diff --git a/freqtrade/freqai/data_kitchen.py b/freqtrade/freqai/data_kitchen.py
index 23bba3f1a..de4a53a50 100644
--- a/freqtrade/freqai/data_kitchen.py
+++ b/freqtrade/freqai/data_kitchen.py
@@ -241,6 +241,7 @@ class FreqaiDataKitchen:
             self.data["filter_drop_index_training"] = drop_index
 
         else:
+            filtered_df = self.check_pred_labels(filtered_df)
             # we are backtesting so we need to preserve row number to send back to strategy,
             # so now we use do_predict to avoid any prediction based on a NaN
             drop_index = pd.isnull(filtered_df).any(axis=1)
@@ -460,7 +461,7 @@ class FreqaiDataKitchen:
 
         return df
 
-    def check_pred_labels(self, df_predictions: DataFrame) -> None:
+    def check_pred_labels(self, df_predictions: DataFrame) -> DataFrame:
         """
         Check that prediction feature labels match training feature labels.
         :params:
@@ -468,9 +469,15 @@ class FreqaiDataKitchen:
         """
         train_labels = self.data_dictionary["train_features"].columns
         pred_labels = df_predictions.columns
-        if len(train_labels.difference(pred_labels)) != 0:
-            self.data_dictionary["prediction_features"] = df_predictions[train_labels]
-        return
+        num_diffs = len(pred_labels.difference(train_labels))
+        if num_diffs != 0:
+            df_predictions = df_predictions[train_labels]
+            logger.warning(
+                f"Removed {num_diffs} features from prediction features, "
+                f"these were likely considered constant values during most recent training."
+            )
+
+        return df_predictions
 
     def principal_component_analysis(self) -> None:
         """
diff --git a/freqtrade/freqai/freqai_interface.py b/freqtrade/freqai/freqai_interface.py
index 62c814c72..5ac7bc32c 100644
--- a/freqtrade/freqai/freqai_interface.py
+++ b/freqtrade/freqai/freqai_interface.py
@@ -492,8 +492,6 @@ class IFreqaiModel(ABC):
         # ensure user is feeding the correct indicators to the model
         self.check_if_feature_list_matches_strategy(dk)
 
-        dk.check_pred_labels(dk.data_dictionary['prediction_features'])
-
         if ft_params.get('inlier_metric_window', 0):
             dk.compute_inlier_metric(set_='predict')
 
diff --git a/tests/freqai/test_freqai_interface.py b/tests/freqai/test_freqai_interface.py
index 238c0418c..c8444067a 100644
--- a/tests/freqai/test_freqai_interface.py
+++ b/tests/freqai/test_freqai_interface.py
@@ -157,7 +157,7 @@ def test_extract_data_and_train_model_Classifiers(mocker, freqai_conf, model):
         ("CatboostClassifier", 6, "freqai_test_classifier")
     ],
     )
-def test_start_backtesting(mocker, freqai_conf, model, num_files, strat):
+def test_start_backtesting(mocker, freqai_conf, model, num_files, strat, caplog):
     freqai_conf.get("freqai", {}).update({"save_backtest_models": True})
     freqai_conf['runmode'] = RunMode.BACKTEST
     Trade.use_db = False
@@ -182,13 +182,21 @@ def test_start_backtesting(mocker, freqai_conf, model, num_files, strat):
 
     df = freqai.dk.use_strategy_to_populate_indicators(strategy, corr_df, base_df, "LTC/BTC")
     for i in range(5):
-        df[f'constant_{i}'] = i
+        df.loc[:, f'%-constant_{i}'] = i
 
     metadata = {"pair": "LTC/BTC"}
     freqai.start_backtesting(df, metadata, freqai.dk)
     model_folders = [x for x in freqai.dd.full_path.iterdir() if x.is_dir()]
 
     assert len(model_folders) == num_files
+    assert log_has_re(
+        "Removed features ",
+        caplog,
+    )
+    assert log_has_re(
+        "Removed 5 features from prediction features, ",
+        caplog,
+    )
     Backtesting.cleanup()
     shutil.rmtree(Path(freqai.dk.full_path))
 
@@ -210,8 +218,6 @@ def test_start_backtesting_subdaily_backtest_period(mocker, freqai_conf):
     corr_df, base_df = freqai.dd.get_base_and_corr_dataframes(sub_timerange, "LTC/BTC", freqai.dk)
 
     df = freqai.dk.use_strategy_to_populate_indicators(strategy, corr_df, base_df, "LTC/BTC")
-    for i in range(5):
-        df[f'constant_{i}'] = i
 
     metadata = {"pair": "LTC/BTC"}
     freqai.start_backtesting(df, metadata, freqai.dk)
@@ -237,8 +243,6 @@ def test_start_backtesting_from_existing_folder(mocker, freqai_conf, caplog):
     corr_df, base_df = freqai.dd.get_base_and_corr_dataframes(sub_timerange, "LTC/BTC", freqai.dk)
 
     df = freqai.dk.use_strategy_to_populate_indicators(strategy, corr_df, base_df, "LTC/BTC")
-    for i in range(5):
-        df[f'constant_{i}'] = i
 
     metadata = {"pair": "ADA/BTC"}
     freqai.start_backtesting(df, metadata, freqai.dk)
@@ -262,8 +266,7 @@ def test_start_backtesting_from_existing_folder(mocker, freqai_conf, caplog):
     corr_df, base_df = freqai.dd.get_base_and_corr_dataframes(sub_timerange, "LTC/BTC", freqai.dk)
 
     df = freqai.dk.use_strategy_to_populate_indicators(strategy, corr_df, base_df, "LTC/BTC")
-    for i in range(5):
-        df[f'constant_{i}'] = i
+
     freqai.start_backtesting(df, metadata, freqai.dk)
 
     assert log_has_re(
@@ -320,8 +323,7 @@ def test_follow_mode(mocker, freqai_conf):
     freqai.dd.load_all_pair_histories(timerange, freqai.dk)
 
     df = strategy.dp.get_pair_dataframe('ADA/BTC', '5m')
-    for i in range(5):
-        df[f'constant_{i}'] = i
+
     freqai.start_live(df, metadata, strategy, freqai.dk)
 
     assert len(freqai.dk.return_dataframe.index) == 5702