diff --git a/toolkit.py b/toolkit.py
index df89ef3b0d36fbd4e4aab7ca34dcaca6322e370d..b1690dfc66b3b64f6788991d2bca3c0561a21804 100755
--- a/toolkit.py
+++ b/toolkit.py
@@ -922,8 +922,8 @@ class ClassificationProject(object):
         significances_train = []
         significances_test = []
         for hist_sig, hist_bkg, rel_errors_sig, rel_errors_bkg, significances, w, y in [
-                (hist_sig_train, hist_bkg_train, rel_errors_bkg_train, rel_errors_sig_train, significances_train, self.w_train, self.y_train),
-                (hist_sig_test, hist_bkg_test, rel_errors_bkg_test, rel_errors_sig_test, significances_test, self.w_test, self.y_test),
+                (hist_sig_train, hist_bkg_train, rel_errors_sig_train, rel_errors_bkg_train, significances_train, self.w_train, self.y_train),
+                (hist_sig_test, hist_bkg_test, rel_errors_sig_test, rel_errors_bkg_test, significances_test, self.w_test, self.y_test),
         ]:
             # factor to rescale due to using only a fraction of events (training and test samples)
             normfactor_sig = (np.sum(self.w_train[self.y_train==1])+np.sum(self.w_test[self.y_test==1]))/np.sum(w[y==1])