diff --git a/pypfopt/discrete_allocation.py b/pypfopt/discrete_allocation.py
index f4033e0d..ac64c10d 100644
--- a/pypfopt/discrete_allocation.py
+++ b/pypfopt/discrete_allocation.py
@@ -297,7 +297,7 @@ def lp_portfolio(self, reinvest=False, verbose=False, solver=None):
             print error analysis? Defaults to False.
         solver : str, optional
             the CVXPY solver to use (must support mixed-integer programs).
-            Defaults to "ECOS_BB" if ecos is installed, else None.
+            Defaults to "ECOS_BB" when that solver is installed, else None.
 
         Returns
         -------
@@ -306,7 +306,13 @@ def lp_portfolio(self, reinvest=False, verbose=False, solver=None):
             of funds leftover.
         """
         # todo 1.7.0: remove this defaulting behavior
-        if solver is None and _check_soft_dependencies("ecos", severity="none"):
+        # cvxpy 1.6 dropped ECOS_BB. The ecos package can still be installed,
+        # but selecting a solver cvxpy no longer registers fails the allocation.
+        if (
+            solver is None
+            and _check_soft_dependencies("ecos", severity="none")
+            and "ECOS_BB" in cp.installed_solvers()
+        ):
             solver = "ECOS_BB"
             warn(
                 "The default solver for lp_portfolio will change from ECOS_BB to"
diff --git a/tests/test_discrete_allocation.py b/tests/test_discrete_allocation.py
index f0953186..56f26d17 100644
--- a/tests/test_discrete_allocation.py
+++ b/tests/test_discrete_allocation.py
@@ -1,3 +1,4 @@
+import cvxpy as cp
 from cvxpy.error import SolverError
 import numpy as np
 import pandas as pd
@@ -223,6 +224,71 @@ def test_greedy_portfolio_allocation_short_different_params_reinvest():
     np.testing.assert_almost_equal(long_total + short_total + leftover, 100000)
 
 
+def test_lp_portfolio_skips_ecos_bb_when_cvxpy_removed_it(monkeypatch):
+    requested = {}
+
+    def fake_solve(self, *args, **kwargs):
+        requested["solver"] = kwargs.get("solver")
+        raise SolverError("stop before a real solve")
+
+    monkeypatch.setattr(
+        "pypfopt.discrete_allocation._check_soft_dependencies",
+        lambda *args, **kwargs: True,
+    )
+    monkeypatch.setattr(
+        "pypfopt.discrete_allocation.cp.installed_solvers",
+        lambda: ["OSQP", "SCS"],
+    )
+    monkeypatch.setattr("pypfopt.discrete_allocation.cp.Problem.solve", fake_solve)
+
+    prices = pd.Series({"A": 10.0, "B": 20.0})
+    da = DiscreteAllocation({"A": 0.5, "B": 0.5}, prices, total_portfolio_value=100)
+    with pytest.raises(SolverError):
+        da.lp_portfolio()
+    assert requested["solver"] is None
+
+
+def test_lp_portfolio_still_defaults_to_ecos_bb_when_installed(monkeypatch):
+    requested = {}
+
+    def fake_solve(self, *args, **kwargs):
+        requested["solver"] = kwargs.get("solver")
+        raise SolverError("stop before a real solve")
+
+    monkeypatch.setattr(
+        "pypfopt.discrete_allocation._check_soft_dependencies",
+        lambda *args, **kwargs: True,
+    )
+    monkeypatch.setattr(
+        "pypfopt.discrete_allocation.cp.installed_solvers",
+        lambda: ["ECOS_BB", "OSQP"],
+    )
+    monkeypatch.setattr("pypfopt.discrete_allocation.cp.Problem.solve", fake_solve)
+
+    prices = pd.Series({"A": 10.0, "B": 20.0})
+    da = DiscreteAllocation({"A": 0.5, "B": 0.5}, prices, total_portfolio_value=100)
+    with pytest.warns(FutureWarning, match="ECOS_BB"):
+        with pytest.raises(SolverError):
+            da.lp_portfolio()
+    assert requested["solver"] == "ECOS_BB"
+
+
+@pytest.mark.skipif(
+    "ECOS_BB" in cp.installed_solvers(),
+    reason="ECOS_BB is registered, so the missing-solver path is not active",
+)
+def test_lp_portfolio_solves_when_ecos_bb_is_absent():
+    """cvxpy 1.6 dropped ECOS. A two-asset allocation must still solve."""
+    prices = pd.Series({"A": 10.0, "B": 20.0})
+    da = DiscreteAllocation({"A": 0.5, "B": 0.5}, prices, total_portfolio_value=1000)
+
+    allocation, leftover = da.lp_portfolio()
+
+    assert allocation["A"] * 10 + allocation["B"] * 20 + leftover == pytest.approx(1000)
+    assert allocation["A"] > 0
+    assert allocation["B"] > 0
+
+
 def test_lp_portfolio_allocation():
     ef = setup_efficient_frontier()
     w = ef.max_sharpe(risk_free_rate=0.02)