diff --git a/brambox/boxes/statistics/pr.py b/brambox/boxes/statistics/pr.py
index 1e687906c7ae0c82e604558d1b68788588a1a682..c2302f8e9d74a2803e49a1f8234816a7cd53aa9f 100644
--- a/brambox/boxes/statistics/pr.py
+++ b/brambox/boxes/statistics/pr.py
@@ -5,13 +5,9 @@
 #
 #   Functions for generating PR-curve values and calculating average precision
 #
-
-import math
-from statistics import mean
 import numpy as np
 import scipy.interpolate
-
-from .util import *
+from .util import match_detections
 
 __all__ = ['pr', 'ap']
 
@@ -54,7 +50,7 @@ def ap(precision, recall, num_of_samples=100):
         p = np.array(precision)
         r = np.array(recall)
         p_start = p[np.argmin(r)]
-        samples = np.arange(0., 1., 1.0/num_of_samples)
+        samples = np.arange(0., 1., 1.0 / num_of_samples)
         interpolated = scipy.interpolate.interp1d(r, p, fill_value=(p_start, 0.), bounds_error=False)(samples)
         avg = sum(interpolated) / len(interpolated)
     elif len(precision) > 0 and len(recall) > 0:
diff --git a/brambox/boxes/statistics/util.py b/brambox/boxes/statistics/util.py
index ddf2f631c4af0e06608e4e3e48ab3d16121f2da5..7dea7599ed92a94fe6fe3ebe6e40041e3e39d427 100644
--- a/brambox/boxes/statistics/util.py
+++ b/brambox/boxes/statistics/util.py
@@ -109,7 +109,7 @@ def match_detections(detection_results, ground_truth, overlap_threshold, overlap
                 positives.append((detection.confidence, False))
 
     # sort matches by confidence from high to low
-    positives = sorted(positives, key=lambda d: d[0], reverse=True)
+    positives = sorted(positives, key=lambda d: (d[0], -d[1]), reverse=True)
 
     tps = []
     fps = []