diff --git a/_questions/machine-learning-zoomcamp/general/112_91ff5cb6b6_fix-parse-xg-output-xgb-evals-result.md b/_questions/machine-learning-zoomcamp/general/112_91ff5cb6b6_fix-parse-xg-output-xgb-evals-result.md
new file mode 100644
index 0000000..e5910de
--- /dev/null
+++ b/_questions/machine-learning-zoomcamp/general/112_91ff5cb6b6_fix-parse-xg-output-xgb-evals-result.md
@@ -0,0 +1,36 @@
+---
+id: 91ff5cb6b6
+question: 'ValueError: not enough values to unpack when parsing XGBoost output in
+  parse_xg_output; how to fix and switch to evals_result?'
+sort_order: 112
+---
+
+Check the following outputs:
+```python
+print(repr(output.stdout))
+```
+and
+```python
+print(repr(output.stderr))
+```
+If they are empty, this means XGBoost does not print training logs to stdout.
+One solution is to use the training results returned by the Python API:
+```python
+evals_result = {}
+model = xgb.train(
+params=xgb_params,
+dtrain=dtrain,
+num_boost_round=200,
+evals=[(dtrain, 'train'), (dval, 'val')],
+evals_result=evals_result,
+verbose_eval=True
+)
+# Convert results to a DataFrame directly
+df_results = pd.DataFrame({
+'num_iter': range(len(evals_result['train']['auc'])),
+'train_auc': evals_result['train']['auc'],
+'val_auc': evals_result['val']['auc']
+})
+df_results.head()
+```
+This will replace relevant code in parse_xg_output(). Also, the function now does not take any parameters.
\ No newline at end of file