diff --git a/docs/BlackLitterman.rst b/docs/BlackLitterman.rst
index 8a6b703d..e96bde9d 100644
--- a/docs/BlackLitterman.rst
+++ b/docs/BlackLitterman.rst
@@ -243,3 +243,8 @@ References
 .. [2] Black, F; Litterman, R. Combining investor views with market equilibrium. The Journal of Fixed Income, 1991.
 .. [3] Walters, Jay, The Factor Tau in the Black-Litterman Model (October 9, 2013). Available at SSRN: https://ssrn.com/abstract=1701467 or http://dx.doi.org/10.2139/ssrn.1701467
 .. [4] Walters J. The Black-Litterman Model in Detail (2014). SSRN Electron J.;(February 2007):1–65. 
+
+When priors and absolute views are total returns, pass ``risk_free_rate`` to
+``BlackLittermanModel``. The market-prior constructor adds this rate;
+``bl_weights()`` subtracts it before recovering implied weights. Leave it at
+zero when the supplied estimates are already excess returns.
diff --git a/pypfopt/black_litterman.py b/pypfopt/black_litterman.py
index ec4bff42..1b208076 100644
--- a/pypfopt/black_litterman.py
+++ b/pypfopt/black_litterman.py
@@ -183,7 +183,9 @@ def __init__(
         market_caps : np.ndarray or pd.Series, optional
             (kwarg) market caps for the assets, required if pi="market"
         risk_free_rate : float, optional
-            (kwarg) risk_free_rate is needed in some methods. Defaults to 0.0.
+            (kwarg) risk-free rate added to a market prior and subtracted when
+            computing implied weights. Defaults to 0.0. If an explicit prior
+            contains total returns, pass its risk-free rate here as well.
         """
         if sys.version_info[1] == 5:  # pragma: no cover
             warnings.warn(
@@ -206,6 +208,7 @@ def __init__(
             self.Q, self.P = self._parse_views(absolute_views)
         else:
             self._set_Q_P(Q, P)
+        self.risk_free_rate = kwargs.get("risk_free_rate", 0.0)
         self._set_risk_aversion(risk_aversion)
         self._set_pi(pi, **kwargs)
         self._set_tau(tau)
@@ -480,7 +483,7 @@ def bl_weights(self, risk_aversion=None):
 
         .. math::
 
-            w = (\delta \Sigma)^{-1} E(R)
+            w = (\delta \Sigma)^{-1} (E(R) - R_f)
 
         Parameters
         ----------
@@ -497,12 +500,12 @@ def bl_weights(self, risk_aversion=None):
 
         self.posterior_rets = self.bl_returns()
         A = risk_aversion * self.cov_matrix
-        b = self.posterior_rets
+        b = self.posterior_rets - self.risk_free_rate
         try:
             weight_solution = np.linalg.solve(A, b)
         except np.linalg.LinAlgError as e:
             if "Singular matrix" in str(e):
-                weight_solution = np.linalg.lstsq(self._A, b, rcond=None)[0]
+                weight_solution = np.linalg.lstsq(A, b, rcond=None)[0]
             else:
                 raise e
         raw_weights = weight_solution
diff --git a/tests/test_black_litterman_regressions.py b/tests/test_black_litterman_regressions.py
new file mode 100644
index 00000000..8f11f20c
--- /dev/null
+++ b/tests/test_black_litterman_regressions.py
@@ -0,0 +1,62 @@
+import numpy as np
+import pandas as pd
+import pytest
+
+from pypfopt.black_litterman import BlackLittermanModel, market_implied_prior_returns
+
+
+@pytest.mark.parametrize("risk_free_rate", [0.0, 0.04])
+@pytest.mark.parametrize("market_prior", [True, False])
+def test_neutral_views_recover_market_weights(risk_free_rate, market_prior):
+    tickers = ["A", "B", "C"]
+    cov = pd.DataFrame(np.diag([0.04, 0.09, 0.16]), index=tickers, columns=tickers)
+    caps = pd.Series([500.0, 300.0, 200.0], index=tickers)
+    prior = market_implied_prior_returns(caps, 2.0, cov, risk_free_rate)
+    model = BlackLittermanModel(
+        cov,
+        pi="market" if market_prior else prior,
+        market_caps=caps,
+        absolute_views=prior,
+        risk_aversion=2.0,
+        risk_free_rate=risk_free_rate,
+    )
+    pd.testing.assert_series_equal(model.bl_returns(), prior)
+    np.testing.assert_allclose(list(model.bl_weights().values()), [0.5, 0.3, 0.2])
+
+
+@pytest.mark.parametrize("views", [{"B": 0.06}, {"A": 0.04, "B": 0.06, "C": 0.04}])
+def test_singular_covariance_uses_asset_space_system(views):
+    # A and C are identical assets. Their total weight is 3/5; B has 2/5.
+    cov = pd.DataFrame(
+        [[0.04, 0.0, 0.04], [0.0, 0.09, 0.0], [0.04, 0.0, 0.04]],
+        index=["A", "B", "C"],
+        columns=["A", "B", "C"],
+    )
+    model = BlackLittermanModel(
+        cov, pi=np.array([0.04, 0.06, 0.04]), absolute_views=views, risk_aversion=2
+    )
+    weights = model.bl_weights()
+    np.testing.assert_allclose(list(weights.values()), [0.3, 0.4, 0.3], atol=1e-12)
+    np.testing.assert_allclose(list(model.optimize().values()), list(weights.values()))
+
+
+def test_duplicate_exact_views_match_single_view():
+    # A singular view-space system is distinct from a singular asset covariance.
+    cov = np.diag([0.04, 0.09, 0.16])
+    prior = np.array([0.05, 0.06, 0.07])
+    single = BlackLittermanModel(
+        cov,
+        pi=prior,
+        P=np.array([[1.0, 0.0, 0.0]]),
+        Q=np.array([0.08]),
+        omega=np.zeros((1, 1)),
+    )
+    duplicate = BlackLittermanModel(
+        cov,
+        pi=prior,
+        P=np.array([[1.0, 0.0, 0.0], [1.0, 0.0, 0.0]]),
+        Q=np.array([0.08, 0.08]),
+        omega=np.zeros((2, 2)),
+    )
+    np.testing.assert_allclose(duplicate.bl_returns(), single.bl_returns())
+    np.testing.assert_allclose(duplicate.bl_cov(), single.bl_cov())