I'm getting the following error AttributeError: 'DecisionTreeRegressor' object has no attribute 'n_features_' when running the app in Streamlit for either prediction. I suspect there's a mismatched version somewhere. Do you have a requirements file and the python version you used?

Well, it is pretty self explanatory, Renata was added to league officially and this program can't handle matches that she exists, heres the full log:

2022-02-20 01:22:46.785 Traceback (most recent call last):
  File "C:\Users\hbnat\AppData\Roaming\Python\Python39\site-packages\streamlit\script_runner.py", line 379, in _run_script
    exec(code, module.__dict__)
  File "C:\Users\hbnat\Desktop\ML-Prediction-LoL-master\src\app.py", line 86, in <module>
    data = get_current_match_prediction(summoner_name, region)
  File "C:\Users\hbnat\Desktop\ML-Prediction-LoL-master\.\src\predict.py", line 127, in get_current_match_prediction
    prediction = predict_match(match, region1)
  File "C:\Users\hbnat\Desktop\ML-Prediction-LoL-master\.\src\predict.py", line 54, in predict_match
    mastery_list = api_calls.get_masteries(summonerName, region)["mastery"]
  File "C:\Users\hbnat\Desktop\ML-Prediction-LoL-master\.\src\api_calls\api_calls.py", line 233, in get_masteries
    championId = int(championIds[data[2]])
KeyError: 'Renata Glasc'

I saw this project on reddit, and I think its really great, the data scraping is really good, and we get a nice data set to work with.

However, I think the winrate is doing most of the work here. I threw together a quick test with GBOOST with just using the mean win rates, and I get a 87.91% accuracy by training from the LAN data set, and testing with the NA dataset. If I only used the Winrates Avg, I get 88.20% accuracy on the data set. I did not use k-fold splitting.

Removing any winrate data dropped the accuracy on the NA dataset to 58.67%.

So either champ winrate is all you really need to predict the outcome of a match, or something about winrate data is actually encoding the match outcome, and the GBOOST is learning that instead.

I've attached the test file so you can verify yourself
LoL Winrate Test.zip