no message

Ajoute Unreal/Plugins/*/Binaries/ et les patterns *.dll/*.pdb/*.exp/*.lib/*.exe/*.dylib au .gitignore, et retire de l'index les 22 binaires précédemment trackés. Évite de polluer le dépôt avec les artefacts de build régénérables.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

.claude/settings.local.json

@@ -0,0 +1,38 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(find /c/ASTERION/GIT/PS_Ballistics/Unreal -name *.bat -o -name Generate*.sh -o -name *Generate*)",
+      "Bash(\"C:\\\\Program Files\\\\Epic Games\\\\UE_5.5\\\\Engine\\\\Build\\\\BatchFiles\\\\Build.bat\" PS_BallisticsEditor Win64 Development -Project=\"C:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Unreal\\\\PS_Ballistics.uproject\" -WaitMutex -FromMsBuild)",
+      "Bash(\"C:\\\\Program Files\\\\Epic Games\\\\UE_5.5\\\\Engine\\\\Build\\\\BatchFiles\\\\Build.bat\" PS_BallisticsEditor Win64 Development -Project=\"C:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Unreal\\\\PS_Ballistics.uproject\" -WaitMutex -FromMsBuild -NoLiveCoding)",
+      "Bash(\"C:\\\\Program Files\\\\Epic Games\\\\UE_5.5\\\\Engine\\\\Build\\\\BatchFiles\\\\Build.bat\" PS_BallisticsEditor Win64 Development -Project=\"C:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Unreal\\\\PS_Ballistics.uproject\" -NoLiveCoding)",
+      "Bash(grep -l \"Shoot\\\\|ClientAim\\\\|ShootRep\" \"E:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Unreal\\\\Plugins\\\\PS_Ballistics\\\\Source\\\\EasyBallistics\\\\Private\"/*.cpp)",
+      "Bash(xargs grep:*)",
+      "Bash(ls Source/EasyBallistics/Private/*.cpp Source/EasyBallistics/Public/*.h)",
+      "Bash(powershell.exe -Command \"& ''''C:\\\\Program Files\\\\Epic Games\\\\UE_5.5\\\\Engine\\\\Build\\\\BatchFiles\\\\RunUAT.bat'''' BuildEditor -project=''''E:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Unreal\\\\PS_Ballistics.uproject'''' -notools -noP4 2>&1\")",
+      "Bash(python \"E:\\\\ASTERION\\\\GIT\\\\PS_Ballistics\\\\Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_150326.csv\")",
+      "Bash(python \"Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_150326.csv\")",
+      "Bash(python \"Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_153607.csv\")",
+      "Bash(python \"Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_160323.csv\")",
+      "Bash(python \"Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_164341.csv\")",
+      "Bash(python \"Tools\\\\analyze_antirecoil.py\" \"E:\\\\ASTERION\\\\SVN\\\\DEV\\\\PROSERVE_UE_5_5\\\\Saved\\\\Logs\\\\AntiRecoil_20260316_170543.csv\")",
+      "Bash(find C:ASTERIONSVNDEVPROSERVE_UE_5_5Plugins -type f \\\\\\(-name *.cpp -o -name *.h \\\\\\))",
+      "Bash(git add:*)",
+      "Bash(git commit:*)",
+      "Bash(find E:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved -name AntiRecoil* -type f)",
+      "Bash(python analyze_antirecoil.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_140946.csv\" --grid)",
+      "Bash(python analyze_antirecoil.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_140946.csv\" \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_141329.csv\")",
+      "Bash(python analyze_antirecoil.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_140946.csv\" \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_141329.csv\" --grid)",
+      "Bash(python analyze_antirecoil.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_140946.csv\" \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_141329.csv\" --grid --strategy worst_case)",
+      "Bash(python analyze_antirecoil.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_162726.csv\" \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_162234.csv\" --grid --strategy worst_case)",
+      "Bash(python analyze_shots.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_162726.csv\")",
+      "Bash(python analyze_shots.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_163533.csv\")",
+      "Bash(python analyze_shots.py \"C:/ASTERION/SVN/DEV/PROSERVE_UE_5_5/Saved/Logs/AntiRecoil_20260318_181404.csv\")",
+      "Bash(python -c \":*)",
+      "Bash(git -C C:/ASTERION/GIT/PS_Ballistics status --short)",
+      "Bash(git -C C:/ASTERION/GIT/PS_Ballistics ls-files)",
+      "Bash(grep -iE \"\\\\.\\(dll|pdb|exp|lib|exe|dylib|so|a|o\\)$\")",
+      "Bash(git -C C:/ASTERION/GIT/PS_Ballistics add .gitignore)",
+      "Bash(git -C C:/ASTERION/GIT/PS_Ballistics commit -m ' *)"
+    ]
+  }
+}

.gitignore

@@ -3,4 +3,19 @@ Unreal/.vs/
 Unreal/Binaries/
 Unreal/Intermediate/
 Unreal/Plugins/EasyBallistics/Intermediate/
+Unreal/Plugins/EasyBallistics/Binaries/
 Unreal/Saved/
+Unreal/Plugins/PS_Ballistics/Intermediate/
+Unreal/Plugins/PS_Ballistics/Binaries/
+
+# Binaires compilés
+*.dll
+*.exp
+*.pdb
+*.lib
+*.obj
+*.exe
+*.so
+*.dylib
+*.a
+*.o

Bind to PROSERVE.bat

@@ -0,0 +1 @@
+powershell -Command "New-Item -ItemType Junction -Path 'E:\ASTERION\SVN\DEV\PROSERVE_UE_5_5\Plugins\PS_Ballistics' -Target 'E:\ASTERION\GIT\PS_Ballistics\Unreal\Plugins\PS_Ballistics'"

Tools/analyze_antirecoil.py

@@ -0,0 +1,779 @@
+"""
+Anti-Recoil Parameter Optimizer
+================================
+Reads CSV files recorded by the EBBarrel CSV recording feature and finds
+optimal parameters for the Adaptive Extrapolation mode.
+
+Usage:
+    python analyze_antirecoil.py <csv_file> [csv_file2 ...] [options]
+
+Options:
+    --plot              Generate comparison plots (requires matplotlib)
+    --grid              Use grid search instead of differential evolution
+    --strategy <s>      Multi-file aggregation: mean (default), worst_case
+    --max-iter <n>      Max optimizer iterations (default: 200)
+
+The script:
+1. Loads per-frame data (real position/aim vs predicted position/aim)
+2. Simulates adaptive extrapolation offline (matching C++ exactly)
+3. Optimizes all 4 parameters: Sensitivity, DeadZone, MinSpeed, Damping
+4. Reports recommended parameters with per-file breakdown
+"""
+
+import csv
+import sys
+import math
+import os
+import argparse
+from dataclasses import dataclass
+from typing import List, Tuple, Optional
+
+
+@dataclass
+class Frame:
+    timestamp: float
+    real_pos: Tuple[float, float, float]
+    real_aim: Tuple[float, float, float]
+    pred_pos: Tuple[float, float, float]
+    pred_aim: Tuple[float, float, float]
+    safe_count: int
+    buffer_count: int
+    extrap_time: float
+    shot_fired: bool = False
+
+
+@dataclass
+class AdaptiveParams:
+    sensitivity: float = 3.0
+    dead_zone: float = 0.95
+    min_speed: float = 0.0
+    damping: float = 5.0
+    buffer_time_ms: float = 200.0
+    discard_time_ms: float = 30.0
+
+
+@dataclass
+class ScoreResult:
+    pos_mean: float
+    pos_p95: float
+    aim_mean: float
+    aim_p95: float
+    jitter: float
+    overshoot: float
+    score: float
+
+
+def load_csv(path: str) -> List[Frame]:
+    frames = []
+    with open(path, 'r') as f:
+        reader = csv.DictReader(f)
+        has_shot_col = False
+        for row in reader:
+            if not has_shot_col and 'ShotFired' in row:
+                has_shot_col = True
+            frames.append(Frame(
+                timestamp=float(row['Timestamp']),
+                real_pos=(float(row['RealPosX']), float(row['RealPosY']), float(row['RealPosZ'])),
+                real_aim=(float(row['RealAimX']), float(row['RealAimY']), float(row['RealAimZ'])),
+                pred_pos=(float(row['PredPosX']), float(row['PredPosY']), float(row['PredPosZ'])),
+                pred_aim=(float(row['PredAimX']), float(row['PredAimY']), float(row['PredAimZ'])),
+                safe_count=int(row['SafeCount']),
+                buffer_count=int(row['BufferCount']),
+                extrap_time=float(row['ExtrapolationTime']),
+                shot_fired=int(row.get('ShotFired', 0)) == 1,
+            ))
+    return frames
+
+
+# --- Vector math helpers ---
+
+def vec_dist(a, b):
+    return math.sqrt(sum((ai - bi) ** 2 for ai, bi in zip(a, b)))
+
+
+def vec_sub(a, b):
+    return tuple(ai - bi for ai, bi in zip(a, b))
+
+
+def vec_add(a, b):
+    return tuple(ai + bi for ai, bi in zip(a, b))
+
+
+def vec_scale(a, s):
+    return tuple(ai * s for ai in a)
+
+
+def vec_len(a):
+    return math.sqrt(sum(ai * ai for ai in a))
+
+
+def vec_normalize(a):
+    l = vec_len(a)
+    if l < 1e-10:
+        return (0, 0, 0)
+    return tuple(ai / l for ai in a)
+
+
+def angle_between(a, b):
+    """Angle in degrees between two direction vectors."""
+    dot = sum(ai * bi for ai, bi in zip(a, b))
+    dot = max(-1.0, min(1.0, dot))
+    return math.degrees(math.acos(dot))
+
+
+# --- Prediction error from recorded data ---
+
+def compute_prediction_error(frames: List[Frame]) -> dict:
+    """Compute error between predicted and actual (real) positions/aims."""
+    pos_errors = []
+    aim_errors = []
+
+    for f in frames:
+        pos_err = vec_dist(f.pred_pos, f.real_pos)
+        pos_errors.append(pos_err)
+
+        aim_a = vec_normalize(f.pred_aim)
+        aim_b = vec_normalize(f.real_aim)
+        if vec_len(aim_a) > 0.5 and vec_len(aim_b) > 0.5:
+            aim_err = angle_between(aim_a, aim_b)
+            aim_errors.append(aim_err)
+
+    if not pos_errors:
+        return {'pos_mean': 0, 'pos_p95': 0, 'pos_max': 0, 'aim_mean': 0, 'aim_p95': 0, 'aim_max': 0}
+
+    pos_errors.sort()
+    aim_errors.sort()
+
+    p95_idx_pos = int(len(pos_errors) * 0.95)
+    p95_idx_aim = int(len(aim_errors) * 0.95) if aim_errors else 0
+
+    return {
+        'pos_mean': sum(pos_errors) / len(pos_errors),
+        'pos_p95': pos_errors[min(p95_idx_pos, len(pos_errors) - 1)],
+        'pos_max': pos_errors[-1],
+        'aim_mean': sum(aim_errors) / len(aim_errors) if aim_errors else 0,
+        'aim_p95': aim_errors[min(p95_idx_aim, len(aim_errors) - 1)] if aim_errors else 0,
+        'aim_max': aim_errors[-1] if aim_errors else 0,
+    }
+
+
+# --- Shot contamination analysis ---
+
+def analyze_shot_contamination(frames: List[Frame], analysis_window_ms: float = 200.0):
+    """
+    Analyze how shots contaminate the tracking data.
+    For each shot, measure the velocity/acceleration spike and how long it takes
+    to return to baseline. This tells us the minimum discard_time needed.
+
+    Returns a dict with analysis results, or None if no shots found.
+    """
+    shot_indices = [i for i, f in enumerate(frames) if f.shot_fired]
+    if not shot_indices:
+        return None
+
+    analysis_window_s = analysis_window_ms / 1000.0
+
+    # Compute per-frame speeds
+    speeds = [0.0]
+    for i in range(1, len(frames)):
+        dt = frames[i].timestamp - frames[i - 1].timestamp
+        if dt > 1e-6:
+            d = vec_dist(frames[i].real_pos, frames[i - 1].real_pos)
+            speeds.append(d / dt)
+        else:
+            speeds.append(speeds[-1] if speeds else 0.0)
+
+    # For each shot, measure the speed profile before and after
+    contamination_durations = []
+    speed_spikes = []
+
+    for si in shot_indices:
+        # Baseline speed: average speed in 100ms BEFORE the shot
+        baseline_speeds = []
+        for j in range(si - 1, -1, -1):
+            if frames[si].timestamp - frames[j].timestamp > 0.1:
+                break
+            baseline_speeds.append(speeds[j])
+
+        if not baseline_speeds:
+            continue
+
+        baseline_mean = sum(baseline_speeds) / len(baseline_speeds)
+        baseline_std = math.sqrt(sum((s - baseline_mean) ** 2 for s in baseline_speeds) / len(baseline_speeds)) if len(baseline_speeds) > 1 else baseline_mean * 0.1
+
+        # Threshold: speed is "contaminated" if it deviates by more than 3 sigma from baseline
+        threshold = baseline_mean + max(3.0 * baseline_std, 10.0)  # at least 10 cm/s spike
+
+        # Find how long after the shot the speed stays above threshold
+        max_speed = 0.0
+        last_contaminated_time = 0.0
+        for j in range(si, len(frames)):
+            dt_from_shot = frames[j].timestamp - frames[si].timestamp
+            if dt_from_shot > analysis_window_s:
+                break
+            if speeds[j] > threshold:
+                last_contaminated_time = dt_from_shot
+            if speeds[j] > max_speed:
+                max_speed = speeds[j]
+
+        contamination_durations.append(last_contaminated_time * 1000.0)  # in ms
+        speed_spikes.append(max_speed - baseline_mean)
+
+    if not contamination_durations:
+        return None
+
+    contamination_durations.sort()
+    return {
+        'num_shots': len(shot_indices),
+        'contamination_mean_ms': sum(contamination_durations) / len(contamination_durations),
+        'contamination_p95_ms': contamination_durations[int(len(contamination_durations) * 0.95)],
+        'contamination_max_ms': contamination_durations[-1],
+        'speed_spike_mean': sum(speed_spikes) / len(speed_spikes) if speed_spikes else 0,
+        'speed_spike_max': max(speed_spikes) if speed_spikes else 0,
+        'recommended_discard_ms': math.ceil(contamination_durations[int(len(contamination_durations) * 0.95)] / 5.0) * 5.0,  # round up to 5ms
+    }
+
+
+# --- Offline adaptive extrapolation simulation (matches C++ exactly) ---
+
+def simulate_adaptive(frames: List[Frame], params: AdaptiveParams) -> Tuple[List[float], List[float]]:
+    """
+    Simulate the adaptive extrapolation offline with given parameters.
+    Matches the C++ PredictAdaptiveExtrapolation algorithm exactly.
+    Optimized for speed: pre-extracts arrays, inlines math, avoids allocations.
+    """
+    pos_errors = []
+    aim_errors = []
+
+    n_frames = len(frames)
+    if n_frames < 4:
+        return pos_errors, aim_errors
+
+    # Pre-extract into flat arrays for speed
+    ts = [f.timestamp for f in frames]
+    px = [f.real_pos[0] for f in frames]
+    py = [f.real_pos[1] for f in frames]
+    pz = [f.real_pos[2] for f in frames]
+    ax = [f.real_aim[0] for f in frames]
+    ay = [f.real_aim[1] for f in frames]
+    az = [f.real_aim[2] for f in frames]
+
+    buffer_s = params.buffer_time_ms / 1000.0
+    discard_s = params.discard_time_ms / 1000.0
+    sensitivity = params.sensitivity
+    dead_zone = params.dead_zone
+    min_speed = params.min_speed
+    damping = params.damping
+    SMALL = 1e-10
+    _sqrt = math.sqrt
+    _exp = math.exp
+    _acos = math.acos
+    _degrees = math.degrees
+    _pow = pow
+
+    for i in range(2, n_frames - 1):
+        ct = ts[i]
+        safe_cutoff = ct - discard_s
+        oldest_allowed = ct - buffer_s
+
+        # Collect safe sample indices (backward scan, then reverse)
+        safe = []
+        for j in range(i, -1, -1):
+            t = ts[j]
+            if t < oldest_allowed:
+                break
+            if t <= safe_cutoff:
+                safe.append(j)
+        safe.reverse()
+
+        ns = len(safe)
+        if ns < 2:
+            continue
+
+        # Build velocity pairs inline
+        vpx = []; vpy = []; vpz = []
+        vax = []; vay = []; vaz = []
+        for k in range(1, ns):
+            p, c = safe[k - 1], safe[k]
+            dt = ts[c] - ts[p]
+            if dt > 1e-6:
+                inv_dt = 1.0 / dt
+                vpx.append((px[c] - px[p]) * inv_dt)
+                vpy.append((py[c] - py[p]) * inv_dt)
+                vpz.append((pz[c] - pz[p]) * inv_dt)
+                vax.append((ax[c] - ax[p]) * inv_dt)
+                vay.append((ay[c] - ay[p]) * inv_dt)
+                vaz.append((az[c] - az[p]) * inv_dt)
+
+        nv = len(vpx)
+        if nv < 2:
+            continue
+
+        # Weighted average velocity (quadratic weights, oldest=index 0)
+        tw = 0.0
+        apx = apy = apz = 0.0
+        aax = aay = aaz = 0.0
+        for k in range(nv):
+            w = (k + 1) * (k + 1)
+            apx += vpx[k] * w; apy += vpy[k] * w; apz += vpz[k] * w
+            aax += vax[k] * w; aay += vay[k] * w; aaz += vaz[k] * w
+            tw += w
+        inv_tw = 1.0 / tw
+        apx *= inv_tw; apy *= inv_tw; apz *= inv_tw
+        aax *= inv_tw; aay *= inv_tw; aaz *= inv_tw
+
+        # Recent velocity (last 25%, unweighted)
+        rs = max(0, nv - max(1, nv // 4))
+        rc = nv - rs
+        rpx = rpy = rpz = 0.0
+        rax = ray = raz = 0.0
+        for k in range(rs, nv):
+            rpx += vpx[k]; rpy += vpy[k]; rpz += vpz[k]
+            rax += vax[k]; ray += vay[k]; raz += vaz[k]
+        inv_rc = 1.0 / rc
+        rpx *= inv_rc; rpy *= inv_rc; rpz *= inv_rc
+        rax *= inv_rc; ray *= inv_rc; raz *= inv_rc
+
+        avg_ps = _sqrt(apx*apx + apy*apy + apz*apz)
+        avg_as = _sqrt(aax*aax + aay*aay + aaz*aaz)
+        rec_ps = _sqrt(rpx*rpx + rpy*rpy + rpz*rpz)
+        rec_as = _sqrt(rax*rax + ray*ray + raz*raz)
+
+        # Position confidence
+        pc = 1.0
+        if avg_ps > min_speed:
+            ratio = rec_ps / avg_ps
+            if ratio > 1.0: ratio = 1.0
+            if ratio < dead_zone:
+                rm = ratio / dead_zone if dead_zone > SMALL else 0.0
+                if rm > 1.0: rm = 1.0
+                pc = _pow(rm, sensitivity)
+
+        # Aim confidence
+        ac = 1.0
+        if avg_as > min_speed:
+            ratio = rec_as / avg_as
+            if ratio > 1.0: ratio = 1.0
+            if ratio < dead_zone:
+                rm = ratio / dead_zone if dead_zone > SMALL else 0.0
+                if rm > 1.0: rm = 1.0
+                ac = _pow(rm, sensitivity)
+
+        # Extrapolation time
+        lsi = safe[-1]
+        edt = ct - ts[lsi]
+        if edt <= 0: edt = 0.011
+
+        # Damping
+        ds = _exp(-damping * edt) if damping > 0.0 else 1.0
+
+        # Predict
+        m = edt * pc * ds
+        ppx = px[lsi] + apx * m
+        ppy = py[lsi] + apy * m
+        ppz = pz[lsi] + apz * m
+
+        ma = edt * ac * ds
+        pax_r = ax[lsi] + aax * ma
+        pay_r = ay[lsi] + aay * ma
+        paz_r = az[lsi] + aaz * ma
+        pa_len = _sqrt(pax_r*pax_r + pay_r*pay_r + paz_r*paz_r)
+
+        # Position error
+        dx = ppx - px[i]; dy = ppy - py[i]; dz = ppz - pz[i]
+        pos_errors.append(_sqrt(dx*dx + dy*dy + dz*dz))
+
+        # Aim error
+        if pa_len > 0.5:
+            inv_pa = 1.0 / pa_len
+            pax_n = pax_r * inv_pa; pay_n = pay_r * inv_pa; paz_n = paz_r * inv_pa
+            ra_len = _sqrt(ax[i]*ax[i] + ay[i]*ay[i] + az[i]*az[i])
+            if ra_len > 0.5:
+                inv_ra = 1.0 / ra_len
+                dot = pax_n * ax[i] * inv_ra + pay_n * ay[i] * inv_ra + paz_n * az[i] * inv_ra
+                if dot > 1.0: dot = 1.0
+                if dot < -1.0: dot = -1.0
+                aim_errors.append(_degrees(_acos(dot)))
+
+    return pos_errors, aim_errors
+
+
+# --- Scoring ---
+
+def compute_score(pos_errors: List[float], aim_errors: List[float]) -> ScoreResult:
+    """Compute a combined score from position and aim errors, including stability metrics."""
+    if not pos_errors:
+        return ScoreResult(0, 0, 0, 0, 0, 0, float('inf'))
+
+    pos_sorted = sorted(pos_errors)
+    aim_sorted = sorted(aim_errors) if aim_errors else [0]
+
+    pos_mean = sum(pos_errors) / len(pos_errors)
+    pos_p95 = pos_sorted[int(len(pos_sorted) * 0.95)]
+    aim_mean = sum(aim_errors) / len(aim_errors) if aim_errors else 0
+    aim_p95 = aim_sorted[int(len(aim_sorted) * 0.95)] if aim_errors else 0
+
+    # Jitter: standard deviation of frame-to-frame error change
+    jitter = 0.0
+    if len(pos_errors) > 1:
+        deltas = [abs(pos_errors[i] - pos_errors[i - 1]) for i in range(1, len(pos_errors))]
+        delta_mean = sum(deltas) / len(deltas)
+        jitter = math.sqrt(sum((d - delta_mean) ** 2 for d in deltas) / len(deltas))
+
+    # Overshoot: percentage of frames where error spikes above 2x mean
+    overshoot = 0.0
+    if pos_mean > 0:
+        overshoot_count = sum(1 for e in pos_errors if e > 2.0 * pos_mean)
+        overshoot = overshoot_count / len(pos_errors)
+
+    # Combined score
+    score = (pos_mean * 0.25 + pos_p95 * 0.15 +
+             aim_mean * 0.25 + aim_p95 * 0.15 +
+             jitter * 0.10 + overshoot * 0.10)
+
+    return ScoreResult(pos_mean, pos_p95, aim_mean, aim_p95, jitter, overshoot, score)
+
+
+def aggregate_scores(per_file_scores: List[Tuple[str, ScoreResult]],
+                     strategy: str = "mean") -> float:
+    """Aggregate scores across multiple files."""
+    scores = [s.score for _, s in per_file_scores]
+    if not scores:
+        return float('inf')
+    if strategy == "worst_case":
+        return max(scores)
+    else:  # mean
+        return sum(scores) / len(scores)
+
+
+# --- Optimizer ---
+
+def objective(x, all_frames, strategy):
+    """Objective function for the optimizer."""
+    params = AdaptiveParams(
+        sensitivity=x[0],
+        dead_zone=x[1],
+        min_speed=x[2],
+        damping=x[3],
+        buffer_time_ms=x[4],
+        discard_time_ms=x[5]
+    )
+    per_file_scores = []
+    for name, frames in all_frames:
+        pos_errors, aim_errors = simulate_adaptive(frames, params)
+        score_result = compute_score(pos_errors, aim_errors)
+        per_file_scores.append((name, score_result))
+    return aggregate_scores(per_file_scores, strategy)
+
+
+def optimize_differential_evolution(all_frames, strategy="mean", max_iter=200, min_discard_ms=10.0):
+    """Find optimal parameters using scipy differential evolution."""
+    try:
+        from scipy.optimize import differential_evolution
+    except ImportError:
+        print("ERROR: scipy is required for optimization.")
+        print("Install with: pip install scipy")
+        sys.exit(1)
+
+    bounds = [
+        (0.1, 5.0),    # sensitivity
+        (0.0, 0.95),   # dead_zone
+        (0.0, 200.0),  # min_speed
+        (0.0, 50.0),   # damping
+        (100.0, 500.0), # buffer_time_ms
+        (max(10.0, min_discard_ms), 100.0),  # discard_time_ms (floor from contamination analysis)
+    ]
+
+    print(f"\nRunning differential evolution (maxiter={max_iter}, popsize=25, min_discard={min_discard_ms:.0f}ms)...")
+    print("This may take a few minutes...\n")
+
+    result = differential_evolution(
+        objective,
+        bounds,
+        args=(all_frames, strategy),
+        maxiter=max_iter,
+        seed=42,
+        tol=1e-4,
+        popsize=25,
+        disp=True,
+        workers=1
+    )
+
+    best_params = AdaptiveParams(
+        sensitivity=round(result.x[0], 2),
+        dead_zone=round(result.x[1], 3),
+        min_speed=round(result.x[2], 1),
+        damping=round(result.x[3], 1),
+        buffer_time_ms=round(result.x[4], 0),
+        discard_time_ms=round(result.x[5], 0)
+    )
+    return best_params, result.fun
+
+
+def optimize_grid_search(all_frames, strategy="mean", min_discard_ms=10.0):
+    """Find optimal parameters using grid search (slower but no scipy needed)."""
+    print(f"\nRunning grid search over 6 parameters (min_discard={min_discard_ms:.0f}ms)...")
+
+    sensitivities = [1.0, 2.0, 3.0, 4.0]
+    dead_zones = [0.7, 0.8, 0.9]
+    min_speeds = [0.0, 30.0]
+    dampings = [5.0, 10.0, 15.0]
+    buffer_times = [300.0, 400.0, 500.0, 600.0, 800.0]
+    discard_times = [d for d in [20.0, 40.0, 60.0, 100.0, 150.0, 200.0] if d >= min_discard_ms]
+    if not discard_times:
+        discard_times = [min_discard_ms]
+
+    total = (len(sensitivities) * len(dead_zones) * len(min_speeds) *
+             len(dampings) * len(buffer_times) * len(discard_times))
+    print(f"Total combinations: {total}")
+
+    best_score = float('inf')
+    best_params = AdaptiveParams()
+    count = 0
+
+    for sens in sensitivities:
+        for dz in dead_zones:
+            for ms in min_speeds:
+                for damp in dampings:
+                    for bt in buffer_times:
+                        for dt in discard_times:
+                            count += 1
+                            if count % 500 == 0:
+                                print(f"  Progress: {count}/{total} ({100 * count / total:.0f}%) best={best_score:.4f}")
+
+                            params = AdaptiveParams(sens, dz, ms, damp, bt, dt)
+                            per_file_scores = []
+                            for name, frames in all_frames:
+                                pos_errors, aim_errors = simulate_adaptive(frames, params)
+                                score_result = compute_score(pos_errors, aim_errors)
+                                per_file_scores.append((name, score_result))
+
+                            score = aggregate_scores(per_file_scores, strategy)
+                            if score < best_score:
+                                best_score = score
+                                best_params = params
+
+    return best_params, best_score
+
+
+# --- Main ---
+
+def print_file_stats(name: str, frames: List[Frame]):
+    """Print basic stats for a CSV file."""
+    duration = frames[-1].timestamp - frames[0].timestamp
+    avg_fps = len(frames) / duration if duration > 0 else 0
+    avg_safe = sum(f.safe_count for f in frames) / len(frames)
+    avg_buffer = sum(f.buffer_count for f in frames) / len(frames)
+    avg_extrap = sum(f.extrap_time for f in frames) / len(frames) * 1000
+    num_shots = sum(1 for f in frames if f.shot_fired)
+    print(f"  {os.path.basename(name)}: {len(frames)} frames, {avg_fps:.0f}fps, "
+          f"{duration:.1f}s, safe={avg_safe:.1f}, extrap={avg_extrap:.1f}ms, shots={num_shots}")
+
+
+def print_score_detail(name: str, score: ScoreResult):
+    """Print detailed score for a file."""
+    print(f"  {os.path.basename(name):30s} Pos: mean={score.pos_mean:.3f}cm p95={score.pos_p95:.3f}cm | "
+          f"Aim: mean={score.aim_mean:.3f}deg p95={score.aim_p95:.3f}deg | "
+          f"jitter={score.jitter:.3f} overshoot={score.overshoot:.1%} | "
+          f"score={score.score:.4f}")
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="Anti-Recoil Parameter Optimizer - finds optimal AdaptiveExtrapolation parameters"
+    )
+    parser.add_argument("csv_files", nargs="+", help="One or more CSV recording files")
+    parser.add_argument("--plot", action="store_true", help="Generate comparison plots (requires matplotlib)")
+    parser.add_argument("--grid", action="store_true", help="Use grid search instead of differential evolution")
+    parser.add_argument("--strategy", choices=["mean", "worst_case"], default="mean",
+                        help="Multi-file score aggregation strategy (default: mean)")
+    parser.add_argument("--max-iter", type=int, default=200, help="Max optimizer iterations (default: 200)")
+    args = parser.parse_args()
+
+    # Load all CSV files
+    all_frames = []
+    for csv_path in args.csv_files:
+        if not os.path.exists(csv_path):
+            print(f"Error: File not found: {csv_path}")
+            sys.exit(1)
+        frames = load_csv(csv_path)
+        if len(frames) < 50:
+            print(f"Warning: {csv_path} has only {len(frames)} frames (need at least 50 for good results)")
+        all_frames.append((csv_path, frames))
+
+    print(f"\nLoaded {len(all_frames)} file(s)")
+    print("=" * 70)
+
+    # Per-file stats
+    print("\n=== FILE STATISTICS ===")
+    for name, frames in all_frames:
+        print_file_stats(name, frames)
+
+    # Shot contamination analysis
+    has_shots = any(any(f.shot_fired for f in frames) for _, frames in all_frames)
+    if has_shots:
+        print("\n=== SHOT CONTAMINATION ANALYSIS ===")
+        max_recommended_discard = 0.0
+        for name, frames in all_frames:
+            result = analyze_shot_contamination(frames)
+            if result:
+                print(f"  {os.path.basename(name)}:")
+                print(f"    Shots detected: {result['num_shots']}")
+                print(f"    Speed spike: mean={result['speed_spike_mean']:.1f} cm/s, max={result['speed_spike_max']:.1f} cm/s")
+                print(f"    Contamination duration: mean={result['contamination_mean_ms']:.1f}ms, "
+                      f"p95={result['contamination_p95_ms']:.1f}ms, max={result['contamination_max_ms']:.1f}ms")
+                print(f"    Recommended discard_time: >= {result['recommended_discard_ms']:.0f}ms")
+                max_recommended_discard = max(max_recommended_discard, result['recommended_discard_ms'])
+            else:
+                print(f"  {os.path.basename(name)}: no shots detected")
+
+        if max_recommended_discard > 0:
+            print(f"\n  >>> MINIMUM SAFE DiscardTime across all files: {max_recommended_discard:.0f}ms <<<")
+    else:
+        print("\n  (No ShotFired data in CSV - record with updated plugin to get contamination analysis)")
+
+    # Baseline: current default parameters
+    default_params = AdaptiveParams()
+    print(f"\n=== BASELINE (defaults: sens={default_params.sensitivity}, dz={default_params.dead_zone}, "
+          f"minspd={default_params.min_speed}, damp={default_params.damping}, "
+          f"buf={default_params.buffer_time_ms}ms, disc={default_params.discard_time_ms}ms) ===")
+
+    baseline_scores = []
+    for name, frames in all_frames:
+        pos_errors, aim_errors = simulate_adaptive(frames, default_params)
+        score = compute_score(pos_errors, aim_errors)
+        baseline_scores.append((name, score))
+        print_score_detail(name, score)
+
+    baseline_agg = aggregate_scores(baseline_scores, args.strategy)
+    print(f"\n  Aggregate score ({args.strategy}): {baseline_agg:.4f}")
+
+    # Also show recorded prediction error (as-is from the engine)
+    print(f"\n=== RECORDED PREDICTION ERROR (as captured in-engine) ===")
+    for name, frames in all_frames:
+        err = compute_prediction_error(frames)
+        print(f"  {os.path.basename(name):30s} Pos: mean={err['pos_mean']:.3f}cm p95={err['pos_p95']:.3f}cm | "
+              f"Aim: mean={err['aim_mean']:.3f}deg p95={err['aim_p95']:.3f}deg")
+
+    # Compute minimum safe discard time from shot contamination analysis
+    min_discard_ms = 10.0  # absolute minimum
+    if has_shots:
+        for name, frames in all_frames:
+            result = analyze_shot_contamination(frames)
+            if result and result['recommended_discard_ms'] > min_discard_ms:
+                min_discard_ms = result['recommended_discard_ms']
+
+    # Optimize
+    print(f"\n=== OPTIMIZATION ({args.strategy}) ===")
+
+    if args.grid:
+        best_params, best_score = optimize_grid_search(all_frames, args.strategy, min_discard_ms)
+    else:
+        best_params, best_score = optimize_differential_evolution(all_frames, args.strategy, args.max_iter, min_discard_ms)
+
+    # Results
+    print(f"\n{'=' * 70}")
+    print(f"  BEST PARAMETERS FOUND:")
+    print(f"    AdaptiveSensitivity    = {best_params.sensitivity}")
+    print(f"    AdaptiveDeadZone       = {best_params.dead_zone}")
+    print(f"    AdaptiveMinSpeed       = {best_params.min_speed}")
+    print(f"    ExtrapolationDamping   = {best_params.damping}")
+    print(f"    AntiRecoilBufferTimeMs = {best_params.buffer_time_ms}")
+    print(f"    AntiRecoilDiscardTimeMs= {best_params.discard_time_ms}")
+    print(f"{'=' * 70}")
+
+    # Per-file breakdown with optimized params
+    print(f"\n=== OPTIMIZED RESULTS ===")
+    opt_scores = []
+    for name, frames in all_frames:
+        pos_errors, aim_errors = simulate_adaptive(frames, best_params)
+        score = compute_score(pos_errors, aim_errors)
+        opt_scores.append((name, score))
+        print_score_detail(name, score)
+
+    opt_agg = aggregate_scores(opt_scores, args.strategy)
+    print(f"\n  Aggregate score ({args.strategy}): {opt_agg:.4f}")
+
+    # Improvement
+    print(f"\n=== IMPROVEMENT vs BASELINE ===")
+    for (name, baseline), (_, optimized) in zip(baseline_scores, opt_scores):
+        pos_pct = ((baseline.pos_mean - optimized.pos_mean) / baseline.pos_mean * 100) if baseline.pos_mean > 0 else 0
+        aim_pct = ((baseline.aim_mean - optimized.aim_mean) / baseline.aim_mean * 100) if baseline.aim_mean > 0 else 0
+        score_pct = ((baseline.score - optimized.score) / baseline.score * 100) if baseline.score > 0 else 0
+        print(f"  {os.path.basename(name):30s} Pos: {pos_pct:+.1f}% | Aim: {aim_pct:+.1f}% | Score: {score_pct:+.1f}%")
+
+    total_pct = ((baseline_agg - opt_agg) / baseline_agg * 100) if baseline_agg > 0 else 0
+    print(f"  {'TOTAL':30s} Score: {total_pct:+.1f}%")
+
+    # Plotting
+    if args.plot:
+        try:
+            import matplotlib.pyplot as plt
+
+            n_files = len(all_frames)
+            fig, axes = plt.subplots(n_files, 3, figsize=(18, 5 * n_files), squeeze=False)
+
+            for row, (name, frames) in enumerate(all_frames):
+                timestamps = [f.timestamp - frames[0].timestamp for f in frames]
+                short_name = os.path.basename(name)
+
+                # Baseline errors
+                bl_pos, bl_aim = simulate_adaptive(frames, default_params)
+                # Optimized errors
+                op_pos, op_aim = simulate_adaptive(frames, best_params)
+
+                # Time axis for simulated errors (offset by window_size)
+                t_start = window_size = 12
+                sim_timestamps = [frames[i].timestamp - frames[0].timestamp
+                                  for i in range(t_start + 1, t_start + 1 + len(bl_pos))]
+
+                # Position error
+                ax = axes[row][0]
+                if len(sim_timestamps) == len(bl_pos):
+                    ax.plot(sim_timestamps, bl_pos, 'r-', alpha=0.4, linewidth=0.5, label='Baseline')
+                    ax.plot(sim_timestamps, op_pos, 'g-', alpha=0.4, linewidth=0.5, label='Optimized')
+                ax.set_ylabel('Position Error (cm)')
+                ax.set_title(f'{short_name} - Position Error')
+                ax.legend()
+
+                # Aim error
+                ax = axes[row][1]
+                if len(sim_timestamps) >= len(bl_aim):
+                    t_aim = sim_timestamps[:len(bl_aim)]
+                    ax.plot(t_aim, bl_aim, 'r-', alpha=0.4, linewidth=0.5, label='Baseline')
+                if len(sim_timestamps) >= len(op_aim):
+                    t_aim = sim_timestamps[:len(op_aim)]
+                    ax.plot(t_aim, op_aim, 'g-', alpha=0.4, linewidth=0.5, label='Optimized')
+                ax.set_ylabel('Aim Error (deg)')
+                ax.set_title(f'{short_name} - Aim Error')
+                ax.legend()
+
+                # Speed profile
+                ax = axes[row][2]
+                speeds = [0]
+                for i in range(1, len(frames)):
+                    dt = frames[i].timestamp - frames[i - 1].timestamp
+                    if dt > 1e-6:
+                        d = vec_dist(frames[i].real_pos, frames[i - 1].real_pos)
+                        speeds.append(d / dt)
+                    else:
+                        speeds.append(speeds[-1])
+                ax.plot(timestamps, speeds, 'b-', alpha=0.7, linewidth=0.5)
+                ax.set_ylabel('Speed (cm/s)')
+                ax.set_xlabel('Time (s)')
+                ax.set_title(f'{short_name} - Speed Profile')
+
+            plt.tight_layout()
+            plot_path = args.csv_files[0].replace('.csv', '_optimizer.png')
+            plt.savefig(plot_path, dpi=150)
+            print(f"\nPlot saved: {plot_path}")
+            plt.show()
+
+        except ImportError:
+            print("\nmatplotlib not installed. Install with: pip install matplotlib")
+
+    print("\nDone.")
+
+
+if __name__ == '__main__':
+    main()

Tools/analyze_shots.py

@@ -0,0 +1,366 @@
+"""
+Shot Contamination Analyzer
+============================
+Analyzes the precise contamination zone around each shot event.
+Shows speed/acceleration profiles before and after each shot to identify
+the exact duration of IMU perturbation vs voluntary movement.
+
+Usage:
+    python analyze_shots.py <csv_file> [--plot] [--window 100]
+
+Protocol for best results:
+    1. Stay stable (no movement) for 2-3 seconds
+    2. Fire a single shot
+    3. Stay stable again for 2-3 seconds
+    4. Repeat 10+ times
+    This isolates the IMU shock from voluntary movement.
+"""
+
+import csv
+import sys
+import math
+import os
+import argparse
+from dataclasses import dataclass
+from typing import List, Tuple
+
+
+@dataclass
+class Frame:
+    timestamp: float
+    real_pos: Tuple[float, float, float]
+    real_aim: Tuple[float, float, float]
+    pred_pos: Tuple[float, float, float]
+    pred_aim: Tuple[float, float, float]
+    safe_count: int
+    buffer_count: int
+    extrap_time: float
+    shot_fired: bool = False
+
+
+def load_csv(path: str) -> List[Frame]:
+    frames = []
+    with open(path, 'r') as f:
+        reader = csv.DictReader(f)
+        for row in reader:
+            frames.append(Frame(
+                timestamp=float(row['Timestamp']),
+                real_pos=(float(row['RealPosX']), float(row['RealPosY']), float(row['RealPosZ'])),
+                real_aim=(float(row['RealAimX']), float(row['RealAimY']), float(row['RealAimZ'])),
+                pred_pos=(float(row['PredPosX']), float(row['PredPosY']), float(row['PredPosZ'])),
+                pred_aim=(float(row['PredAimX']), float(row['PredAimY']), float(row['PredAimZ'])),
+                safe_count=int(row['SafeCount']),
+                buffer_count=int(row['BufferCount']),
+                extrap_time=float(row['ExtrapolationTime']),
+                shot_fired=int(row.get('ShotFired', 0)) == 1,
+            ))
+    return frames
+
+
+def vec_dist(a, b):
+    return math.sqrt(sum((ai - bi) ** 2 for ai, bi in zip(a, b)))
+
+
+def vec_sub(a, b):
+    return tuple(ai - bi for ai, bi in zip(a, b))
+
+
+def vec_len(a):
+    return math.sqrt(sum(ai * ai for ai in a))
+
+
+def vec_normalize(a):
+    l = vec_len(a)
+    if l < 1e-10:
+        return (0, 0, 0)
+    return tuple(ai / l for ai in a)
+
+
+def angle_between(a, b):
+    dot = sum(ai * bi for ai, bi in zip(a, b))
+    dot = max(-1.0, min(1.0, dot))
+    return math.degrees(math.acos(dot))
+
+
+def compute_per_frame_metrics(frames):
+    """Compute speed, acceleration, and aim angular speed per frame."""
+    n = len(frames)
+    pos_speed = [0.0] * n
+    aim_speed = [0.0] * n
+    pos_accel = [0.0] * n
+
+    for i in range(1, n):
+        dt = frames[i].timestamp - frames[i - 1].timestamp
+        if dt > 1e-6:
+            pos_speed[i] = vec_dist(frames[i].real_pos, frames[i - 1].real_pos) / dt
+
+            aim_a = vec_normalize(frames[i].real_aim)
+            aim_b = vec_normalize(frames[i - 1].real_aim)
+            if vec_len(aim_a) > 0.5 and vec_len(aim_b) > 0.5:
+                aim_speed[i] = angle_between(aim_a, aim_b) / dt  # deg/s
+
+    for i in range(1, n):
+        dt = frames[i].timestamp - frames[i - 1].timestamp
+        if dt > 1e-6:
+            pos_accel[i] = (pos_speed[i] - pos_speed[i - 1]) / dt
+
+    return pos_speed, aim_speed, pos_accel
+
+
+def analyze_single_shot(frames, shot_idx, pos_speed, aim_speed, pos_accel, window_ms=200.0):
+    """Analyze contamination around a single shot event."""
+    window_s = window_ms / 1000.0
+    shot_time = frames[shot_idx].timestamp
+
+    # Collect frames in window before and after shot
+    before = []  # (time_relative_ms, pos_speed, aim_speed, pos_accel)
+    after = []
+
+    for i in range(max(0, shot_idx - 100), min(len(frames), shot_idx + 100)):
+        dt_ms = (frames[i].timestamp - shot_time) * 1000.0
+        if -window_ms <= dt_ms < 0:
+            before.append((dt_ms, pos_speed[i], aim_speed[i], pos_accel[i]))
+        elif dt_ms >= 0 and dt_ms <= window_ms:
+            after.append((dt_ms, pos_speed[i], aim_speed[i], pos_accel[i]))
+
+    if not before:
+        return None
+
+    # Baseline: average speed in the window before the shot
+    baseline_pos_speed = sum(s for _, s, _, _ in before) / len(before)
+    baseline_aim_speed = sum(s for _, _, s, _ in before) / len(before)
+    baseline_pos_std = math.sqrt(sum((s - baseline_pos_speed) ** 2 for _, s, _, _ in before) / len(before)) if len(before) > 1 else 0.0
+    baseline_aim_std = math.sqrt(sum((s - baseline_aim_speed) ** 2 for _, _, s, _ in before) / len(before)) if len(before) > 1 else 0.0
+
+    # Find contamination end: when speed returns to within 2 sigma of baseline
+    pos_threshold = baseline_pos_speed + max(2.0 * baseline_pos_std, 5.0)  # at least 5 cm/s
+    aim_threshold = baseline_aim_speed + max(2.0 * baseline_aim_std, 5.0)  # at least 5 deg/s
+
+    pos_contamination_end_ms = 0.0
+    aim_contamination_end_ms = 0.0
+    max_pos_spike = 0.0
+    max_aim_spike = 0.0
+
+    for dt_ms, ps, ais, _ in after:
+        if ps > pos_threshold:
+            pos_contamination_end_ms = dt_ms
+        if ais > aim_threshold:
+            aim_contamination_end_ms = dt_ms
+        max_pos_spike = max(max_pos_spike, ps - baseline_pos_speed)
+        max_aim_spike = max(max_aim_spike, ais - baseline_aim_speed)
+
+    return {
+        'shot_time': shot_time,
+        'baseline_pos_speed': baseline_pos_speed,
+        'baseline_aim_speed': baseline_aim_speed,
+        'baseline_pos_std': baseline_pos_std,
+        'baseline_aim_std': baseline_aim_std,
+        'pos_contamination_ms': pos_contamination_end_ms,
+        'aim_contamination_ms': aim_contamination_end_ms,
+        'max_contamination_ms': max(pos_contamination_end_ms, aim_contamination_end_ms),
+        'max_pos_spike': max_pos_spike,
+        'max_aim_spike': max_aim_spike,
+        'pos_threshold': pos_threshold,
+        'aim_threshold': aim_threshold,
+        'before': before,
+        'after': after,
+        'is_stable': baseline_pos_speed < 30.0 and baseline_aim_speed < 200.0,
+    }
+
+
+def main():
+    parser = argparse.ArgumentParser(description="Shot Contamination Analyzer")
+    parser.add_argument("csv_file", help="CSV recording file with ShotFired column")
+    parser.add_argument("--plot", action="store_true", help="Generate per-shot plots (requires matplotlib)")
+    parser.add_argument("--window", type=float, default=200.0, help="Analysis window in ms before/after shot (default: 200)")
+    args = parser.parse_args()
+
+    if not os.path.exists(args.csv_file):
+        print(f"Error: File not found: {args.csv_file}")
+        sys.exit(1)
+
+    frames = load_csv(args.csv_file)
+    print(f"Loaded {len(frames)} frames from {os.path.basename(args.csv_file)}")
+
+    duration = frames[-1].timestamp - frames[0].timestamp
+    fps = len(frames) / duration if duration > 0 else 0
+    print(f"Duration: {duration:.1f}s | FPS: {fps:.0f}")
+
+    shot_indices = [i for i, f in enumerate(frames) if f.shot_fired]
+    print(f"Shots detected: {len(shot_indices)}")
+
+    if not shot_indices:
+        print("No shots found! Make sure the CSV has a ShotFired column.")
+        sys.exit(1)
+
+    pos_speed, aim_speed, pos_accel = compute_per_frame_metrics(frames)
+
+    # Analyze each shot
+    results = []
+    print(f"\n{'=' * 90}")
+    print(f"{'Shot':>4} {'Time':>8} {'Stable':>7} {'PosSpike':>10} {'AimSpike':>10} "
+          f"{'PosContam':>10} {'AimContam':>10} {'MaxContam':>10}")
+    print(f"{'':>4} {'(s)':>8} {'':>7} {'(cm/s)':>10} {'(deg/s)':>10} "
+          f"{'(ms)':>10} {'(ms)':>10} {'(ms)':>10}")
+    print(f"{'-' * 90}")
+
+    for idx, si in enumerate(shot_indices):
+        result = analyze_single_shot(frames, si, pos_speed, aim_speed, pos_accel, args.window)
+        if result is None:
+            continue
+        results.append(result)
+
+        stable_str = "YES" if result['is_stable'] else "no"
+        print(f"{idx + 1:>4} {result['shot_time']:>8.2f} {stable_str:>7} "
+              f"{result['max_pos_spike']:>10.1f} {result['max_aim_spike']:>10.1f} "
+              f"{result['pos_contamination_ms']:>10.1f} {result['aim_contamination_ms']:>10.1f} "
+              f"{result['max_contamination_ms']:>10.1f}")
+
+    # Summary: only stable shots (user was not moving)
+    stable_results = [r for r in results if r['is_stable']]
+    all_results = results
+
+    print(f"\n{'=' * 90}")
+    print(f"SUMMARY - ALL SHOTS ({len(all_results)} shots)")
+    if all_results:
+        contam_all = sorted([r['max_contamination_ms'] for r in all_results])
+        pos_spikes = sorted([r['max_pos_spike'] for r in all_results])
+        aim_spikes = sorted([r['max_aim_spike'] for r in all_results])
+        p95_idx = int(len(contam_all) * 0.95)
+        print(f"  Contamination: mean={sum(contam_all)/len(contam_all):.1f}ms, "
+              f"median={contam_all[len(contam_all)//2]:.1f}ms, "
+              f"p95={contam_all[min(p95_idx, len(contam_all)-1)]:.1f}ms, "
+              f"max={contam_all[-1]:.1f}ms")
+        print(f"  Pos spike:     mean={sum(pos_spikes)/len(pos_spikes):.1f}cm/s, "
+              f"max={pos_spikes[-1]:.1f}cm/s")
+        print(f"  Aim spike:     mean={sum(aim_spikes)/len(aim_spikes):.1f}deg/s, "
+              f"max={aim_spikes[-1]:.1f}deg/s")
+
+    print(f"\nSUMMARY - STABLE SHOTS ONLY ({len(stable_results)} shots, baseline speed < 20cm/s)")
+    if stable_results:
+        contam_stable = sorted([r['max_contamination_ms'] for r in stable_results])
+        pos_spikes_s = sorted([r['max_pos_spike'] for r in stable_results])
+        aim_spikes_s = sorted([r['max_aim_spike'] for r in stable_results])
+        p95_idx = int(len(contam_stable) * 0.95)
+        print(f"  Contamination: mean={sum(contam_stable)/len(contam_stable):.1f}ms, "
+              f"median={contam_stable[len(contam_stable)//2]:.1f}ms, "
+              f"p95={contam_stable[min(p95_idx, len(contam_stable)-1)]:.1f}ms, "
+              f"max={contam_stable[-1]:.1f}ms")
+        print(f"  Pos spike:     mean={sum(pos_spikes_s)/len(pos_spikes_s):.1f}cm/s, "
+              f"max={pos_spikes_s[-1]:.1f}cm/s")
+        print(f"  Aim spike:     mean={sum(aim_spikes_s)/len(aim_spikes_s):.1f}deg/s, "
+              f"max={aim_spikes_s[-1]:.1f}deg/s")
+
+        recommended = math.ceil(contam_stable[min(p95_idx, len(contam_stable)-1)] / 5.0) * 5.0
+        print(f"\n  >>> RECOMMENDED DiscardTime (from stable shots P95): {recommended:.0f}ms <<<")
+    else:
+        print("  No stable shots found! Make sure you stay still before firing.")
+        print("  Shots where baseline speed > 20cm/s are excluded as 'not stable'.")
+
+    # Plotting
+    if args.plot:
+        try:
+            import matplotlib.pyplot as plt
+
+            # Plot each shot individually
+            n_shots = len(results)
+            cols = min(4, n_shots)
+            rows = math.ceil(n_shots / cols)
+            fig, axes = plt.subplots(rows, cols, figsize=(5 * cols, 4 * rows), squeeze=False)
+            fig.suptitle(f'Per-Shot Speed Profile ({os.path.basename(args.csv_file)})', fontsize=14)
+
+            for idx, result in enumerate(results):
+                r, c = divmod(idx, cols)
+                ax = axes[r][c]
+
+                # Before shot
+                if result['before']:
+                    t_before = [b[0] for b in result['before']]
+                    s_before = [b[1] for b in result['before']]
+                    ax.plot(t_before, s_before, 'b-', linewidth=1, label='Before')
+
+                # After shot
+                if result['after']:
+                    t_after = [a[0] for a in result['after']]
+                    s_after = [a[1] for a in result['after']]
+                    ax.plot(t_after, s_after, 'r-', linewidth=1, label='After')
+
+                # Shot line
+                ax.axvline(x=0, color='red', linestyle='--', alpha=0.7, label='Shot')
+
+                # Threshold
+                ax.axhline(y=result['pos_threshold'], color='orange', linestyle=':', alpha=0.5, label='Threshold')
+
+                # Contamination zone
+                if result['pos_contamination_ms'] > 0:
+                    ax.axvspan(0, result['pos_contamination_ms'], alpha=0.15, color='red')
+
+                stable_str = "STABLE" if result['is_stable'] else "MOVING"
+                ax.set_title(f"Shot {idx+1} ({stable_str}) - {result['max_contamination_ms']:.0f}ms",
+                             fontsize=9, color='green' if result['is_stable'] else 'orange')
+                ax.set_xlabel('Time from shot (ms)', fontsize=8)
+                ax.set_ylabel('Pos Speed (cm/s)', fontsize=8)
+                ax.tick_params(labelsize=7)
+                if idx == 0:
+                    ax.legend(fontsize=6)
+
+            # Hide unused subplots
+            for idx in range(n_shots, rows * cols):
+                r, c = divmod(idx, cols)
+                axes[r][c].set_visible(False)
+
+            plt.tight_layout()
+            plot_path = args.csv_file.replace('.csv', '_shots.png')
+            plt.savefig(plot_path, dpi=150)
+            print(f"\nPlot saved: {plot_path}")
+            plt.show()
+
+            # Also plot aim speed
+            fig2, axes2 = plt.subplots(rows, cols, figsize=(5 * cols, 4 * rows), squeeze=False)
+            fig2.suptitle(f'Per-Shot Aim Angular Speed ({os.path.basename(args.csv_file)})', fontsize=14)
+
+            for idx, result in enumerate(results):
+                r, c = divmod(idx, cols)
+                ax = axes2[r][c]
+
+                if result['before']:
+                    t_before = [b[0] for b in result['before']]
+                    a_before = [b[2] for b in result['before']]  # aim_speed
+                    ax.plot(t_before, a_before, 'b-', linewidth=1)
+
+                if result['after']:
+                    t_after = [a[0] for a in result['after']]
+                    a_after = [a[2] for a in result['after']]  # aim_speed
+                    ax.plot(t_after, a_after, 'r-', linewidth=1)
+
+                ax.axvline(x=0, color='red', linestyle='--', alpha=0.7)
+                ax.axhline(y=result['aim_threshold'], color='orange', linestyle=':', alpha=0.5)
+
+                if result['aim_contamination_ms'] > 0:
+                    ax.axvspan(0, result['aim_contamination_ms'], alpha=0.15, color='red')
+
+                stable_str = "STABLE" if result['is_stable'] else "MOVING"
+                ax.set_title(f"Shot {idx+1} ({stable_str}) - Aim {result['aim_contamination_ms']:.0f}ms",
+                             fontsize=9, color='green' if result['is_stable'] else 'orange')
+                ax.set_xlabel('Time from shot (ms)', fontsize=8)
+                ax.set_ylabel('Aim Speed (deg/s)', fontsize=8)
+                ax.tick_params(labelsize=7)
+
+            for idx in range(n_shots, rows * cols):
+                r, c = divmod(idx, cols)
+                axes2[r][c].set_visible(False)
+
+            plt.tight_layout()
+            plot_path2 = args.csv_file.replace('.csv', '_shots_aim.png')
+            plt.savefig(plot_path2, dpi=150)
+            print(f"Plot saved: {plot_path2}")
+            plt.show()
+
+        except ImportError:
+            print("\nmatplotlib not installed. Install with: pip install matplotlib")
+
+    print("\nDone.")
+
+
+if __name__ == '__main__':
+    main()

Unreal/Config/DefaultEngine.ini

@@ -1,7 +1,8 @@
 
 
 [/Script/EngineSettings.GameMapsSettings]
-GameDefaultMap=/Engine/Maps/Templates/OpenWorld
+GameDefaultMap=/Game/PACKS/Blueprints/EasyBallistics/world/ExampleMap.ExampleMap
+EditorStartupMap=/Game/PACKS/Blueprints/EasyBallistics/world/ExampleMap.ExampleMap
 
 [/Script/Engine.RendererSettings]
 r.AllowStaticLighting=False