Fix 4 test failures from first CI run

CHANGELOG.md

@@ -6,6 +6,13 @@ Format: [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); versioning: [S
 ## [Unreleased]
 
 ### Fixed
+- `utils.determine_strategy_side`: signal dtype was platform-dependent (int32 on
+  Windows); now always int64.
+- `data.labeling.calculate_t_value_linear_regression`: constant series now returns
+  NaN (0/0 is undefined) as documented, instead of 0.0.
+- `data.differentiation.fractional_difference_std`: result had object dtype; now float.
+- Raw-string docstrings in `denoising`/`bekker_parkinson` (LaTeX `\(` raised
+  SyntaxWarning on Python 3.12).
 - `backtest.bet_sizing`: `avgActiveSignals`/`Signal` silently returned empty results — the module
   imported `mpPandasObj` while `RiskLabAI.hpc` exports `mp_pandas_obj`, so a placeholder returning
   an empty DataFrame always took over. The silent placeholder is removed; the real import is used.

RiskLabAI/data/denoise/denoising.py

@@ -95,7 +95,7 @@ def fit_kde(
 def _mp_pdf_fit_error(
     variance: float, q: float, eigenvalues: np.ndarray, bandwidth: float
 ) -> float: # <-- FIX 2: Added bandwidth
-    """
+    r"""
     Error function for fitting the MP PDF to observed eigenvalues.
 
     Calculates the sum of squared errors between the theoretical
@@ -136,7 +136,7 @@ def _mp_pdf_fit_error(
 def find_max_eval(
     eigenvalues: np.ndarray, q: float, bandwidth: float
 ) -> Tuple[float, float]:
-    """
+    r"""
     Find the maximum theoretical eigenvalue (\(\lambda_{max}\))
     by fitting the Marcenko-Pastur distribution.
 

RiskLabAI/data/differentiation/differentiation.py

@@ -111,7 +111,7 @@ def fractional_difference_std(
     weights_cumsum_abs /= weights_cumsum_abs[-1]
     skip = np.searchsorted(weights_cumsum_abs, threshold)
 
-    result_df = pd.DataFrame(index=series.index, columns=series.columns)
+    result_df = pd.DataFrame(index=series.index, columns=series.columns, dtype=float)
 
     for name in series.columns:
         # Use .ffill() - fillna(method=) is deprecated

RiskLabAI/data/labeling/financial_labels.py

@@ -41,8 +41,9 @@ def calculate_t_value_linear_regression(prices: pd.Series) -> float:
         return np.nan
 
     if ols.stderr == 0:
-        # Handle perfect fit (vertical line, or constant)
-        return np.sign(ols.slope) * np.inf if ols.slope != 0 else 0.0
+        # Perfect trend: t -> +/- inf. Constant series: 0/0 -> undefined (NaN),
+        # consistent with the documented contract.
+        return np.sign(ols.slope) * np.inf if ols.slope != 0 else np.nan
 
     return ols.slope / ols.stderr
 

RiskLabAI/features/microstructural_features/bekker_parkinson_volatility_estimator.py

@@ -52,7 +52,7 @@ def sigma_estimates(beta: pd.Series, gamma: pd.Series) -> pd.Series:
 def bekker_parkinson_volatility_estimates(
     high_prices: pd.Series, low_prices: pd.Series, window_span: int = 20
 ) -> pd.Series:
-    """
+    r"""
     Compute Bekker-Parkinson volatility estimates from high and low prices.
 
     This function first calculates the Corwin-Schultz \(\beta\) and \(\gamma\)

RiskLabAI/utils/momentum_mean_reverting_strategy_sides.py

@@ -58,7 +58,9 @@ def determine_strategy_side(
         slow_ma = prices.rolling(window=slow_window, min_periods=1).mean()
 
     # Create signal: 1 if fast > slow, -1 if fast < slow
-    signal = (fast_ma >= slow_ma).astype(int) * 2 - 1
+    # Explicit int64: plain `int` maps to int32 on Windows, which makes the
+    # returned dtype platform-dependent.
+    signal = (fast_ma >= slow_ma).astype("int64") * 2 - 1
 
     if mean_reversion:
         # Invert the signal for mean reversion