- 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>
Two bugs fixed:
1. Multi-tick gaps: turn rate assumed 1 tick between observations, but scans can be 5+ ticks apart. Now divides by actual tickDelta.
2. Per-power-bin state corruption: predict() called 4x per tick (per power bin). After first call, prevHeading was already updated, causing subsequent calls to compute 0° delta. Now captures oldHeading/oldTick before updating.
Verified: OscillatorBot at 4°/tick captured correctly; normalization [-180°,180°] works; tickDelta=1 typical; first bin gets delta, subsequent bins see 0 (expected).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace binary hit/miss reward with exponential decay based on miss distance:
- reward = 2.0 * exp(-missDistance / 36.0) - 1.0
- At 0px: +1.0 (perfect hit)
- At 36px: -0.26 (near miss, small penalty)
- At 100px: -0.87 (big miss, large penalty)
Maintains virtual hit/miss counters for display (threshold: 36px).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Swap sin/cos in virtual bullet position calc to match Tank Royale convention
- Change exploration from flipping each bit with 50% chance to flipping exactly 1 random bit
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Added myEnergy, enemyEnergy, and canFire to ScanData. Energy fields quantize 0-150 → 8 bits each; gun ready is 1 bit (1 = can fire, 0 = can't). Updated total encoding from 100 to 117 bits.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
8 new fields in ScanData (myWallN/S/E/W, enemyWallN/S/E/W), 7 bits each via Gray code,
appended after existing 44 bits. TOTAL_BITS 44 → 100.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Output format: <tick> <44 bits> <hamming> <similarity> <overlap>
- Similarity: bits that are the same (TOTAL_BITS - hamming)
- Overlap: bits that are 1 in both vectors (popcount of bitwise AND)
- First tick: all metrics printed as 0
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Replace population/thermometer/signed-thermometer encoding with simple 10-bit binary per value
- 5 values × 10 bits = 50 bits total (was 112)
- Each value normalized to 0-1023, then bit-extracted MSB-first
- Output is now just tick and raw binary string (no field labels, no debug log)
- Remove file logging entirely; stdout only
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Pre-populate all LeadGrid cells with arcsin(vPerp/bulletSpeed) so the
grid starts warm instead of cold, avoiding the early-round 0-data
fallback to -999 (no-lead) firing.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reverted fire gate from 1° back to 2°, removed gridHasData cold-start
guard so bot fires on straight aim during cold-start. Added
/tmp/snnbot_bias_debug.log BIAS entries in EVALUATE phase logging
gridOffset, correctOffset, bias, vPerp, dist per cycle.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- P3 at err<0.5°, P1 at err<2.0° (was fixed P2); no fire at err>=2.0°
- Scale grid lead offset by bullet speed ratio (sin(θ) ∝ 1/v_bullet)
- ENERGY_GUARD 5→15 to survive Walls' sustained fire
- Suppress fire when |vPerp delta| > 2.0 (enemy changing direction)
- Store lastBulletSpeed = 20-3*chosenPower for correct EVALUATE travel time
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Reverts reservoir.nim to the neighbor-blending forward() + single-cell
learn() version (pre-bilinear-interpolation). Adds /tmp/snnbot_round_stats.log
and /tmp/snnbot_aim_debug.log for diagnostics independent of test framework stdout.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Each learn() call now updates a 3×3 neighborhood (center w=1.0, cardinal w=0.3,
diagonal w=0.1). count field changed to float64 to support fractional weights.
Fills grid ~5× faster and eliminates one-sided interpolation at bin boundaries.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Ring buffer had amnesia — cycled out all data every 128 ticks,
preventing convergence. Grid accumulator permanently stores average
lead offsets indexed by (v_perp, distance). 136 cells, 1.5 KB.
Knowledge accumulates across rounds → convergence guaranteed for
stationary velocity patterns.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
relVelDir ±5° shared 80% of bits despite needing opposite lead.
Now encodes v_perp = speed×sin(relVelDir) directly with signed
thermometer coding — positive and negative crossing velocities
have zero bit overlap. Also adds v_parallel for approach/recede.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
SirStone