upgrade xauusd atr breakout continuation

templates/xauusd-atr-breakout/README.md

@@ -1,13 +1,27 @@
-# XAUUSD ATR Breakout
+# XAUUSD ATR Breakout Continuation
 
-Custom Python breakout baseline for XAUUSD.
+Custom Python breakout + continuation baseline for XAUUSD.
 
-This project avoids ML on purpose. It is useful as a transparent baseline to compare against heavier gold models like XGBoost or LightGBM.
+This project avoids ML on purpose. It is useful as a transparent benchmark to compare against heavier gold models like XGBoost or LightGBM.
 
-The default gate is intentionally responsive so PPE produces more events than a strict long-range breakout filter:
+The default gate is intentionally responsive so PPE produces more events than a strict long-range breakout filter. It combines:
 
+- hard ATR breakout entries
+- near-breakout pressure entries
+- trend-continuation entries after range pressure
+- EMA trend alignment
+- short momentum measured in ATR units
+
+Current defaults:
+
+- `lookback`: `96`
 - `breakout_window`: `12`
 - `atr_mult`: `0.05`
+- `near_breakout_atr`: `0.18`
+- `pullback_atr`: `0.35`
+- `min_momentum_atr`: `0.08`
+- `fast_ema`: `8`
+- `slow_ema`: `34`
 - assigned runtime: Modal, because this starter uses `pandas`
 
-If it overtrades, raise `atr_mult` first. If it is still too quiet, shorten `breakout_window`.
+If it overtrades, raise `near_breakout_atr` more carefully than `atr_mult`: `atr_mult` controls hard breakout distance, while `near_breakout_atr` and `pullback_atr` control the extra continuation paths.

templates/xauusd-atr-breakout/quant.config.json

@@ -5,10 +5,15 @@
   "runtime_target": "modal",
   "artifact_format": "python_bundle",
   "parameters": {
-    "lookback": 64,
+    "lookback": 96,
     "atr_window": 14,
     "breakout_window": 12,
-    "atr_mult": 0.05
+    "atr_mult": 0.05,
+    "near_breakout_atr": 0.18,
+    "pullback_atr": 0.35,
+    "min_momentum_atr": 0.08,
+    "fast_ema": 8,
+    "slow_ema": 34
   },
   "training_requirements": [
     "numpy",

templates/xauusd-atr-breakout/strategy.py

@@ -15,42 +15,118 @@ def _atr(df, n):
     return tr.ewm(span=n, adjust=False).mean()
 
 
+def _ema(series, n):
+    return series.ewm(span=n, adjust=False).mean()
+
+
+def _clip01(value):
+    return float(np.clip(value, 0.0, 1.0))
+
+
 def train(data, config):
     params = config.get("parameters", {})
     return {
         "params": {
-            "lookback": int(params.get("lookback", 64)),
+            "lookback": int(params.get("lookback", 96)),
             "atr_window": int(params.get("atr_window", 14)),
             "breakout_window": int(params.get("breakout_window", 12)),
             "atr_mult": float(params.get("atr_mult", 0.05)),
+            "near_breakout_atr": float(params.get("near_breakout_atr", 0.18)),
+            "pullback_atr": float(params.get("pullback_atr", 0.35)),
+            "min_momentum_atr": float(params.get("min_momentum_atr", 0.08)),
+            "fast_ema": int(params.get("fast_ema", 8)),
+            "slow_ema": int(params.get("slow_ema", 34)),
         },
-        "name": "xauusd_atr_breakout",
-    }, {"training_bars": int(len(data)), "model": "modal_python_rule_baseline"}
+        "name": "xauusd_atr_breakout_v2",
+    }, {"training_bars": int(len(data)), "model": "modal_python_rule_breakout_continuation"}
 
 
 def predict(model, market_data, config):
     params = {**model.get("params", {}), **config.get("parameters", {})}
-    lookback = int(params.get("lookback", 64))
+    lookback = int(params.get("lookback", 96))
     atr_window = int(params.get("atr_window", 14))
     breakout_window = int(params.get("breakout_window", 12))
     atr_mult = float(params.get("atr_mult", 0.05))
+    near_breakout_atr = float(params.get("near_breakout_atr", 0.18))
+    pullback_atr = float(params.get("pullback_atr", 0.35))
+    min_momentum_atr = float(params.get("min_momentum_atr", 0.08))
+    fast_ema = int(params.get("fast_ema", 8))
+    slow_ema = int(params.get("slow_ema", 34))
     candles = market_data.get("candles", [])
     if len(candles) < lookback:
         return {"signal": "HOLD", "confidence": 0.0, "metadata": {"reason": "not_enough_candles"}}
 
     df = _df(candles[-lookback:])
     atr = float(_atr(df, atr_window).iloc[-1])
+    if not np.isfinite(atr) or atr <= 0:
+        return {"signal": "HOLD", "confidence": 0.0, "metadata": {"reason": "invalid_atr"}}
+
     close = float(df["close"].iloc[-1])
+    prev_close = float(df["close"].iloc[-2])
     high = float(df["high"].iloc[-breakout_window:-1].max())
     low = float(df["low"].iloc[-breakout_window:-1].min())
     upper = high + atr * atr_mult
     lower = low - atr * atr_mult
+    range_width = max(high - low, atr)
+
+    ema_fast = float(_ema(df["close"], fast_ema).iloc[-1])
+    ema_slow = float(_ema(df["close"], slow_ema).iloc[-1])
+    ema_gap_atr = (ema_fast - ema_slow) / atr
+    momentum_3 = (close - float(df["close"].iloc[-4])) / atr if len(df) >= 4 else 0.0
+    momentum_6 = (close - float(df["close"].iloc[-7])) / atr if len(df) >= 7 else momentum_3
+    body_atr = (close - float(df["open"].iloc[-1])) / atr
+    close_position = (close - low) / range_width
+
+    trend_up = ema_fast > ema_slow and close >= ema_slow and momentum_3 > -min_momentum_atr
+    trend_down = ema_fast < ema_slow and close <= ema_slow and momentum_3 < min_momentum_atr
+
+    meta = {
+        "atr": round(atr, 5),
+        "high": round(high, 5),
+        "low": round(low, 5),
+        "upper": round(upper, 5),
+        "lower": round(lower, 5),
+        "ema_gap_atr": round(float(ema_gap_atr), 4),
+        "momentum_3_atr": round(float(momentum_3), 4),
+        "momentum_6_atr": round(float(momentum_6), 4),
+        "close_position": round(float(close_position), 4),
+    }
+
+    def confidence(edge_atr, quality, floor=0.56):
+        return round(min(floor + edge_atr * 0.22 + quality * 0.16, 0.92), 4)
 
     if close > upper:
         edge = (close - upper) / max(atr, 1e-9)
-        return {"signal": "UP", "confidence": round(min(0.55 + edge, 0.92), 4), "metadata": {"breakout": "high", "atr": atr, "upper": upper}}
+        quality = _clip01(max(ema_gap_atr, 0.0) * 0.35 + max(momentum_3, 0.0) * 0.35 + max(body_atr, 0.0) * 0.20)
+        return {"signal": "UP", "confidence": confidence(edge, quality), "metadata": {**meta, "setup": "breakout_high"}}
     if close < lower:
         edge = (lower - close) / max(atr, 1e-9)
-        return {"signal": "DOWN", "confidence": round(min(0.55 + edge, 0.92), 4), "metadata": {"breakout": "low", "atr": atr, "lower": lower}}
+        quality = _clip01(max(-ema_gap_atr, 0.0) * 0.35 + max(-momentum_3, 0.0) * 0.35 + max(-body_atr, 0.0) * 0.20)
+        return {"signal": "DOWN", "confidence": confidence(edge, quality), "metadata": {**meta, "setup": "breakout_low"}}
 
-    return {"signal": "HOLD", "confidence": 0.2, "metadata": {"high": high, "low": low, "atr": atr, "upper": upper, "lower": lower}}
+    # If price is pressing a range boundary with trend/momentum confirmation,
+    # enter before the hard breakout instead of waiting for an extreme close.
+    upper_pressure = (high - close) / atr
+    lower_pressure = (close - low) / atr
+    if trend_up and upper_pressure <= near_breakout_atr and momentum_3 >= min_momentum_atr:
+        edge = max(near_breakout_atr - upper_pressure, 0.0)
+        quality = _clip01(max(ema_gap_atr, 0.0) * 0.35 + max(momentum_3, 0.0) * 0.35 + close_position * 0.20)
+        return {"signal": "UP", "confidence": confidence(edge, quality, floor=0.57), "metadata": {**meta, "setup": "near_breakout_high"}}
+    if trend_down and lower_pressure <= near_breakout_atr and momentum_3 <= -min_momentum_atr:
+        edge = max(near_breakout_atr - lower_pressure, 0.0)
+        quality = _clip01(max(-ema_gap_atr, 0.0) * 0.35 + max(-momentum_3, 0.0) * 0.35 + (1.0 - close_position) * 0.20)
+        return {"signal": "DOWN", "confidence": confidence(edge, quality, floor=0.57), "metadata": {**meta, "setup": "near_breakout_low"}}
+
+    # Continuation path: after a breakout, gold often retests the fast EMA
+    # without closing outside the range again. This keeps the system alive
+    # while still requiring trend and momentum context.
+    if trend_up and close > high - atr * pullback_atr and prev_close <= close and momentum_6 >= 0:
+        edge = max((close - (high - atr * pullback_atr)) / atr, 0.0)
+        quality = _clip01(max(ema_gap_atr, 0.0) * 0.35 + max(momentum_6, 0.0) * 0.25 + close_position * 0.25)
+        return {"signal": "UP", "confidence": confidence(edge, quality, floor=0.55), "metadata": {**meta, "setup": "trend_continuation_high"}}
+    if trend_down and close < low + atr * pullback_atr and prev_close >= close and momentum_6 <= 0:
+        edge = max(((low + atr * pullback_atr) - close) / atr, 0.0)
+        quality = _clip01(max(-ema_gap_atr, 0.0) * 0.35 + max(-momentum_6, 0.0) * 0.25 + (1.0 - close_position) * 0.25)
+        return {"signal": "DOWN", "confidence": confidence(edge, quality, floor=0.55), "metadata": {**meta, "setup": "trend_continuation_low"}}
+
+    return {"signal": "HOLD", "confidence": 0.2, "metadata": {**meta, "setup": "inside_range"}}