diff --git a/test/train-sets/ref/rcv1_raw_cb_explore.stderr b/test/train-sets/ref/rcv1_raw_cb_explore.stderr
index 966ecc82cc9..d70d35e4b0c 100644
--- a/test/train-sets/ref/rcv1_raw_cb_explore.stderr
+++ b/test/train-sets/ref/rcv1_raw_cb_explore.stderr
@@ -10,20 +10,20 @@ Reading datafile = train-sets/rcv1_raw_cb_small.vw
 num sources = 1
 average  since         example        example  current  current  current
 loss     last          counter         weight    label  predict features
-0.050000 0.050000            1            1.0        1 1:0.975000      280
-0.025000 0.000000            2            2.0        1 2:0.975000      598
-0.012500 0.000000            4            4.0        1 2:0.975000      784
-0.250000 0.487500            8            8.0        1 2:0.975000      850
-0.368750 0.487500           16           16.0        1 2:0.975000      118
-0.245312 0.121875           32           32.0        1 1:0.975000      166
-0.216406 0.187500           64           64.0        2 1:0.975000      340
-0.216406 0.216406          128          128.0        1 2:0.975000      610
-0.300195 0.383984          256          256.0        2 2:0.975000      712
-0.333984 0.367773          512          512.0        2 1:0.975000      424
+1.950000 1.950000            1            1.0        1 1:0.975000      280
+0.975000 0.000000            2            2.0        1 2:0.975000      598
+0.487500 0.000000            4            4.0        1 2:0.975000      784
+0.250000 0.012500            8            8.0        1 2:0.975000      850
+0.618750 0.987500           16           16.0        1 2:0.975000      118
+0.318750 0.018750           32           32.0        1 1:0.975000      166
+0.321875 0.325000           64           64.0        2 1:0.975000      340
+0.244922 0.167969          128          128.0        1 2:0.975000      610
+0.231250 0.217578          256          256.0        2 2:0.975000      712
+0.216992 0.202734          512          512.0        2 1:0.975000      424
 
 finished run
 number of examples = 1000
 weighted example sum = 1000.000000
 weighted label sum = 0.000000
-average loss = 0.362700
+average loss = 0.160650
 total feature number = 440020
diff --git a/vowpalwabbit/cb_explore.cc b/vowpalwabbit/cb_explore.cc
index 6fd15e15c57..16d7f05b68e 100644
--- a/vowpalwabbit/cb_explore.cc
+++ b/vowpalwabbit/cb_explore.cc
@@ -241,7 +241,7 @@ void output_example(vw& all, cb_explore& data, example& ec, CB::label& ld)
 
   if ((c.known_cost = get_observed_cost(ld)) != nullptr)
     for (uint32_t i = 0; i < ec.pred.a_s.size(); i++)
-      loss += get_unbiased_cost(c.known_cost, c.pred_scores, i) * ec.pred.a_s[i].score;
+      loss += get_unbiased_cost(c.known_cost, c.pred_scores, i+1) * ec.pred.a_s[i].score;
 
   all.sd->update(ec.test_only, get_observed_cost(ld) != nullptr, loss, 1.f, ec.num_features);