diff --git a/app/strategies/quant_algos/momentum_algos.py b/app/strategies/quant_algos/momentum_algos.py
index 8dcbd6b..d0cb938 100644
--- a/app/strategies/quant_algos/momentum_algos.py
+++ b/app/strategies/quant_algos/momentum_algos.py
@@ -14,12 +14,13 @@ def _closes(candles: List[dict]) -> List[float]:
 def _atr(candles: List[dict], period: int = 14) -> Optional[float]:
     if not candles or len(candles) < period + 1:
         return None
-    tr_list = []
-    for i in range(1, len(candles)):
-        h, l_ = candles[i]["high"], candles[i]["low"]
-        pc = candles[i - 1]["close"]
-        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
-    return sum(tr_list[-period:]) / period if len(tr_list) >= period else None
+    recent = candles[-(period + 1):]
+    tr_sum = 0.0
+    for i in range(1, len(recent)):
+        h, l_ = recent[i]["high"], recent[i]["low"]
+        pc = recent[i - 1]["close"]
+        tr_sum += max(h - l_, abs(h - pc), abs(l_ - pc))
+    return tr_sum / period
 
 
 def simple_momentum(
@@ -163,15 +164,16 @@ def adx_filter(
     if not candles or len(candles) < period * 2:
         return 0.0, f"warming up ({len(candles)}/{period*2})"
     # Simplified ADX: use +DM -DM from high/low/close
+    recent = candles[-(period + 1):]
     plus_dm = []
     minus_dm = []
     tr_list = []
-    for i in range(1, len(candles)):
-        h, l_, c = candles[i]["high"], candles[i]["low"], candles[i]["close"]
-        pc = candles[i - 1]["close"]
+    for i in range(1, len(recent)):
+        h, l_, c = recent[i]["high"], recent[i]["low"], recent[i]["close"]
+        pc = recent[i - 1]["close"]
         tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
-        up = h - candles[i - 1]["high"]
-        down = candles[i - 1]["low"] - l_
+        up = h - recent[i - 1]["high"]
+        down = recent[i - 1]["low"] - l_
         plus_dm.append(up if up > down and up > 0 else 0)
         minus_dm.append(down if down > up and down > 0 else 0)
     # Smooth and compute +DI -DI then ADX (simplified)
diff --git a/app/strategies/quant_algos/risk_sizing.py b/app/strategies/quant_algos/risk_sizing.py
index 06b7dc5..244d47f 100644
--- a/app/strategies/quant_algos/risk_sizing.py
+++ b/app/strategies/quant_algos/risk_sizing.py
@@ -7,12 +7,13 @@
 def _atr(candles: List[dict], period: int = 14) -> Optional[float]:
     if not candles or len(candles) < period + 1:
         return None
-    tr_list = []
-    for i in range(1, len(candles)):
-        h, l_ = candles[i]["high"], candles[i]["low"]
-        pc = candles[i - 1]["close"]
-        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
-    return sum(tr_list[-period:]) / period if len(tr_list) >= period else None
+    recent = candles[-(period + 1):]
+    tr_sum = 0.0
+    for i in range(1, len(recent)):
+        h, l_ = recent[i]["high"], recent[i]["low"]
+        pc = recent[i - 1]["close"]
+        tr_sum += max(h - l_, abs(h - pc), abs(l_ - pc))
+    return tr_sum / period
 
 
 def kelly_size(
diff --git a/app/strategies/quant_algos/smart_money.py b/app/strategies/quant_algos/smart_money.py
index e482d9d..42b6ffd 100644
--- a/app/strategies/quant_algos/smart_money.py
+++ b/app/strategies/quant_algos/smart_money.py
@@ -22,12 +22,13 @@
 def _atr(candles: List[dict], period: int = 14) -> float:
     if len(candles) < period + 1:
         return 0.0
-    tr_list = []
-    for i in range(1, len(candles)):
-        h, l_ = candles[i]["high"], candles[i]["low"]
-        pc = candles[i - 1]["close"]
-        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
-    return sum(tr_list[-period:]) / period if len(tr_list) >= period else 0.0
+    recent = candles[-(period + 1):]
+    tr_sum = 0.0
+    for i in range(1, len(recent)):
+        h, l_ = recent[i]["high"], recent[i]["low"]
+        pc = recent[i - 1]["close"]
+        tr_sum += max(h - l_, abs(h - pc), abs(l_ - pc))
+    return tr_sum / period
 
 
 def _obv_series(candles: List[dict]) -> List[float]:
diff --git a/app/strategies/quant_algos/volatility_algos.py b/app/strategies/quant_algos/volatility_algos.py
index a8aa1f1..f065921 100644
--- a/app/strategies/quant_algos/volatility_algos.py
+++ b/app/strategies/quant_algos/volatility_algos.py
@@ -8,12 +8,13 @@
 def _atr(candles: List[dict], period: int = 14) -> Optional[float]:
     if not candles or len(candles) < period + 1:
         return None
-    tr_list = []
-    for i in range(1, len(candles)):
-        h, l_ = candles[i]["high"], candles[i]["low"]
-        pc = candles[i - 1]["close"]
-        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
-    return sum(tr_list[-period:]) / period if len(tr_list) >= period else None
+    recent = candles[-(period + 1):]
+    tr_sum = 0.0
+    for i in range(1, len(recent)):
+        h, l_ = recent[i]["high"], recent[i]["low"]
+        pc = recent[i - 1]["close"]
+        tr_sum += max(h - l_, abs(h - pc), abs(l_ - pc))
+    return tr_sum / period
 
 
 def atr_breakout(
diff --git a/shared/tests/test_atr_tail_window.py b/shared/tests/test_atr_tail_window.py
new file mode 100644
index 0000000..aea9c31
--- /dev/null
+++ b/shared/tests/test_atr_tail_window.py
@@ -0,0 +1,81 @@
+from importlib import import_module
+
+from app.strategies.quant_algos import momentum_algos, risk_sizing, volatility_algos
+
+
+smart_money = import_module("app.strategies.quant_algos.smart_money")
+
+
+def _candles(count):
+    out = []
+    for i in range(count):
+        close = 100 + ((i * 7) % 31) + (i * 0.02)
+        high = close + 1.5 + ((i * 3) % 5) * 0.1
+        low = close - 1.2 - ((i * 5) % 7) * 0.1
+        out.append({"high": high, "low": low, "close": close, "volume": 1000 + i})
+    return out
+
+
+def _legacy_atr(candles, period=14, none_on_warmup=True):
+    if not candles or len(candles) < period + 1:
+        return None if none_on_warmup else 0.0
+    tr_list = []
+    for i in range(1, len(candles)):
+        h, l_ = candles[i]["high"], candles[i]["low"]
+        pc = candles[i - 1]["close"]
+        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
+    fallback = None if none_on_warmup else 0.0
+    return sum(tr_list[-period:]) / period if len(tr_list) >= period else fallback
+
+
+def _legacy_adx_filter(candles, period=14, threshold=25.0):
+    if not candles or len(candles) < period * 2:
+        return 0.0, f"warming up ({len(candles)}/{period*2})"
+    plus_dm = []
+    minus_dm = []
+    tr_list = []
+    for i in range(1, len(candles)):
+        h, l_ = candles[i]["high"], candles[i]["low"]
+        pc = candles[i - 1]["close"]
+        tr_list.append(max(h - l_, abs(h - pc), abs(l_ - pc)))
+        up = h - candles[i - 1]["high"]
+        down = candles[i - 1]["low"] - l_
+        plus_dm.append(up if up > down and up > 0 else 0)
+        minus_dm.append(down if down > up and down > 0 else 0)
+    n = period
+    if len(tr_list) < n:
+        return 0.0, "insufficient data"
+    atr = sum(tr_list[-n:]) / n
+    plus_di = 100 * sum(plus_dm[-n:]) / n / atr if atr else 0
+    minus_di = 100 * sum(minus_dm[-n:]) / n / atr if atr else 0
+    dx = 100 * abs(plus_di - minus_di) / (plus_di + minus_di) if (plus_di + minus_di) else 0
+    sig = 1.0 if dx >= threshold else 0.0
+    return sig, f"ADX={dx:.1f} (threshold={threshold})"
+
+
+def test_atr_helpers_match_legacy_tail_result_on_long_history():
+    candles = _candles(5000)
+
+    assert risk_sizing._atr(candles, period=14) == _legacy_atr(candles, period=14)
+    assert volatility_algos._atr(candles, period=14) == _legacy_atr(candles, period=14)
+    assert momentum_algos._atr(candles, period=14) == _legacy_atr(candles, period=14)
+    assert smart_money._atr(candles, period=14) == _legacy_atr(
+        candles, period=14, none_on_warmup=False
+    )
+
+
+def test_adx_filter_matches_legacy_tail_result_on_long_history():
+    candles = _candles(5000)
+
+    assert momentum_algos.adx_filter(candles, period=14) == _legacy_adx_filter(
+        candles, period=14
+    )
+
+
+def test_atr_helpers_keep_warmup_contracts():
+    candles = _candles(4)
+
+    assert risk_sizing._atr(candles, period=14) is None
+    assert volatility_algos._atr(candles, period=14) is None
+    assert momentum_algos._atr(candles, period=14) is None
+    assert smart_money._atr(candles, period=14) == 0.0