diff --git a/pyirf/benchmarks/angular_resolution.py b/pyirf/benchmarks/angular_resolution.py
index 9400b03a9..91e0cd094 100644
--- a/pyirf/benchmarks/angular_resolution.py
+++ b/pyirf/benchmarks/angular_resolution.py
@@ -58,4 +58,5 @@ def angular_resolution(
     by_bin = table[valid].group_by(bin_index[valid])
     for bin_idx, group in zip(by_bin.groups.keys, by_bin.groups):
         result[key][bin_idx] = np.nanquantile(group["theta"], ONE_SIGMA_QUANTILE)
+        result["n_events"][bin_idx] = len(group)
     return result
diff --git a/pyirf/benchmarks/tests/test_angular_resolution.py b/pyirf/benchmarks/tests/test_angular_resolution.py
index 629544721..1a1c179f1 100644
--- a/pyirf/benchmarks/tests/test_angular_resolution.py
+++ b/pyirf/benchmarks/tests/test_angular_resolution.py
@@ -45,11 +45,15 @@ def test_angular_resolution(unit):
     events["true_energy"].value[500] = np.nan
     events["true_energy"].value[1500] = np.nan
 
-    ang_res = angular_resolution(
+    table = angular_resolution(
         events,
         [1, 10, 100] * u.TeV,
-    )['angular_resolution']
+    )
 
+    ang_res = table['angular_resolution']
     assert len(ang_res) == 2
     assert u.isclose(ang_res[0], TRUE_RES_1 * u.deg, rtol=0.05)
     assert u.isclose(ang_res[1], TRUE_RES_2 * u.deg, rtol=0.05)
+
+    # one value in each bin is nan, which is ignored
+    np.testing.assert_array_equal(table["n_events"], [998, 998])