diff --git a/02 - Feature engineering.ipynb b/02 - Feature engineering.ipynb index 857cc4d..49aede8 100644 --- a/02 - Feature engineering.ipynb +++ b/02 - Feature engineering.ipynb @@ -794,7 +794,7 @@ "cell_type": "markdown", "metadata": {}, "source": [ - "Finally, let's exclude unrated games, because people play differently when it \"doesn't count.\" We'll also throw out untimed (\"Correspondence\") games, because we've mentioned that we're especially interested in the effect of time pressure on live win probability, and there is no real time pressure (or clock) when players have unlimited time." + "Finally, let's exclude unrated games, because people play differently when it \"doesn't count.\" We'll also throw out untimed (\"Correspondence\") games, because we've mentioned that we're especially interested in the effect of time pressure on live win probability, and there is no real time pressure (or clock) when players have unlimited time. Last but not least, we'll also remove all games involving non-human players." ] }, { @@ -806,6 +806,8 @@ "model_input_table = model_input_table.filter(\n", " (model_input_table.is_rated)\n", " & (model_input_table.lichess_time_control_type != \"Correspondence\")\n", + " & (model_input_table.white_title.fill_null(\"\") != \"BOT\")\n", + " & (model_input_table.black_title.fill_null(\"\") != \"BOT\")\n", ")" ] },