixed-input probe converges to a 9.6px correction), but
the TM's clause feedback itself is broken: ~131 of 1740 literals end up
included per clause, so its conjunction never fires. Sweeping TM_S,
TM_N_CLAUSES and a two-branch Type-I update did not change the correction
from 0. Needs a real TM fix or removal, not another bug fix.
First-ever guard tests for the gun selector: common_libs/tests/
test_gun_harness.nim (14 checks, headless, no Java). There were none before,
which is how six broken guns survived a full analysis cycle. Against the
previous HEAD, 5 of these checks FAIL - that is the regression guard.
Wave queues (guess_factor, decay_gf, knn_gun): predict() stored ONE wave
per tick while onResult() popped one per resolved bullet (~4/tick), so the
queue drained to empty within a few dozen ticks, ~3 of every 4 resolutions
returned without learning, and the survivor paired with a same-tick wave
(bearingDelta ~= 0) pinning the histogram at centre. PROOF: GF.vHits ==
HeadOn.vHits and DecayGF.vHits == HeadOn.vHits byte-for-byte in every one
of 50 rounds — the guns had degenerated to HeadOn.
Now each gun keeps a per-bin FIFO with an O(1) head cursor. At most one
push per (tick, bin) so the fire site's 5th predict() call is a no-op, and
onResult pops the oldest wave of its OWN bin via e.bulletPower. Aiming
math untouched (it was already correct: 0 deg = East, CCW+).
maxBullets 2048 -> 8192: the rack spawns 52 bullets/tick so the ring wrapped
every ~39 ticks while a long power-3 shot needs ~90, silently discarding
unresolved bullets and biasing every measured hit rate by range. Added a
droppedBullets counter so a future overflow is measurable, and wavePushes/
waveStarved counters on the three guns. After the fix: vDropped = 0 and
vStarved = 0 across all 48 recorded rounds.
fitnessFor is now exported, deterministic (enemies iterated in ascending id
order) and shared by the selector and the stats dump, replacing a hand-rolled
merge in ModularBot that never advanced its window head.
Round lines gain additive keys: vDropped, vStarved.
nimble bin output lands at the garage root, so the existing
'*_garage/out/' ignore rule never covered it and every build dirtied
the tree with a 1.1 MB binary. File stays on disk; build regenerates it.
Attribution is proven, not guessed: the server assigns a per-round-unique
bulletId (GunEngine.nextBulletId) and stamps the same id on BulletFired,
BulletHitBot, BulletHitWall and BulletHitBullet. Keep a FIFO of fired gun
ids, stamp bulletId -> gunId on onBulletFired, resolve through that map.
Hits are deferred when onBulletHit precedes onBulletFired in the same
turn (client dispatches priority 70 > 60), which recovered 14
unattributed hits. 99.9% of shots and 99.8% of hits attributed.
Stats lines now carry per-gun realShots/realHits/realHitRate; the old
keys and round-level totals are unchanged.
bestPower: a gun with zero observations in every bin previously returned
the HIGHEST bin (power 3.0) because an empty bin satisfied the
'count == 0' clause on the first countdown iteration. Cold guns now
return the lowest bin as the docstring always claimed. Warm-gun path
untouched.
- bestGun: replace first-index-wins argmax with random pick among guns
within TieMargin (2%) of best rate. HeadOn at index 0 was silently
winning every tie, starving Tsetlin/Linear/etc.
- MinObsBeforeCompete 15 -> 50 (Pattern entered competition on noise)
- add MinHitRateFloor 0.10: if no gun clears it, fall back to HeadOn
instead of selecting the best of a bad field
- ModularBot: remove AntiSurfer gun (0% virtual hit rate everywhere),
14 -> 13 guns, renumber ids and selection counters
- 30-tick cooldown after ghost-stuck/timeout ram exit prevents re-entry loop
- enemy_tracker.update() skips dead bots to prevent same-tick scan resurrection
- TFIL graphics cleared when ramming is active movement
- [config] logs: white base with green-highlighted changes only
- [ram:enter] logs trigger reason and key values on false→true transition
- [death] and [target-invalid] logs retained for diagnostics
PhantomMeteor:
- Ram finisher: charge at enemy when <200px and their energy <10
- Ram opportunity: charge when <60px and we have >20 energy advantage
- Integrated gunheat tracker for 1-2 tick earlier wave detection
- Distance control: smooth linear ramp toward preferred engagement distance
- Phantom range expanded 150→250px to catch closer threats
WaveSurfer:
- Wall-aware dodge bin selection: penalize bins leading off-arena
- Dodge timing: predict future position 15 ticks ahead for safety
- Distance control: radial blend when outside deadband (350±50px)
- Wall escape: invert strafe if pushing further into wall, blend toward center
ModularBot:
- Wired KNN gun (purple/magenta)
- Shadows tracked for movement (safer GF prediction)
- Bullet lifecycle management (onBulletFired/onBulletHitBot/onBulletHitWall)
- Unified phantom_meteor movement (wave_surfer unplugged)
- Config logging on round start + gun switch
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Gun selector now gates competition: guns with <15 total observations across
all power bins sit out until at least one gun qualifies. Falls back to ungated
selection if no gun reaches threshold, preventing cold-start stalls.
Sliding window increased from 50 to 100 ticks to reduce switching noise.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New DrussGT-inspired modules:
- KNNGun: K-nearest-neighbor statistical targeting using GF density peaks
- GunheatTracker: dual-heat system (predicted + confirmed) for 1-2 tick lead
- ShadowTracker: computes GF regions safe from in-flight bullets (enemy wave dodge)
VirtualBodyTracker now integrates gunheat for earlier fire detection and shadows
for safe-zone multiplier (90% reduction in danger zones).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Renames PatternMover_garage → PatternMover, RandomMover_garage → RandomMover,
WaveSurfer_garage → WaveSurfer. Updates all .json, .sh, .nimble, and config.nims
files to match TR Booter naming convention (directory name = bot name).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implements VirtualBodyTracker (wave-based hit/miss scoring) instead of EMA damage accumulation. Movement switching now happens every tick, not every 3 rounds. Also refactors radar colors to dark teal (#004444/#0D4D4D) for faint visibility.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Adds WaveSurferModule as second movement option
- Tracks damage per round via onHitByBullet, EMA decay α=0.5
- Every round after round 2: switches to lower-damage movement
- Body color signals active mover (orange=phantom, blue=wave)
- vs SpinBot 10r: ModularBot 1412 vs SpinBot 650
- vs WaveSurfer 10r: ModularBot 1858 vs WaveSurfer 10
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- GF gun: triangular head-on prior replaces flat bins (was always aiming at GF=-1)
- GF gun: recompute MEA per bullet power in onResult (was using stale first-bin value)
- Scores improved: SpinBot 1805-48, Crazy 1557-46, WaveSurfer 1868-10, PatternMover 1846-4
- All 6 guns active in selection across gauntlet
- Cold-start bug: uniform bins[0..30]=0.1 made peakBin() always return
0 (first-wins tie), giving GF=-1 (max CW escape) before any learning.
Fixed with a triangular head-on bump at bin 15 (GF=0) as the prior.
- onResult now recomputes mea from FeedbackEvent.bulletPower instead of
the stale first-bin mea cached by predict; correct per-power-bin GF.
- Add DebugGF const (default false) with [gf-dbg] echoes in predict/onResult.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
SirStone