diff --git a/tests/indicators/__init__.py b/tests/indicators/__init__.py
new file mode 100644
index 0000000..b382bfd
--- /dev/null
+++ b/tests/indicators/__init__.py
@@ -0,0 +1 @@
+# Marker for indicators test package
\ No newline at end of file
diff --git a/tests/indicators/conftest.py b/tests/indicators/conftest.py
new file mode 100644
index 0000000..8c7c842
--- /dev/null
+++ b/tests/indicators/conftest.py
@@ -0,0 +1,31 @@
+import pytest
+import polars as pl
+
+@pytest.fixture
+def sample_ohlcv_df():
+    """
+    Returns a deterministic polars DataFrame with 120 rows and standard OHLCV columns.
+    Values are monotonically increasing for easy/deterministic indicator output.
+    """
+    n = 120
+    return pl.DataFrame({
+        "open":   [float(i) for i in range(n)],
+        "high":   [float(i) + 1 for i in range(n)],
+        "low":    [float(i) - 1 for i in range(n)],
+        "close":  [float(i) + 0.5 for i in range(n)],
+        "volume": [100 + i for i in range(n)],
+    })
+
+@pytest.fixture
+def small_ohlcv_df():
+    """
+    Returns a polars DataFrame with 5 rows to trigger insufficient data paths.
+    """
+    n = 5
+    return pl.DataFrame({
+        "open":   [float(i) for i in range(n)],
+        "high":   [float(i) + 1 for i in range(n)],
+        "low":    [float(i) - 1 for i in range(n)],
+        "close":  [float(i) + 0.5 for i in range(n)],
+        "volume": [100 + i for i in range(n)],
+    })
\ No newline at end of file
diff --git a/tests/indicators/test_all_indicators.py b/tests/indicators/test_all_indicators.py
new file mode 100644
index 0000000..ecd34ed
--- /dev/null
+++ b/tests/indicators/test_all_indicators.py
@@ -0,0 +1,77 @@
+import pytest
+import polars as pl
+import inspect
+import importlib
+import pkgutil
+from project_x_py.indicators.base import BaseIndicator
+
+def _concrete_indicator_classes():
+    # Recursively discover all non-abstract subclasses of BaseIndicator in project_x_py.indicators.*
+    import project_x_py.indicators
+    seen = set()
+    result = []
+
+    def onclass(cls):
+        if cls in seen:
+            return
+        seen.add(cls)
+        # Must be subclass of BaseIndicator but not the base class itself
+        if not issubclass(cls, BaseIndicator) or cls is BaseIndicator:
+            return
+        # Skip abstract classes (those with any abstractmethods)
+        if getattr(cls, "__abstractmethods__", None):
+            return
+        # Only include classes defined in project_x_py.indicators.*
+        if not cls.__module__.startswith("project_x_py.indicators."):
+            return
+        result.append(cls)
+
+    # Walk all modules in project_x_py.indicators package
+    for finder, name, ispkg in pkgutil.walk_packages(project_x_py.indicators.__path__, project_x_py.indicators.__name__ + "."):
+        try:
+            mod = importlib.import_module(name)
+        except Exception:
+            continue  # If import fails, skip that module
+        for _, obj in inspect.getmembers(mod, inspect.isclass):
+            onclass(obj)
+    # Remove duplicates, sort by class name for determinism
+    return sorted(set(result), key=lambda cls: cls.__name__)
+
+@pytest.mark.parametrize("indicator_cls", _concrete_indicator_classes(), ids=lambda cls: cls.__name__)
+def test_indicator_calculate_adds_new_column(indicator_cls, sample_ohlcv_df):
+    """
+    For every indicator class: instantiate with default ctor, call .calculate() or __call__ on sample data.
+    - No exception is raised.
+    - Result is a polars.DataFrame with same row count.
+    - At least one new column is present.
+    """
+    instance = indicator_cls()
+    input_cols = set(sample_ohlcv_df.columns)
+    # Try __call__ first (uses caching), then fallback to .calculate
+    try:
+        out_df = instance(sample_ohlcv_df)
+    except Exception:
+        out_df = instance.calculate(sample_ohlcv_df)
+
+    assert isinstance(out_df, pl.DataFrame), f"{indicator_cls.__name__} output is not a polars.DataFrame"
+    assert out_df.height == sample_ohlcv_df.height, (
+        f"{indicator_cls.__name__} output row count {out_df.height} != input {sample_ohlcv_df.height}"
+    )
+    new_cols = set(out_df.columns) - input_cols
+    assert new_cols, f"{indicator_cls.__name__} did not add any new columns"
+
+def _get_new_column_names(indicator_cls, input_cols, df):
+    return set(df.columns) - set(input_cols)
+
+@pytest.mark.parametrize("indicator_cls", _concrete_indicator_classes(), ids=lambda cls: cls.__name__)
+def test_indicator_caching_returns_same_object(indicator_cls, sample_ohlcv_df):
+    """
+    Calling the indicator twice with the same df on the same instance should return the exact same DataFrame object (proves internal cache).
+    """
+    instance = indicator_cls()
+    # Use __call__ to trigger cache logic
+    out1 = instance(sample_ohlcv_df)
+    out2 = instance(sample_ohlcv_df)
+    assert out1 is out2, (
+        f"{indicator_cls.__name__} did not return identical object on repeated call (cache broken?)"
+    )
\ No newline at end of file
diff --git a/tests/indicators/test_base_utils.py b/tests/indicators/test_base_utils.py
new file mode 100644
index 0000000..35f5680
--- /dev/null
+++ b/tests/indicators/test_base_utils.py
@@ -0,0 +1,51 @@
+import pytest
+import polars as pl
+
+from project_x_py.indicators.base import (
+    BaseIndicator,
+    safe_division,
+    IndicatorError,
+)
+from project_x_py.indicators.overlap import calculate_sma, SMA
+from project_x_py.indicators.momentum import calculate_rsi
+from project_x_py.indicators.volatility import calculate_atr
+from project_x_py.indicators.volume import calculate_obv
+
+def test_validate_data_missing_column(sample_ohlcv_df):
+    df_missing = sample_ohlcv_df.drop("open")
+    sma = SMA()
+    with pytest.raises(IndicatorError, match="Missing required columns?"):
+        sma.validate_data(df_missing, required_cols=["open", "close"])
+
+def test_validate_data_length_too_short(small_ohlcv_df):
+    sma = SMA()
+    with pytest.raises(IndicatorError, match="at least"):
+        sma.validate_data_length(small_ohlcv_df, min_length=10)
+
+def test_validate_period_negative_or_zero():
+    sma = SMA()
+    for val in [0, -1, -10]:
+        with pytest.raises(IndicatorError, match="period"):
+            sma.validate_period(val)
+
+def test_safe_division_behavior():
+    df = pl.DataFrame({"numerator": [1, 2], "denominator": [0, 2]})
+    out = df.with_columns(
+        result=safe_division(pl.col("numerator"), pl.col("denominator"), default=-1)
+    )
+    # Should be Series [-1, 1]
+    assert out["result"].to_list() == [-1, 1], f"safe_division gave {out['result'].to_list()}"
+
+@pytest.mark.parametrize("func, kwargs, exp_col", [
+    (calculate_sma, {"period": 5}, "sma_5"),
+    (calculate_rsi, {"period": 14}, "rsi_14"),
+    (calculate_atr, {"period": 14}, "atr_14"),
+    (calculate_obv, {}, "obv"),
+])
+def test_convenience_functions_expected_column_and_shape(sample_ohlcv_df, func, kwargs, exp_col):
+    """
+    Convenience functions (calculate_sma etc) add expected columns and preserve row count.
+    """
+    df_func = func(sample_ohlcv_df, **kwargs)
+    assert exp_col in df_func.columns, f"{func.__name__} did not add expected column '{exp_col}'"
+    assert df_func.height == sample_ohlcv_df.height
\ No newline at end of file