TabNet_resultados_pasados_premier.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 105,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pytorch_tabnet.tab_model import TabNetClassifier\n",
    "\n",
    "import os\n",
    "import torch\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import itertools\n",
    "import matplotlib.pyplot as plt\n",
    " \n",
    "from sklearn.model_selection import KFold, train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler, StandardScaler\n",
    "from sklearn.metrics import accuracy_score, classification_report, confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['Season_End_Year', 'Team', 'Rk', 'MP', 'W', 'D', 'L', 'GF', 'GA', 'GD',\n",
      "       'Pts', 'Notes'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "historical_standings_pl = pd.read_csv('data/Premier/premier-league-standing-1993-2023.csv')\n",
    "\n",
    "print(historical_standings_pl.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['Season_End_Year', 'Team', 'Rk', 'MP', 'W', 'D', 'L', 'GF', 'GA', 'GD',\n",
      "       'Pts'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "#columna Notes no aporta.\n",
    "historical_standings_pl = historical_standings_pl.drop(columns=['Notes'])\n",
    "print(historical_standings_pl.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Arsenal' 'Aston Villa' 'Blackburn' 'Chelsea' 'Coventry City'\n",
      " 'Crystal Palace' 'Everton' 'Ipswich Town' 'Leeds United' 'Liverpool'\n",
      " 'Manchester City' 'Manchester Utd' 'Middlesbrough' 'Norwich City'\n",
      " \"Nott'ham Forest\" 'Oldham Athletic' 'QPR' 'Sheffield Utd'\n",
      " 'Sheffield Weds' 'Southampton' 'Tottenham' 'Wimbledon' 'Newcastle Utd'\n",
      " 'Swindon Town' 'West Ham' 'Leicester City' 'Bolton' 'Derby County'\n",
      " 'Sunderland' 'Barnsley' 'Charlton Ath' 'Bradford City' 'Watford' 'Fulham'\n",
      " 'Birmingham City' 'West Brom' 'Portsmouth' 'Wolves' 'Wigan Athletic'\n",
      " 'Reading' 'Hull City' 'Stoke City' 'Burnley' 'Blackpool' 'Swansea City'\n",
      " 'Cardiff City' 'Bournemouth' 'Brighton' 'Huddersfield' 'Brentford']\n",
      "\n",
      "cantidad:  50\n"
     ]
    }
   ],
   "source": [
    "#Revisamos la cantidad y los nombres de los equipos\n",
    "print(historical_standings_pl['Team'].unique())\n",
    "print(\"\\ncantidad: \", len(historical_standings_pl['Team'].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['Season_End_Year', 'Wk', 'Date', 'Home', 'HomeGoals', 'AwayGoals',\n",
      "       'Away', 'FTR'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "#Cargar archivo de los partidos\n",
    "matches = pd.read_csv('data/Premier/premier-league-matches-1993-2023.csv')\n",
    "\n",
    "print(matches.columns)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "equipos locales:  ['Coventry City' 'Leeds United' 'Sheffield Utd' 'Crystal Palace' 'Arsenal'\n",
      " 'Ipswich Town' 'Everton' 'Southampton' 'Chelsea' \"Nott'ham Forest\"\n",
      " 'Manchester City' 'Blackburn' 'Wimbledon' 'Tottenham' 'Liverpool'\n",
      " 'Aston Villa' 'Oldham Athletic' 'Middlesbrough' 'Norwich City' 'QPR'\n",
      " 'Manchester Utd' 'Sheffield Weds' 'Newcastle Utd' 'West Ham'\n",
      " 'Swindon Town' 'Leicester City' 'Bolton' 'Sunderland' 'Derby County'\n",
      " 'Barnsley' 'Charlton Ath' 'Watford' 'Bradford City' 'Fulham'\n",
      " 'Birmingham City' 'West Brom' 'Portsmouth' 'Wolves' 'Wigan Athletic'\n",
      " 'Reading' 'Hull City' 'Stoke City' 'Burnley' 'Blackpool' 'Swansea City'\n",
      " 'Cardiff City' 'Bournemouth' 'Brighton' 'Huddersfield' 'Brentford']\n",
      "\n",
      "cantidad locales:  50\n",
      "\n",
      "\n",
      "equipos visitantes:  ['Middlesbrough' 'Wimbledon' 'Manchester Utd' 'Blackburn' 'Norwich City'\n",
      " 'Aston Villa' 'Sheffield Weds' 'Tottenham' 'Oldham Athletic' 'Liverpool'\n",
      " 'QPR' 'Arsenal' 'Ipswich Town' 'Coventry City' 'Sheffield Utd'\n",
      " 'Leeds United' 'Crystal Palace' 'Manchester City' 'Chelsea' 'Southampton'\n",
      " 'Everton' \"Nott'ham Forest\" 'Swindon Town' 'West Ham' 'Newcastle Utd'\n",
      " 'Leicester City' 'Bolton' 'Derby County' 'Sunderland' 'Barnsley'\n",
      " 'Charlton Ath' 'Bradford City' 'Watford' 'Fulham' 'West Brom'\n",
      " 'Birmingham City' 'Wolves' 'Portsmouth' 'Wigan Athletic' 'Reading'\n",
      " 'Stoke City' 'Hull City' 'Burnley' 'Blackpool' 'Swansea City'\n",
      " 'Cardiff City' 'Bournemouth' 'Huddersfield' 'Brighton' 'Brentford']\n",
      "\n",
      "cantidad visitantes:  50\n"
     ]
    }
   ],
   "source": [
    "#Revisan los equipos\n",
    "print(\"equipos locales: \", matches['Home'].unique())\n",
    "print(\"\\ncantidad locales: \", len(matches['Home'].unique()))\n",
    "\n",
    "print(\"\\n\\nequipos visitantes: \", matches['Away'].unique())\n",
    "print(\"\\ncantidad visitantes: \", len(matches['Away'].unique()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006\n",
      " 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020\n",
      " 2021 2022]\n",
      "['Arsenal' 'Aston Villa' 'Blackburn' 'Chelsea' 'Coventry City'\n",
      " 'Crystal Palace' 'Everton' 'Ipswich Town' 'Leeds United' 'Liverpool'\n",
      " 'Manchester City' 'Manchester Utd' 'Middlesbrough' 'Norwich City'\n",
      " \"Nott'ham Forest\" 'Oldham Athletic' 'QPR' 'Sheffield Utd'\n",
      " 'Sheffield Weds' 'Southampton' 'Tottenham' 'Wimbledon' 'Newcastle Utd'\n",
      " 'Swindon Town' 'West Ham' 'Leicester City' 'Bolton' 'Derby County'\n",
      " 'Sunderland' 'Barnsley' 'Charlton Ath' 'Bradford City' 'Watford' 'Fulham'\n",
      " 'Birmingham City' 'West Brom' 'Portsmouth' 'Wolves' 'Wigan Athletic'\n",
      " 'Reading' 'Hull City' 'Stoke City' 'Burnley' 'Blackpool' 'Swansea City'\n",
      " 'Cardiff City' 'Bournemouth' 'Brighton' 'Huddersfield' 'Brentford']\n",
      "\n",
      "cantidad:  50\n"
     ]
    }
   ],
   "source": [
    "#Se eliminan los resultados de la temporada 2023 de las historical standings\n",
    "historical_standings_pl = historical_standings_pl[historical_standings_pl['Season_End_Year'] != 2023]\n",
    "\n",
    "#Revisar seasons\n",
    "print(historical_standings_pl['Season_End_Year'].unique())\n",
    "#Se revisan los equipos d nuevo\n",
    "print(historical_standings_pl['Team'].unique())\n",
    "print(\"\\ncantidad: \", len(historical_standings_pl['Team'].unique()))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Se elimina la temporada 1993 de los partidos porque no hay datos de la tabla de posiciones de la temp anterior.\n",
    "matches = matches[matches['Season_End_Year'] != 1993]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Equipos en partidos pero no en standings:  []\n",
      "Equipos en standings pero no en partidos:  []\n"
     ]
    }
   ],
   "source": [
    "#Comparar los equipos de los partidos con los equipos de las standings\n",
    "teams_matches_home = matches['Home'].unique()\n",
    "teams_standings = historical_standings_pl['Team'].unique()\n",
    "\n",
    "teams_not_in_standings = []\n",
    "for team in teams_matches_home:\n",
    "    if team not in teams_standings:\n",
    "        teams_not_in_standings.append(team)\n",
    "\n",
    "print(\"Equipos en partidos pero no en standings: \", teams_not_in_standings)\n",
    "\n",
    "#Comparamos las standings con partidos\n",
    "teams_not_in_matches = []\n",
    "for team in teams_standings:\n",
    "    if team not in teams_matches_home:\n",
    "        teams_not_in_matches.append(team)\n",
    "\n",
    "print(\"Equipos en standings pero no en partidos: \", teams_not_in_matches)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<bound method NDFrame.head of               Team  Rk  MP   W   D   L  GF  GA  GD  Pts  Season_Start_Year\n",
      "0          Arsenal  10  42  15  11  16  40  38   2   56               1994\n",
      "1      Aston Villa   2  42  21  11  10  57  40  17   74               1994\n",
      "2        Blackburn   4  42  20  11  11  68  46  22   71               1994\n",
      "3          Chelsea  11  42  14  14  14  51  54  -3   56               1994\n",
      "4    Coventry City  15  42  13  13  16  52  57  -5   52               1994\n",
      "..             ...  ..  ..  ..  ..  ..  ..  ..  ..  ...                ...\n",
      "601    Southampton  15  38   9  13  16  43  67 -24   40               2023\n",
      "602      Tottenham   4  38  22   5  11  69  40  29   71               2023\n",
      "603        Watford  19  38   6   5  27  34  77 -43   23               2023\n",
      "604       West Ham   7  38  16   8  14  60  51   9   56               2023\n",
      "605         Wolves  10  38  15   6  17  38  43  -5   51               2023\n",
      "\n",
      "[606 rows x 11 columns]>\n",
      "[10  2  4 11 15 20 13 16 17  6  9  1 21  3 22 19  5 14  7 18  8 12]\n"
     ]
    }
   ],
   "source": [
    "#Ahora, se crea una columna de start year en las standings para hacer el merge.\n",
    "historical_standings_pl['Season_Start_Year'] = historical_standings_pl['Season_End_Year'] + 1\n",
    "\n",
    "#Se elimina la columna de season end year\n",
    "historical_standings_pl = historical_standings_pl.drop(columns=['Season_End_Year'])\n",
    "print(historical_standings_pl.head)\n",
    "print(historical_standings_pl['Rk'].unique())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "               Team  Rk  MP   W   D   L  GF   GA  GD  Pts  Season_Start_Year\n",
      "12    Middlesbrough  21  42  11  11  20  54   75 -21   44               1994\n",
      "14  Nott'ham Forest  22  42  10  10  22  41   62 -21   40               1994\n",
      "35  Oldham Athletic  21  42   9  13  20  42   68 -26   40               1995\n",
      "40     Swindon Town  22  42   5  15  22  47  100 -53   30               1995\n",
      "51     Ipswich Town  22  42   7   6  29  36   93 -57   27               1996\n",
      "53   Leicester City  21  42   6  11  25  45   80 -35   29               1996\n"
     ]
    }
   ],
   "source": [
    "#Seleccionar equipos con Rk menor que 20.\n",
    "rk_over_20 = historical_standings_pl[historical_standings_pl['Rk'] > 20]\n",
    "print(rk_over_20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index(['Season_End_Year', 'Wk', 'Date', 'Home', 'HomeGoals', 'AwayGoals',\n",
      "       'Away', 'FTR', 'Position_last_year_home', 'Points_last_year_home',\n",
      "       'Position_last_year_away', 'Points_last_year_away'],\n",
      "      dtype='object')\n",
      "Season_End_Year                      1994\n",
      "Wk                                      1\n",
      "Date                           1993-08-14\n",
      "Home                            Liverpool\n",
      "HomeGoals                               2\n",
      "AwayGoals                               0\n",
      "Away                       Sheffield Weds\n",
      "FTR                                     H\n",
      "Position_last_year_home               6.0\n",
      "Points_last_year_home                59.0\n",
      "Position_last_year_away               7.0\n",
      "Points_last_year_away                59.0\n",
      "Name: 0, dtype: object\n"
     ]
    }
   ],
   "source": [
    "#Solo nos importan las columnas de la posición y los puntos\n",
    "historical_standings_pl = historical_standings_pl[['Team', 'Season_Start_Year', 'Rk', 'Pts']]\n",
    "\n",
    "#merge de los partidos con las standings del local\n",
    "matches = matches.merge(historical_standings_pl, left_on=['Home', 'Season_End_Year'], right_on=['Team', 'Season_Start_Year'], how='left')\n",
    "\n",
    "#Se renombran las columnas last_season\n",
    "matches = matches.rename(columns={'Rk': 'Position_last_year_home', 'Pts': 'Points_last_year_home'})\n",
    "\n",
    "#Se eliminan las columnas de team y season start year\n",
    "matches = matches.drop(columns=['Team', 'Season_Start_Year'])\n",
    "\n",
    "#merge de los partidos con las standings del visitante\n",
    "matches = matches.merge(historical_standings_pl, left_on=['Away', 'Season_End_Year'], right_on=['Team', 'Season_Start_Year'], how='left')\n",
    "\n",
    "#Se renombran las columnas last_season\n",
    "matches = matches.rename(columns={'Rk': 'Position_last_year_away', 'Pts': 'Points_last_year_away'})\n",
    "\n",
    "#Se eliminan las columnas de team y season start year\n",
    "matches = matches.drop(columns=['Team', 'Season_Start_Year'])\n",
    "\n",
    "print(matches.columns)\n",
    "#print el primer partido\n",
    "print(matches.iloc[0])\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Si no tiene posición en la temporada pasada, es porque el equipo **ascendió de categoría**. Se pone la posición 18."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[ 6. 11.  2. 19. 10.  9. 18. 14.  3.  8. 12. 17. 16. 13.  1.  7.  4. 15.\n",
      "  5.]\n",
      "[ 7.  4.  5. 16. 15. 17.  8. 18. 13. 12.  1. 10. 11.  9. 14.  2.  3. 19.\n",
      "  6.]\n"
     ]
    }
   ],
   "source": [
    "matches['Position_last_year_home'] = matches['Position_last_year_home'].fillna(18)\n",
    "print(matches['Position_last_year_home'].unique())\n",
    "\n",
    "matches['Position_last_year_away'] = matches['Position_last_year_away'].fillna(18)\n",
    "print(matches['Position_last_year_away'].unique())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Para el tema de los puntos vacios, se rellenan con el puntaje que tuvo el puesto 17 ya que es el último puesto que no desciende."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Season_End_Year                      1994\n",
      "Wk                                      1\n",
      "Date                           1993-08-14\n",
      "Home                            Liverpool\n",
      "HomeGoals                               2\n",
      "AwayGoals                               0\n",
      "Away                       Sheffield Weds\n",
      "FTR                                     H\n",
      "Position_last_year_home               6.0\n",
      "Points_last_year_home                59.0\n",
      "Position_last_year_away               7.0\n",
      "Points_last_year_away                59.0\n",
      "Name: 0, dtype: object\n",
      "Empty DataFrame\n",
      "Columns: [Season_End_Year, Wk, Date, Home, HomeGoals, AwayGoals, Away, FTR, Position_last_year_home, Points_last_year_home, Position_last_year_away, Points_last_year_away]\n",
      "Index: []\n",
      "Empty DataFrame\n",
      "Columns: [Season_End_Year, Wk, Date, Home, HomeGoals, AwayGoals, Away, FTR, Position_last_year_home, Points_last_year_home, Position_last_year_away, Points_last_year_away]\n",
      "Index: []\n"
     ]
    }
   ],
   "source": [
    "seasons = matches['Season_End_Year'].unique()\n",
    "\n",
    "for season in seasons:\n",
    "    \n",
    "    points_17 = historical_standings_pl[(historical_standings_pl['Season_Start_Year'] == season) & (historical_standings_pl['Rk'] == 17.0)]['Pts']\n",
    "\n",
    "    points_17 = points_17.values[0]\n",
    "    matches.loc[(matches['Season_End_Year'] == season) & (matches['Points_last_year_home'].isnull()), 'Points_last_year_home'] = points_17\n",
    "    matches.loc[(matches['Season_End_Year'] == season) & (matches['Points_last_year_away'].isnull()), 'Points_last_year_away'] = points_17\n",
    "\n",
    "print(matches.iloc[0])\n",
    "print(matches[matches['Points_last_year_home'].isnull()])\n",
    "print(matches[matches['Points_last_year_away'].isnull()])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Se normaliza los nombres de las columnas: Home -> Team_home, Away -> Team_away, HomeGoals -> Goals_home, awayGoals -> Goals_away."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Season_End_Year', 'Wk', 'Date', 'Team_home', 'Goals_home',\n",
       "       'Goals_away', 'Team_away', 'Position_last_year_home',\n",
       "       'Points_last_year_home', 'Position_last_year_away',\n",
       "       'Points_last_year_away'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "matches = matches.rename(columns={'Home': 'Team_home', 'Away': 'Team_away', 'HomeGoals': 'Goals_home', 'AwayGoals': 'Goals_away'})\n",
    "#Se elimina columna FTR\n",
    "matches = matches.drop(columns=['FTR'])\n",
    "matches.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Season_End_Year', 'Wk', 'Date', 'Team_home', 'Team_away', 'Goals_home',\n",
       "       'Goals_away', 'Position_last_year_home', 'Position_last_year_away',\n",
       "       'Points_last_year_home', 'Points_last_year_away'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Reordenar columnas de matches: 'Season_End_Year', 'Wk', 'Date', Team_home, Team_away, Goals_home, Goals_away, Position_last_year_home, Position_last_year_away, Points_last_year_home, Points_last_year_away\n",
    "matches = matches[['Season_End_Year', 'Wk', 'Date', 'Team_home', 'Team_away', 'Goals_home', 'Goals_away', 'Position_last_year_home', 'Position_last_year_away', 'Points_last_year_home', 'Points_last_year_away']]\n",
    "matches.columns"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Se dividen los partidos por temporada."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(462, 9)\n",
      "(462, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n",
      "(380, 9)\n"
     ]
    }
   ],
   "source": [
    "seasons = []\n",
    "seasons_year = matches['Season_End_Year'].unique()\n",
    "\n",
    "for season in seasons_year:\n",
    "    season_data = matches[matches['Season_End_Year'] == season]\n",
    "    season_data = season_data.drop(columns=['Season_End_Year'])\n",
    "    season_data = season_data.drop(columns=['Date'])\n",
    "    seasons.append(season_data)\n",
    "    print(season_data.shape)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Se generan las tablas para cada temporada."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['score_prom', 'Position_last_year', 'Points_last_year']\n",
      "3\n"
     ]
    }
   ],
   "source": [
    "def remove_suffix(item):\n",
    "    return item.replace(\"_home\", \"\").replace(\"_away\", \"\")\n",
    "\n",
    "column = [remove_suffix(item) for item in matches.columns]\n",
    "columnas = []\n",
    "[columnas.append(x) for x in column if x not in columnas]\n",
    "columnas.remove(\"Wk\")\n",
    "columnas.remove(\"Season_End_Year\")\n",
    "columnas.remove(\"Date\")\n",
    "columnas.remove(\"Team\")\n",
    "columnas[0] = \"score_prom\"\n",
    "print(columnas)\n",
    "print(len(columnas))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [],
   "source": [
    "columnas.append(\"wins\")\n",
    "columnas.append(\"ties\")\n",
    "columnas.append(\"losses\")\n",
    "columnas.append(\"total_goals\")\n",
    "columnas.append(\"total_goals_conceded\")\n",
    "columnas.append(\"points\")\n",
    "columnas.append(\"streak\")\n",
    "columnas.append(\"wins_in_last_5_matches\")\n",
    "columnas.append(\"ties_in_last_5_matches\")\n",
    "columnas.append(\"losses_in_last_5_matches\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['score_prom', 'Position_last_year', 'Points_last_year', 'wins', 'ties', 'losses', 'total_goals', 'total_goals_conceded', 'points', 'streak', 'wins_in_last_5_matches', 'ties_in_last_5_matches', 'losses_in_last_5_matches']\n"
     ]
    }
   ],
   "source": [
    "tablas = []\n",
    "\n",
    "for torneo, i in zip(seasons, range(len(seasons))):\n",
    "    equipos = torneo[\"Team_home\"].unique()\n",
    "    equipos2 = torneo[\"Team_away\"].unique()\n",
    "    stats = columnas\n",
    "\n",
    "    tabla = {}\n",
    "\n",
    "    for equipo in equipos:\n",
    "\n",
    "        tabla[equipo] = dict.fromkeys(stats, 0)\n",
    "        tabla[equipo][\"last_5_matches\"] = list()\n",
    "\n",
    "    tablas.append(tabla)\n",
    "\n",
    "print(stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wk</th>\n",
       "      <th>Team_home</th>\n",
       "      <th>Team_away</th>\n",
       "      <th>Goals_home</th>\n",
       "      <th>Goals_away</th>\n",
       "      <th>score_prom_home</th>\n",
       "      <th>score_prom_away</th>\n",
       "      <th>Position_last_year_home</th>\n",
       "      <th>Position_last_year_away</th>\n",
       "      <th>Points_last_year_home</th>\n",
       "      <th>...</th>\n",
       "      <th>points_home</th>\n",
       "      <th>points_away</th>\n",
       "      <th>streak_home</th>\n",
       "      <th>streak_away</th>\n",
       "      <th>wins_in_last_5_matches_home</th>\n",
       "      <th>wins_in_last_5_matches_away</th>\n",
       "      <th>ties_in_last_5_matches_home</th>\n",
       "      <th>ties_in_last_5_matches_away</th>\n",
       "      <th>losses_in_last_5_matches_home</th>\n",
       "      <th>losses_in_last_5_matches_away</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>0 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [Wk, Team_home, Team_away, Goals_home, Goals_away, score_prom_home, score_prom_away, Position_last_year_home, Position_last_year_away, Points_last_year_home, Points_last_year_away, wins_home, wins_away, ties_home, ties_away, losses_home, losses_away, total_goals_home, total_goals_away, total_goals_conceded_home, total_goals_conceded_away, points_home, points_away, streak_home, streak_away, wins_in_last_5_matches_home, wins_in_last_5_matches_away, ties_in_last_5_matches_home, ties_in_last_5_matches_away, losses_in_last_5_matches_home, losses_in_last_5_matches_away]\n",
       "Index: []\n",
       "\n",
       "[0 rows x 31 columns]"
      ]
     },
     "execution_count": 125,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_stats = [\"wins\",\n",
    "\"ties\",\n",
    "\"losses\",\n",
    "\"total_goals\",\n",
    "\"total_goals_conceded\",\n",
    "\"points\",\n",
    "\"streak\",\n",
    "\"wins_in_last_5_matches\",\n",
    "\"ties_in_last_5_matches\",\n",
    "\"losses_in_last_5_matches\"]\n",
    "procesated_matches_columns = list(seasons[0].columns)\n",
    "procesated_matches_columns.insert(5, \"score_prom_away\")\n",
    "procesated_matches_columns.insert(5, \"score_prom_home\")\n",
    "\n",
    "for stat in new_stats:\n",
    "    procesated_matches_columns.append(f\"{stat}_home\")\n",
    "    procesated_matches_columns.append(f\"{stat}_away\")\n",
    "\n",
    "procesated_matches = pd.DataFrame(columns = procesated_matches_columns)\n",
    "\n",
    "procesated_matches.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6.0\n",
      "7.0\n"
     ]
    }
   ],
   "source": [
    "contador = 0\n",
    "for torneo, i in zip(seasons, range(len(seasons))):\n",
    "    torneo.sort_values(by='Wk', inplace=True)\n",
    "    tabla = tablas[i]\n",
    "    for index in torneo.index:\n",
    "    \n",
    "        \"\"\"Escribir info en el nuevo df de partidos con la información hasta el partido la cual esta almacenada en 'tabla'\"\"\"\n",
    "        match = torneo.loc[index]\n",
    "\n",
    "        team_home = match[\"Team_home\"]\n",
    "        team_away = match[\"Team_away\"]\n",
    "\n",
    "        stats_home = tabla[team_home]\n",
    "        stats_away = tabla[team_away]\n",
    "\n",
    "        fila = []\n",
    "        fila.append(match[\"Wk\"])\n",
    "        fila.append(team_home)\n",
    "        fila.append(team_away)\n",
    "        fila.append(match[\"Goals_home\"])\n",
    "        fila.append(match[\"Goals_away\"])\n",
    "\n",
    "        for stat in stats:\n",
    "            fila.append(tabla[team_home][stat])\n",
    "            fila.append(tabla[team_away][stat])\n",
    "\n",
    "        procesated_matches.loc[len(procesated_matches)] = fila\n",
    "        \n",
    "        \n",
    "        \"\"\" Actualizar la tabla con lo que realmente ocurrió en ese partido\"\"\"\n",
    "\n",
    "        matches_played_home = tabla[team_home][\"wins\"] + tabla[team_home][\"ties\"] + tabla[team_home][\"losses\"]+1\n",
    "        matches_played_away = tabla[team_away][\"wins\"] + tabla[team_away][\"ties\"] + tabla[team_away][\"losses\"]+1\n",
    "        \n",
    "        ## stats promediadas:\n",
    "        for stat in stats[:stats.index(\"wins\")]:\n",
    "            if stat == \"score_prom\":\n",
    "                tabla[team_home][stat] = (tabla[team_home][stat]*(matches_played_home-1) + match[\"Goals_home\"])/matches_played_home\n",
    "                tabla[team_away][stat] = (tabla[team_away][stat]*(matches_played_away-1) + match[\"Goals_away\"])/matches_played_away\n",
    "\n",
    "\n",
    "        if contador == 0:\n",
    "            print(match[\"Position_last_year_home\"])\n",
    "            print(match[\"Position_last_year_away\"])\n",
    "        contador = 1\n",
    "        ## stats acumuladas\n",
    "        \n",
    "        empate = False\n",
    "        if match[\"Goals_home\"] > match[\"Goals_away\"]:\n",
    "\n",
    "            #victorias / derrotas\n",
    "            tabla[team_home][\"wins\"] += 1\n",
    "            tabla[team_away][\"losses\"] += 1\n",
    "\n",
    "            #goles a favor\n",
    "            tabla[team_home][\"total_goals\"] += match[\"Goals_home\"]\n",
    "            tabla[team_away][\"total_goals\"] += match[\"Goals_away\"]\n",
    "\n",
    "            #goles en contra\n",
    "            tabla[team_home][\"total_goals_conceded\"] += match[\"Goals_away\"]\n",
    "            tabla[team_away][\"total_goals_conceded\"] += match[\"Goals_home\"]\n",
    "\n",
    "            #puntos\n",
    "            tabla[team_home][\"points\"] += 3\n",
    "\n",
    "            # racha\n",
    "            tabla[team_home][\"streak\"] += 1\n",
    "            tabla[team_away][\"streak\"] = 0\n",
    "\n",
    "            # ultimos 5 partidos\n",
    "            if len(tabla[team_home][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_home][\"last_5_matches\"].pop(0)\n",
    "            if len(tabla[team_away][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_away][\"last_5_matches\"].pop(0)\n",
    "\n",
    "            tabla[team_home][\"last_5_matches\"].append(\"W\")\n",
    "            tabla[team_away][\"last_5_matches\"].append(\"L\")\n",
    "\n",
    "\n",
    "        elif match[\"Goals_home\"] < match[\"Goals_away\"]:\n",
    "\n",
    "            #victorias / derrotas\n",
    "            tabla[team_away][\"wins\"] += 1\n",
    "            tabla[team_home][\"losses\"] += 1\n",
    "\n",
    "            #goles a favor\n",
    "            tabla[team_away][\"total_goals\"] += match[\"Goals_away\"]\n",
    "            tabla[team_home][\"total_goals\"] += match[\"Goals_home\"]\n",
    "\n",
    "            #goles en contra\n",
    "            tabla[team_away][\"total_goals_conceded\"] += match[\"Goals_home\"]\n",
    "            tabla[team_home][\"total_goals_conceded\"] += match[\"Goals_away\"]\n",
    "\n",
    "            #puntos\n",
    "            tabla[team_away][\"points\"] += 3\n",
    "\n",
    "            # racha\n",
    "            tabla[team_away][\"streak\"] += 1\n",
    "            tabla[team_home][\"streak\"] = 0\n",
    "\n",
    "            # ultimos 5 partidos\n",
    "            if len(tabla[team_home][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_home][\"last_5_matches\"].pop(0)\n",
    "            if len(tabla[team_away][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_away][\"last_5_matches\"].pop(0)\n",
    "\n",
    "            tabla[team_away][\"last_5_matches\"].append(\"W\")\n",
    "            tabla[team_home][\"last_5_matches\"].append(\"L\")\n",
    "\n",
    "        else:\n",
    "            #empate\n",
    "            tabla[team_home][\"ties\"] += 1\n",
    "            tabla[team_away][\"ties\"] += 1\n",
    "\n",
    "            #goles a favor\n",
    "            tabla[team_home][\"total_goals\"] += match[\"Goals_home\"]\n",
    "            tabla[team_away][\"total_goals\"] += match[\"Goals_away\"]\n",
    "\n",
    "            #goles en contra\n",
    "            tabla[team_home][\"total_goals_conceded\"] += match[\"Goals_away\"]\n",
    "            tabla[team_away][\"total_goals_conceded\"] += match[\"Goals_home\"]\n",
    "\n",
    "            #puntos\n",
    "            tabla[team_home][\"points\"] += 1\n",
    "            tabla[team_away][\"points\"] += 1\n",
    "\n",
    "            # racha\n",
    "            tabla[team_away][\"streak\"] += 1\n",
    "            tabla[team_home][\"streak\"] += 1\n",
    "\n",
    "            # ultimos 5 partidos\n",
    "            if len(tabla[team_home][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_home][\"last_5_matches\"].pop(0)\n",
    "            if len(tabla[team_away][\"last_5_matches\"]) == 5:\n",
    "                tabla[team_away][\"last_5_matches\"].pop(0)\n",
    "\n",
    "            tabla[team_away][\"last_5_matches\"].append(\"T\")\n",
    "            tabla[team_home][\"last_5_matches\"].append(\"T\")\n",
    "\n",
    "        tabla[team_home][\"wins_in_last_5_matches\"] = tabla[team_home][\"last_5_matches\"].count(\"W\")\n",
    "        tabla[team_home][\"ties_in_last_5_matches\"] = tabla[team_home][\"last_5_matches\"].count(\"T\")\n",
    "        tabla[team_home][\"losses_in_last_5_matches\"] = tabla[team_home][\"last_5_matches\"].count(\"L\")\n",
    "\n",
    "        tabla[team_away][\"wins_in_last_5_matches\"] = tabla[team_away][\"last_5_matches\"].count(\"W\")\n",
    "        tabla[team_away][\"ties_in_last_5_matches\"] = tabla[team_away][\"last_5_matches\"].count(\"T\")\n",
    "        tabla[team_away][\"losses_in_last_5_matches\"] = tabla[team_away][\"last_5_matches\"].count(\"L\")\n",
    "\n",
    "        tabla[team_home][\"Points_last_year\"] = match[\"Points_last_year_home\"]\n",
    "        tabla[team_away][\"Points_last_year\"] = match[\"Points_last_year_away\"]\n",
    "\n",
    "        tabla[team_home][\"Position_last_year\"] = match[\"Position_last_year_home\"]\n",
    "        tabla[team_away][\"Position_last_year\"] = match[\"Position_last_year_away\"]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Wk</th>\n",
       "      <th>Team_home</th>\n",
       "      <th>Team_away</th>\n",
       "      <th>Goals_home</th>\n",
       "      <th>Goals_away</th>\n",
       "      <th>score_prom_home</th>\n",
       "      <th>score_prom_away</th>\n",
       "      <th>Position_last_year_home</th>\n",
       "      <th>Position_last_year_away</th>\n",
       "      <th>Points_last_year_home</th>\n",
       "      <th>...</th>\n",
       "      <th>points_home</th>\n",
       "      <th>points_away</th>\n",
       "      <th>streak_home</th>\n",
       "      <th>streak_away</th>\n",
       "      <th>wins_in_last_5_matches_home</th>\n",
       "      <th>wins_in_last_5_matches_away</th>\n",
       "      <th>ties_in_last_5_matches_home</th>\n",
       "      <th>ties_in_last_5_matches_away</th>\n",
       "      <th>losses_in_last_5_matches_home</th>\n",
       "      <th>losses_in_last_5_matches_away</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>11559</th>\n",
       "      <td>38</td>\n",
       "      <td>Southampton</td>\n",
       "      <td>Liverpool</td>\n",
       "      <td>4</td>\n",
       "      <td>4</td>\n",
       "      <td>0.864865</td>\n",
       "      <td>1.918919</td>\n",
       "      <td>15.0</td>\n",
       "      <td>2.0</td>\n",
       "      <td>40.0</td>\n",
       "      <td>...</td>\n",
       "      <td>24</td>\n",
       "      <td>66</td>\n",
       "      <td>0</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11560</th>\n",
       "      <td>38</td>\n",
       "      <td>Everton</td>\n",
       "      <td>Bournemouth</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.891892</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>16.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>39.0</td>\n",
       "      <td>...</td>\n",
       "      <td>33</td>\n",
       "      <td>39</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11561</th>\n",
       "      <td>38</td>\n",
       "      <td>Leicester City</td>\n",
       "      <td>West Ham</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1.324324</td>\n",
       "      <td>1.108108</td>\n",
       "      <td>8.0</td>\n",
       "      <td>7.0</td>\n",
       "      <td>52.0</td>\n",
       "      <td>...</td>\n",
       "      <td>31</td>\n",
       "      <td>40</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11562</th>\n",
       "      <td>38</td>\n",
       "      <td>Aston Villa</td>\n",
       "      <td>Brighton</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1.324324</td>\n",
       "      <td>1.918919</td>\n",
       "      <td>14.0</td>\n",
       "      <td>9.0</td>\n",
       "      <td>45.0</td>\n",
       "      <td>...</td>\n",
       "      <td>58</td>\n",
       "      <td>62</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11563</th>\n",
       "      <td>38</td>\n",
       "      <td>Brentford</td>\n",
       "      <td>Manchester City</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1.540541</td>\n",
       "      <td>2.540541</td>\n",
       "      <td>13.0</td>\n",
       "      <td>1.0</td>\n",
       "      <td>46.0</td>\n",
       "      <td>...</td>\n",
       "      <td>56</td>\n",
       "      <td>89</td>\n",
       "      <td>2</td>\n",
       "      <td>15</td>\n",
       "      <td>4</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 31 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       Wk       Team_home        Team_away  Goals_home  Goals_away  \\\n",
       "11559  38     Southampton        Liverpool           4           4   \n",
       "11560  38         Everton      Bournemouth           1           0   \n",
       "11561  38  Leicester City         West Ham           2           1   \n",
       "11562  38     Aston Villa         Brighton           2           1   \n",
       "11563  38       Brentford  Manchester City           1           0   \n",
       "\n",
       "       score_prom_home  score_prom_away  Position_last_year_home  \\\n",
       "11559         0.864865         1.918919                     15.0   \n",
       "11560         0.891892         1.000000                     16.0   \n",
       "11561         1.324324         1.108108                      8.0   \n",
       "11562         1.324324         1.918919                     14.0   \n",
       "11563         1.540541         2.540541                     13.0   \n",
       "\n",
       "       Position_last_year_away  Points_last_year_home  ...  points_home  \\\n",
       "11559                      2.0                   40.0  ...           24   \n",
       "11560                     18.0                   39.0  ...           33   \n",
       "11561                      7.0                   52.0  ...           31   \n",
       "11562                      9.0                   45.0  ...           58   \n",
       "11563                      1.0                   46.0  ...           56   \n",
       "\n",
       "       points_away  streak_home  streak_away  wins_in_last_5_matches_home  \\\n",
       "11559           66            0            8                            0   \n",
       "11560           39            1            0                            1   \n",
       "11561           40            1            1                            0   \n",
       "11562           62            2            2                            2   \n",
       "11563           89            2           15                            4   \n",
       "\n",
       "       wins_in_last_5_matches_away  ties_in_last_5_matches_home  \\\n",
       "11559                            4                            0   \n",
       "11560                            2                            2   \n",
       "11561                            2                            3   \n",
       "11562                            3                            1   \n",
       "11563                            5                            0   \n",
       "\n",
       "       ties_in_last_5_matches_away  losses_in_last_5_matches_home  \\\n",
       "11559                            1                              5   \n",
       "11560                            0                              2   \n",
       "11561                            0                              2   \n",
       "11562                            0                              2   \n",
       "11563                            0                              1   \n",
       "\n",
       "       losses_in_last_5_matches_away  \n",
       "11559                              0  \n",
       "11560                              3  \n",
       "11561                              3  \n",
       "11562                              2  \n",
       "11563                              0  \n",
       "\n",
       "[5 rows x 31 columns]"
      ]
     },
     "execution_count": 127,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "procesated_matches.tail()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "using mps\n"
     ]
    }
   ],
   "source": [
    "#Usamos el GPU si esta disponible, mps porque es apple silicon.\n",
    "device = 'mps' if torch.backends.mps.is_available() else 'cpu'\n",
    "print(f\"using {device}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "       Wk       Team_home        Team_away  Goals_home  Goals_away  \\\n",
      "11559  38     Southampton        Liverpool           4           4   \n",
      "11560  38         Everton      Bournemouth           1           0   \n",
      "11561  38  Leicester City         West Ham           2           1   \n",
      "11562  38     Aston Villa         Brighton           2           1   \n",
      "11563  38       Brentford  Manchester City           1           0   \n",
      "\n",
      "       score_prom_home  score_prom_away  Position_last_year_home  \\\n",
      "11559         0.864865         1.918919                     15.0   \n",
      "11560         0.891892         1.000000                     16.0   \n",
      "11561         1.324324         1.108108                      8.0   \n",
      "11562         1.324324         1.918919                     14.0   \n",
      "11563         1.540541         2.540541                     13.0   \n",
      "\n",
      "       Position_last_year_away  Points_last_year_home  ...  points_away  \\\n",
      "11559                      2.0                   40.0  ...           66   \n",
      "11560                     18.0                   39.0  ...           39   \n",
      "11561                      7.0                   52.0  ...           40   \n",
      "11562                      9.0                   45.0  ...           62   \n",
      "11563                      1.0                   46.0  ...           89   \n",
      "\n",
      "       streak_home  streak_away  wins_in_last_5_matches_home  \\\n",
      "11559            0            8                            0   \n",
      "11560            1            0                            1   \n",
      "11561            1            1                            0   \n",
      "11562            2            2                            2   \n",
      "11563            2           15                            4   \n",
      "\n",
      "       wins_in_last_5_matches_away  ties_in_last_5_matches_home  \\\n",
      "11559                            4                            0   \n",
      "11560                            2                            2   \n",
      "11561                            2                            3   \n",
      "11562                            3                            1   \n",
      "11563                            5                            0   \n",
      "\n",
      "       ties_in_last_5_matches_away  losses_in_last_5_matches_home  \\\n",
      "11559                            1                              5   \n",
      "11560                            0                              2   \n",
      "11561                            0                              2   \n",
      "11562                            0                              2   \n",
      "11563                            0                              1   \n",
      "\n",
      "       losses_in_last_5_matches_away  result  \n",
      "11559                              0       0  \n",
      "11560                              3       1  \n",
      "11561                              3       1  \n",
      "11562                              2       1  \n",
      "11563                              0       1  \n",
      "\n",
      "[5 rows x 32 columns]\n"
     ]
    }
   ],
   "source": [
    "matches = procesated_matches\n",
    "\n",
    "#Aqui agrega la columna de resultados\n",
    "\n",
    "results = []\n",
    "\n",
    "for index in matches.index:\n",
    "\n",
    "    match = matches.loc[index]\n",
    "    if match[\"Goals_home\"] > match[\"Goals_away\"]:\n",
    "        results.append(1)\n",
    "    elif match[\"Goals_home\"] < match[\"Goals_away\"]:\n",
    "        results.append(2)\n",
    "    else:\n",
    "        results.append(0)\n",
    "\n",
    "matches[\"result\"] = results\n",
    "print(matches.tail())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Resultado 0, hay: 2967 partidos\n",
      "Resultado 1, hay: 5305 partidos\n",
      "Resultado 2, hay: 3292 partidos\n",
      "azar, siempre al local: 0.4587512971290211\n"
     ]
    }
   ],
   "source": [
    "#Cantidad de partidos por resultado (local, empate, visita)\n",
    "\n",
    "for i in range(0,3):\n",
    "    cantidad = matches[matches[\"result\"]==i].shape[0]\n",
    "    print(f\"Resultado {i}, hay: {cantidad} partidos\")\n",
    "num = 5305/(2967+3292+5305)\n",
    "print(\"azar, siempre al local:\", num)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(       score_prom_home  score_prom_away  Position_last_year_home  \\\n",
       " 11179         1.324324         1.540541                     14.0   \n",
       " 11180         0.621622         1.729730                     18.0   \n",
       " 11181         0.891892         1.135135                     17.0   \n",
       " 11182         2.459459         1.000000                      3.0   \n",
       " 11183         2.000000         0.891892                      4.0   \n",
       " \n",
       "        Position_last_year_away  Points_last_year_home  Points_last_year_away  \\\n",
       " 11179                      2.0                   44.0                   74.0   \n",
       " 11180                      7.0                   39.0                   62.0   \n",
       " 11181                     12.0                   39.0                   45.0   \n",
       " 11182                     13.0                   69.0                   45.0   \n",
       " 11183                     18.0                   67.0                   39.0   \n",
       " \n",
       "        wins_home  wins_away  ties_home  ties_away  ...  points_home  \\\n",
       " 11179         10         16         15         10  ...           45   \n",
       " 11180          5         21          7          5  ...           22   \n",
       " 11181          7         12         14         10  ...           35   \n",
       " 11182         27         15          8          6  ...           89   \n",
       " 11183         20          6         11          5  ...           71   \n",
       " \n",
       "        points_away  streak_home  streak_away  wins_in_last_5_matches_home  \\\n",
       " 11179           58            4            1                            2   \n",
       " 11180           68            1            4                            0   \n",
       " 11181           46            0            1                            2   \n",
       " 11182           51           17            2                            4   \n",
       " 11183           23            2            0                            2   \n",
       " \n",
       "        wins_in_last_5_matches_away  ties_in_last_5_matches_home  \\\n",
       " 11179                            2                            2   \n",
       " 11180                            2                            1   \n",
       " 11181                            3                            1   \n",
       " 11182                            0                            1   \n",
       " 11183                            0                            2   \n",
       " \n",
       "        ties_in_last_5_matches_away  losses_in_last_5_matches_home  \\\n",
       " 11179                            1                              1   \n",
       " 11180                            2                              4   \n",
       " 11181                            0                              2   \n",
       " 11182                            2                              0   \n",
       " 11183                            0                              1   \n",
       " \n",
       "        losses_in_last_5_matches_away  \n",
       " 11179                              2  \n",
       " 11180                              1  \n",
       " 11181                              2  \n",
       " 11182                              3  \n",
       " 11183                              5  \n",
       " \n",
       " [5 rows x 26 columns],\n",
       "        score_prom_home  score_prom_away  Position_last_year_home  \\\n",
       " 11559         0.864865         1.918919                     15.0   \n",
       " 11560         0.891892         1.000000                     16.0   \n",
       " 11561         1.324324         1.108108                      8.0   \n",
       " 11562         1.324324         1.918919                     14.0   \n",
       " 11563         1.540541         2.540541                     13.0   \n",
       " \n",
       "        Position_last_year_away  Points_last_year_home  Points_last_year_away  \\\n",
       " 11559                      2.0                   40.0                   92.0   \n",
       " 11560                     18.0                   39.0                   38.0   \n",
       " 11561                      7.0                   52.0                   56.0   \n",
       " 11562                      9.0                   45.0                   51.0   \n",
       " 11563                      1.0                   46.0                   93.0   \n",
       " \n",
       "        wins_home  wins_away  ties_home  ties_away  ...  points_home  \\\n",
       " 11559          6         19          6          9  ...           24   \n",
       " 11560          7         11         12          6  ...           33   \n",
       " 11561          8         11          7          7  ...           31   \n",
       " 11562         17         18          7          8  ...           58   \n",
       " 11563         14         28         14          5  ...           56   \n",
       " \n",
       "        points_away  streak_home  streak_away  wins_in_last_5_matches_home  \\\n",
       " 11559           66            0            8                            0   \n",
       " 11560           39            1            0                            1   \n",
       " 11561           40            1            1                            0   \n",
       " 11562           62            2            2                            2   \n",
       " 11563           89            2           15                            4   \n",
       " \n",
       "        wins_in_last_5_matches_away  ties_in_last_5_matches_home  \\\n",
       " 11559                            4                            0   \n",
       " 11560                            2                            2   \n",
       " 11561                            2                            3   \n",
       " 11562                            3                            1   \n",
       " 11563                            5                            0   \n",
       " \n",
       "        ties_in_last_5_matches_away  losses_in_last_5_matches_home  \\\n",
       " 11559                            1                              5   \n",
       " 11560                            0                              2   \n",
       " 11561                            0                              2   \n",
       " 11562                            0                              2   \n",
       " 11563                            0                              1   \n",
       " \n",
       "        losses_in_last_5_matches_away  \n",
       " 11559                              0  \n",
       " 11560                              3  \n",
       " 11561                              3  \n",
       " 11562                              2  \n",
       " 11563                              0  \n",
       " \n",
       " [5 rows x 26 columns])"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Ahora, vamos a separar el último torneo\n",
    "\n",
    "cant_partidos_ultima_temp = len(seasons[-1])\n",
    "\n",
    "matches_train = matches.iloc[:-cant_partidos_ultima_temp]\n",
    "matches_test = matches.iloc[-cant_partidos_ultima_temp:]\n",
    "\n",
    "result_train = matches_train[\"result\"]\n",
    "result_test = matches_test[\"result\"]\n",
    "matches_train = matches_train.drop(columns = [\"Team_home\", \"Team_away\", \"result\", \"Goals_home\", \"Goals_away\", \"Wk\"])\n",
    "matches_test = matches_test.drop(columns = [\"Team_home\", \"Team_away\", \"result\", \"Goals_home\", \"Goals_away\", \"Wk\"])\n",
    "\n",
    "matches_train.tail(), matches_test.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = matches_train.to_numpy()\n",
    "scaler = StandardScaler()\n",
    "X_train = scaler.fit_transform(X)\n",
    "y_train = result_train\n",
    "\n",
    "X_test = scaler.transform(matches_test.to_numpy())\n",
    "y_test = result_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/gustavofernandez/Desktop/iPre-resultados-deportivos/env/lib/python3.8/site-packages/pytorch_tabnet/abstract_model.py:82: UserWarning: Device used : cpu\n",
      "  warnings.warn(f\"Device used : {self.device}\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch 0  | loss: 1.26721 | train_accuracy: 0.44599 | valid_accuracy: 0.47895 |  0:00:04s\n",
      "epoch 1  | loss: 1.06558 | train_accuracy: 0.45771 | valid_accuracy: 0.45789 |  0:00:08s\n",
      "epoch 2  | loss: 1.04599 | train_accuracy: 0.4738  | valid_accuracy: 0.46842 |  0:00:12s\n",
      "epoch 3  | loss: 1.03572 | train_accuracy: 0.48364 | valid_accuracy: 0.51053 |  0:00:17s\n",
      "epoch 4  | loss: 1.03101 | train_accuracy: 0.4932  | valid_accuracy: 0.50263 |  0:00:21s\n",
      "epoch 5  | loss: 1.02194 | train_accuracy: 0.49946 | valid_accuracy: 0.51053 |  0:00:25s\n",
      "epoch 6  | loss: 1.02159 | train_accuracy: 0.50277 | valid_accuracy: 0.52368 |  0:00:30s\n",
      "epoch 7  | loss: 1.01591 | train_accuracy: 0.5076  | valid_accuracy: 0.52105 |  0:00:34s\n",
      "epoch 8  | loss: 1.0147  | train_accuracy: 0.50653 | valid_accuracy: 0.51842 |  0:00:39s\n",
      "epoch 9  | loss: 1.01318 | train_accuracy: 0.50975 | valid_accuracy: 0.50263 |  0:00:43s\n",
      "epoch 10 | loss: 1.0108  | train_accuracy: 0.50948 | valid_accuracy: 0.51316 |  0:00:47s\n",
      "epoch 11 | loss: 1.00698 | train_accuracy: 0.51046 | valid_accuracy: 0.51053 |  0:00:52s\n",
      "epoch 12 | loss: 1.00613 | train_accuracy: 0.50984 | valid_accuracy: 0.51579 |  0:00:56s\n",
      "epoch 13 | loss: 1.00988 | train_accuracy: 0.51144 | valid_accuracy: 0.51316 |  0:01:00s\n",
      "epoch 14 | loss: 1.00657 | train_accuracy: 0.51109 | valid_accuracy: 0.50789 |  0:01:04s\n",
      "epoch 15 | loss: 1.00457 | train_accuracy: 0.51377 | valid_accuracy: 0.52368 |  0:01:08s\n",
      "epoch 16 | loss: 1.0053  | train_accuracy: 0.51225 | valid_accuracy: 0.51842 |  0:01:13s\n",
      "epoch 17 | loss: 1.00531 | train_accuracy: 0.51457 | valid_accuracy: 0.52632 |  0:01:17s\n",
      "epoch 18 | loss: 1.00379 | train_accuracy: 0.51413 | valid_accuracy: 0.52105 |  0:01:21s\n",
      "epoch 19 | loss: 1.00491 | train_accuracy: 0.51484 | valid_accuracy: 0.53421 |  0:01:25s\n",
      "epoch 20 | loss: 1.00338 | train_accuracy: 0.51502 | valid_accuracy: 0.52368 |  0:01:29s\n",
      "epoch 21 | loss: 1.00207 | train_accuracy: 0.51359 | valid_accuracy: 0.53421 |  0:01:34s\n",
      "epoch 22 | loss: 1.00249 | train_accuracy: 0.51627 | valid_accuracy: 0.52105 |  0:01:38s\n",
      "epoch 23 | loss: 1.00007 | train_accuracy: 0.51493 | valid_accuracy: 0.53684 |  0:01:43s\n",
      "epoch 24 | loss: 1.00137 | train_accuracy: 0.51681 | valid_accuracy: 0.53684 |  0:01:47s\n",
      "epoch 25 | loss: 1.00082 | train_accuracy: 0.51395 | valid_accuracy: 0.53158 |  0:01:51s\n",
      "epoch 26 | loss: 0.9998  | train_accuracy: 0.5152  | valid_accuracy: 0.52368 |  0:01:55s\n",
      "epoch 27 | loss: 0.99993 | train_accuracy: 0.5177  | valid_accuracy: 0.52895 |  0:02:00s\n",
      "epoch 28 | loss: 0.99925 | train_accuracy: 0.51377 | valid_accuracy: 0.54211 |  0:02:04s\n",
      "epoch 29 | loss: 0.99916 | train_accuracy: 0.51726 | valid_accuracy: 0.53421 |  0:02:08s\n",
      "epoch 30 | loss: 0.99577 | train_accuracy: 0.5169  | valid_accuracy: 0.53684 |  0:02:12s\n",
      "epoch 31 | loss: 0.99959 | train_accuracy: 0.51663 | valid_accuracy: 0.54474 |  0:02:17s\n",
      "epoch 32 | loss: 0.99595 | train_accuracy: 0.51529 | valid_accuracy: 0.53684 |  0:02:21s\n",
      "epoch 33 | loss: 0.99744 | train_accuracy: 0.51815 | valid_accuracy: 0.53421 |  0:02:25s\n",
      "epoch 34 | loss: 0.99731 | train_accuracy: 0.51761 | valid_accuracy: 0.53158 |  0:02:29s\n",
      "epoch 35 | loss: 0.99644 | train_accuracy: 0.51699 | valid_accuracy: 0.52632 |  0:02:34s\n",
      "epoch 36 | loss: 0.99572 | train_accuracy: 0.51949 | valid_accuracy: 0.54211 |  0:02:38s\n",
      "epoch 37 | loss: 0.99697 | train_accuracy: 0.51761 | valid_accuracy: 0.53684 |  0:02:42s\n",
      "epoch 38 | loss: 0.99878 | train_accuracy: 0.5203  | valid_accuracy: 0.52895 |  0:02:47s\n",
      "epoch 39 | loss: 0.99656 | train_accuracy: 0.51663 | valid_accuracy: 0.54211 |  0:02:51s\n",
      "epoch 40 | loss: 0.99492 | train_accuracy: 0.51842 | valid_accuracy: 0.53947 |  0:02:56s\n",
      "epoch 41 | loss: 0.99616 | train_accuracy: 0.51753 | valid_accuracy: 0.53684 |  0:03:00s\n",
      "epoch 42 | loss: 0.99765 | train_accuracy: 0.52092 | valid_accuracy: 0.55    |  0:03:04s\n",
      "epoch 43 | loss: 0.99642 | train_accuracy: 0.51905 | valid_accuracy: 0.52368 |  0:03:09s\n",
      "epoch 44 | loss: 0.9956  | train_accuracy: 0.51833 | valid_accuracy: 0.53421 |  0:03:13s\n",
      "epoch 45 | loss: 0.99495 | train_accuracy: 0.51708 | valid_accuracy: 0.53684 |  0:03:17s\n",
      "epoch 46 | loss: 0.99372 | train_accuracy: 0.52003 | valid_accuracy: 0.54211 |  0:03:21s\n",
      "epoch 47 | loss: 0.99338 | train_accuracy: 0.51994 | valid_accuracy: 0.54474 |  0:03:26s\n",
      "epoch 48 | loss: 0.99088 | train_accuracy: 0.52164 | valid_accuracy: 0.53947 |  0:03:30s\n",
      "epoch 49 | loss: 0.99428 | train_accuracy: 0.52235 | valid_accuracy: 0.54211 |  0:03:34s\n",
      "epoch 50 | loss: 0.99421 | train_accuracy: 0.52262 | valid_accuracy: 0.53684 |  0:03:39s\n",
      "epoch 51 | loss: 0.99231 | train_accuracy: 0.52092 | valid_accuracy: 0.54474 |  0:03:43s\n",
      "epoch 52 | loss: 0.99169 | train_accuracy: 0.52262 | valid_accuracy: 0.52895 |  0:03:48s\n",
      "epoch 53 | loss: 0.99189 | train_accuracy: 0.522   | valid_accuracy: 0.53421 |  0:03:52s\n",
      "epoch 54 | loss: 0.99272 | train_accuracy: 0.522   | valid_accuracy: 0.55263 |  0:03:57s\n",
      "epoch 55 | loss: 0.98985 | train_accuracy: 0.52298 | valid_accuracy: 0.54737 |  0:04:01s\n",
      "epoch 56 | loss: 0.99124 | train_accuracy: 0.52244 | valid_accuracy: 0.54211 |  0:04:06s\n",
      "epoch 57 | loss: 0.99036 | train_accuracy: 0.52217 | valid_accuracy: 0.53684 |  0:04:11s\n",
      "epoch 58 | loss: 0.99012 | train_accuracy: 0.5245  | valid_accuracy: 0.53947 |  0:04:18s\n",
      "epoch 59 | loss: 0.99357 | train_accuracy: 0.52271 | valid_accuracy: 0.52895 |  0:04:23s\n",
      "epoch 60 | loss: 0.99047 | train_accuracy: 0.52217 | valid_accuracy: 0.54211 |  0:04:27s\n",
      "epoch 61 | loss: 0.99115 | train_accuracy: 0.52289 | valid_accuracy: 0.53158 |  0:04:32s\n",
      "epoch 62 | loss: 0.9898  | train_accuracy: 0.52316 | valid_accuracy: 0.54474 |  0:04:36s\n",
      "epoch 63 | loss: 0.99117 | train_accuracy: 0.52146 | valid_accuracy: 0.54474 |  0:04:44s\n",
      "epoch 64 | loss: 0.99135 | train_accuracy: 0.52101 | valid_accuracy: 0.53684 |  0:04:48s\n",
      "epoch 65 | loss: 0.99103 | train_accuracy: 0.52459 | valid_accuracy: 0.54211 |  0:04:52s\n",
      "epoch 66 | loss: 0.98968 | train_accuracy: 0.52459 | valid_accuracy: 0.54474 |  0:04:57s\n",
      "epoch 67 | loss: 0.98899 | train_accuracy: 0.5245  | valid_accuracy: 0.53684 |  0:05:01s\n",
      "epoch 68 | loss: 0.99117 | train_accuracy: 0.52584 | valid_accuracy: 0.54211 |  0:05:05s\n",
      "epoch 69 | loss: 0.99109 | train_accuracy: 0.52647 | valid_accuracy: 0.53947 |  0:05:09s\n",
      "epoch 70 | loss: 0.98959 | train_accuracy: 0.52414 | valid_accuracy: 0.53684 |  0:05:13s\n",
      "epoch 71 | loss: 0.98983 | train_accuracy: 0.52513 | valid_accuracy: 0.54474 |  0:05:18s\n",
      "epoch 72 | loss: 0.98924 | train_accuracy: 0.52334 | valid_accuracy: 0.54211 |  0:05:22s\n",
      "epoch 73 | loss: 0.98764 | train_accuracy: 0.52477 | valid_accuracy: 0.54211 |  0:05:26s\n",
      "epoch 74 | loss: 0.98884 | train_accuracy: 0.52593 | valid_accuracy: 0.54474 |  0:05:31s\n",
      "epoch 75 | loss: 0.98777 | train_accuracy: 0.52423 | valid_accuracy: 0.54474 |  0:05:36s\n",
      "epoch 76 | loss: 0.9885  | train_accuracy: 0.52369 | valid_accuracy: 0.53947 |  0:05:41s\n",
      "epoch 77 | loss: 0.98931 | train_accuracy: 0.52459 | valid_accuracy: 0.53947 |  0:05:45s\n",
      "epoch 78 | loss: 0.98725 | train_accuracy: 0.52513 | valid_accuracy: 0.55    |  0:05:50s\n",
      "epoch 79 | loss: 0.98637 | train_accuracy: 0.5253  | valid_accuracy: 0.53947 |  0:05:54s\n",
      "epoch 80 | loss: 0.98612 | train_accuracy: 0.52387 | valid_accuracy: 0.53947 |  0:05:58s\n",
      "epoch 81 | loss: 0.98867 | train_accuracy: 0.52387 | valid_accuracy: 0.53947 |  0:06:02s\n",
      "epoch 82 | loss: 0.98798 | train_accuracy: 0.52665 | valid_accuracy: 0.54211 |  0:06:07s\n",
      "epoch 83 | loss: 0.98606 | train_accuracy: 0.52781 | valid_accuracy: 0.55526 |  0:06:11s\n",
      "epoch 84 | loss: 0.98884 | train_accuracy: 0.52709 | valid_accuracy: 0.54474 |  0:06:15s\n",
      "epoch 85 | loss: 0.98681 | train_accuracy: 0.52808 | valid_accuracy: 0.53421 |  0:06:19s\n",
      "epoch 86 | loss: 0.98724 | train_accuracy: 0.52718 | valid_accuracy: 0.55    |  0:06:24s\n",
      "epoch 87 | loss: 0.98595 | train_accuracy: 0.52736 | valid_accuracy: 0.54474 |  0:06:28s\n",
      "epoch 88 | loss: 0.98813 | train_accuracy: 0.52682 | valid_accuracy: 0.55    |  0:06:32s\n",
      "epoch 89 | loss: 0.98724 | train_accuracy: 0.52629 | valid_accuracy: 0.54211 |  0:06:36s\n",
      "epoch 90 | loss: 0.98811 | train_accuracy: 0.52879 | valid_accuracy: 0.53947 |  0:06:40s\n",
      "epoch 91 | loss: 0.98696 | train_accuracy: 0.52647 | valid_accuracy: 0.54474 |  0:06:45s\n",
      "epoch 92 | loss: 0.98463 | train_accuracy: 0.52852 | valid_accuracy: 0.54474 |  0:06:49s\n",
      "epoch 93 | loss: 0.98392 | train_accuracy: 0.52843 | valid_accuracy: 0.52895 |  0:06:53s\n",
      "epoch 94 | loss: 0.9883  | train_accuracy: 0.52852 | valid_accuracy: 0.53421 |  0:06:57s\n",
      "epoch 95 | loss: 0.98761 | train_accuracy: 0.52691 | valid_accuracy: 0.53947 |  0:07:02s\n",
      "epoch 96 | loss: 0.98528 | train_accuracy: 0.52682 | valid_accuracy: 0.53947 |  0:07:06s\n",
      "epoch 97 | loss: 0.98731 | train_accuracy: 0.5262  | valid_accuracy: 0.55789 |  0:07:10s\n",
      "epoch 98 | loss: 0.98603 | train_accuracy: 0.52843 | valid_accuracy: 0.55263 |  0:07:14s\n",
      "epoch 99 | loss: 0.98422 | train_accuracy: 0.52754 | valid_accuracy: 0.54474 |  0:07:18s\n",
      "epoch 100| loss: 0.98311 | train_accuracy: 0.53058 | valid_accuracy: 0.54211 |  0:07:23s\n",
      "epoch 101| loss: 0.98298 | train_accuracy: 0.52897 | valid_accuracy: 0.54211 |  0:07:27s\n",
      "epoch 102| loss: 0.98379 | train_accuracy: 0.52933 | valid_accuracy: 0.53421 |  0:07:31s\n",
      "epoch 103| loss: 0.98491 | train_accuracy: 0.52995 | valid_accuracy: 0.53947 |  0:07:35s\n",
      "epoch 104| loss: 0.98789 | train_accuracy: 0.52977 | valid_accuracy: 0.53947 |  0:07:39s\n",
      "epoch 105| loss: 0.98314 | train_accuracy: 0.52897 | valid_accuracy: 0.53947 |  0:07:44s\n",
      "epoch 106| loss: 0.9865  | train_accuracy: 0.52906 | valid_accuracy: 0.53684 |  0:07:48s\n",
      "epoch 107| loss: 0.98419 | train_accuracy: 0.53156 | valid_accuracy: 0.54474 |  0:07:52s\n",
      "epoch 108| loss: 0.98439 | train_accuracy: 0.53138 | valid_accuracy: 0.55263 |  0:07:56s\n",
      "epoch 109| loss: 0.98319 | train_accuracy: 0.52888 | valid_accuracy: 0.53421 |  0:08:00s\n",
      "epoch 110| loss: 0.98417 | train_accuracy: 0.53058 | valid_accuracy: 0.53684 |  0:08:05s\n",
      "epoch 111| loss: 0.98379 | train_accuracy: 0.53192 | valid_accuracy: 0.54211 |  0:08:09s\n",
      "epoch 112| loss: 0.9855  | train_accuracy: 0.53192 | valid_accuracy: 0.54211 |  0:08:13s\n",
      "epoch 113| loss: 0.98122 | train_accuracy: 0.52951 | valid_accuracy: 0.53947 |  0:08:17s\n",
      "epoch 114| loss: 0.98506 | train_accuracy: 0.52888 | valid_accuracy: 0.53421 |  0:08:22s\n",
      "epoch 115| loss: 0.98287 | train_accuracy: 0.52942 | valid_accuracy: 0.53158 |  0:08:26s\n",
      "epoch 116| loss: 0.98675 | train_accuracy: 0.53192 | valid_accuracy: 0.54737 |  0:08:30s\n",
      "epoch 117| loss: 0.98342 | train_accuracy: 0.53264 | valid_accuracy: 0.54737 |  0:08:34s\n",
      "epoch 118| loss: 0.98439 | train_accuracy: 0.5304  | valid_accuracy: 0.53421 |  0:08:38s\n",
      "epoch 119| loss: 0.98268 | train_accuracy: 0.53156 | valid_accuracy: 0.53684 |  0:08:43s\n",
      "epoch 120| loss: 0.98264 | train_accuracy: 0.53174 | valid_accuracy: 0.53947 |  0:08:47s\n",
      "epoch 121| loss: 0.98197 | train_accuracy: 0.53067 | valid_accuracy: 0.54737 |  0:08:51s\n",
      "epoch 122| loss: 0.98183 | train_accuracy: 0.53156 | valid_accuracy: 0.54474 |  0:08:55s\n",
      "epoch 123| loss: 0.98259 | train_accuracy: 0.53147 | valid_accuracy: 0.53421 |  0:09:00s\n",
      "epoch 124| loss: 0.98303 | train_accuracy: 0.53371 | valid_accuracy: 0.55    |  0:09:04s\n",
      "epoch 125| loss: 0.98029 | train_accuracy: 0.53183 | valid_accuracy: 0.53684 |  0:09:08s\n",
      "epoch 126| loss: 0.98257 | train_accuracy: 0.5329  | valid_accuracy: 0.53158 |  0:09:12s\n",
      "epoch 127| loss: 0.97999 | train_accuracy: 0.53273 | valid_accuracy: 0.52895 |  0:09:16s\n",
      "epoch 128| loss: 0.98291 | train_accuracy: 0.53237 | valid_accuracy: 0.54474 |  0:09:21s\n",
      "epoch 129| loss: 0.98176 | train_accuracy: 0.53416 | valid_accuracy: 0.53158 |  0:09:25s\n",
      "epoch 130| loss: 0.98215 | train_accuracy: 0.5338  | valid_accuracy: 0.53947 |  0:09:29s\n",
      "epoch 131| loss: 0.98071 | train_accuracy: 0.53308 | valid_accuracy: 0.53947 |  0:09:33s\n",
      "epoch 132| loss: 0.98179 | train_accuracy: 0.53416 | valid_accuracy: 0.53158 |  0:09:38s\n",
      "epoch 133| loss: 0.98027 | train_accuracy: 0.53433 | valid_accuracy: 0.54474 |  0:09:42s\n",
      "epoch 134| loss: 0.98316 | train_accuracy: 0.53129 | valid_accuracy: 0.53684 |  0:09:46s\n",
      "epoch 135| loss: 0.98026 | train_accuracy: 0.53478 | valid_accuracy: 0.54737 |  0:09:50s\n",
      "epoch 136| loss: 0.98049 | train_accuracy: 0.5346  | valid_accuracy: 0.52632 |  0:09:55s\n",
      "epoch 137| loss: 0.98376 | train_accuracy: 0.53541 | valid_accuracy: 0.54474 |  0:09:59s\n",
      "epoch 138| loss: 0.97801 | train_accuracy: 0.53442 | valid_accuracy: 0.54474 |  0:10:03s\n",
      "epoch 139| loss: 0.98087 | train_accuracy: 0.53433 | valid_accuracy: 0.54474 |  0:10:07s\n",
      "epoch 140| loss: 0.97825 | train_accuracy: 0.53371 | valid_accuracy: 0.54737 |  0:10:12s\n",
      "epoch 141| loss: 0.97919 | train_accuracy: 0.53621 | valid_accuracy: 0.54737 |  0:10:16s\n",
      "epoch 142| loss: 0.98004 | train_accuracy: 0.53326 | valid_accuracy: 0.55    |  0:10:20s\n",
      "epoch 143| loss: 0.97908 | train_accuracy: 0.53523 | valid_accuracy: 0.54737 |  0:10:24s\n",
      "epoch 144| loss: 0.98156 | train_accuracy: 0.53478 | valid_accuracy: 0.56053 |  0:10:28s\n",
      "epoch 145| loss: 0.98242 | train_accuracy: 0.53612 | valid_accuracy: 0.54737 |  0:10:33s\n",
      "epoch 146| loss: 0.9821  | train_accuracy: 0.53523 | valid_accuracy: 0.54737 |  0:10:37s\n",
      "epoch 147| loss: 0.97821 | train_accuracy: 0.53729 | valid_accuracy: 0.54211 |  0:10:41s\n",
      "epoch 148| loss: 0.97897 | train_accuracy: 0.53496 | valid_accuracy: 0.54474 |  0:10:45s\n",
      "epoch 149| loss: 0.97599 | train_accuracy: 0.5363  | valid_accuracy: 0.55526 |  0:10:49s\n",
      "epoch 150| loss: 0.97874 | train_accuracy: 0.53621 | valid_accuracy: 0.54211 |  0:10:54s\n",
      "epoch 151| loss: 0.97633 | train_accuracy: 0.53621 | valid_accuracy: 0.54737 |  0:10:58s\n",
      "epoch 152| loss: 0.97625 | train_accuracy: 0.53505 | valid_accuracy: 0.53684 |  0:11:02s\n",
      "epoch 153| loss: 0.97893 | train_accuracy: 0.53425 | valid_accuracy: 0.54737 |  0:11:06s\n",
      "epoch 154| loss: 0.97707 | train_accuracy: 0.53362 | valid_accuracy: 0.54474 |  0:11:11s\n",
      "epoch 155| loss: 0.97823 | train_accuracy: 0.53612 | valid_accuracy: 0.53684 |  0:11:15s\n",
      "epoch 156| loss: 0.97843 | train_accuracy: 0.53487 | valid_accuracy: 0.53684 |  0:11:19s\n",
      "epoch 157| loss: 0.97791 | train_accuracy: 0.53532 | valid_accuracy: 0.53947 |  0:11:23s\n",
      "epoch 158| loss: 0.97912 | train_accuracy: 0.53317 | valid_accuracy: 0.54737 |  0:11:27s\n",
      "epoch 159| loss: 0.97796 | train_accuracy: 0.53344 | valid_accuracy: 0.54211 |  0:11:32s\n",
      "epoch 160| loss: 0.97822 | train_accuracy: 0.53478 | valid_accuracy: 0.55526 |  0:11:36s\n",
      "epoch 161| loss: 0.97999 | train_accuracy: 0.53568 | valid_accuracy: 0.54737 |  0:11:40s\n",
      "epoch 162| loss: 0.98027 | train_accuracy: 0.53577 | valid_accuracy: 0.56053 |  0:11:44s\n",
      "epoch 163| loss: 0.97622 | train_accuracy: 0.53603 | valid_accuracy: 0.55263 |  0:11:48s\n",
      "epoch 164| loss: 0.97814 | train_accuracy: 0.53746 | valid_accuracy: 0.55263 |  0:11:53s\n",
      "epoch 165| loss: 0.97707 | train_accuracy: 0.53416 | valid_accuracy: 0.55263 |  0:11:57s\n",
      "epoch 166| loss: 0.97624 | train_accuracy: 0.53442 | valid_accuracy: 0.55    |  0:12:01s\n",
      "epoch 167| loss: 0.97707 | train_accuracy: 0.53711 | valid_accuracy: 0.54211 |  0:12:05s\n",
      "epoch 168| loss: 0.97524 | train_accuracy: 0.53684 | valid_accuracy: 0.55789 |  0:12:09s\n",
      "epoch 169| loss: 0.97467 | train_accuracy: 0.53782 | valid_accuracy: 0.55    |  0:12:14s\n",
      "epoch 170| loss: 0.97614 | train_accuracy: 0.53737 | valid_accuracy: 0.55263 |  0:12:18s\n",
      "epoch 171| loss: 0.97471 | train_accuracy: 0.53934 | valid_accuracy: 0.54474 |  0:12:22s\n",
      "epoch 172| loss: 0.975   | train_accuracy: 0.53711 | valid_accuracy: 0.54737 |  0:12:26s\n",
      "epoch 173| loss: 0.97569 | train_accuracy: 0.53907 | valid_accuracy: 0.54211 |  0:12:31s\n",
      "epoch 174| loss: 0.97734 | train_accuracy: 0.53881 | valid_accuracy: 0.53421 |  0:12:35s\n",
      "epoch 175| loss: 0.97481 | train_accuracy: 0.54068 | valid_accuracy: 0.54474 |  0:12:39s\n",
      "epoch 176| loss: 0.97509 | train_accuracy: 0.54292 | valid_accuracy: 0.53158 |  0:12:43s\n",
      "epoch 177| loss: 0.9772  | train_accuracy: 0.53854 | valid_accuracy: 0.54211 |  0:12:47s\n",
      "epoch 178| loss: 0.97382 | train_accuracy: 0.53961 | valid_accuracy: 0.54474 |  0:12:52s\n",
      "epoch 179| loss: 0.97706 | train_accuracy: 0.53729 | valid_accuracy: 0.54211 |  0:12:56s\n",
      "epoch 180| loss: 0.97572 | train_accuracy: 0.53755 | valid_accuracy: 0.54474 |  0:13:00s\n",
      "epoch 181| loss: 0.97574 | train_accuracy: 0.53711 | valid_accuracy: 0.54737 |  0:13:04s\n",
      "epoch 182| loss: 0.97531 | train_accuracy: 0.53997 | valid_accuracy: 0.55526 |  0:13:09s\n",
      "epoch 183| loss: 0.97568 | train_accuracy: 0.53809 | valid_accuracy: 0.55    |  0:13:13s\n",
      "epoch 184| loss: 0.97369 | train_accuracy: 0.53952 | valid_accuracy: 0.55263 |  0:13:17s\n",
      "epoch 185| loss: 0.97531 | train_accuracy: 0.53845 | valid_accuracy: 0.55    |  0:13:21s\n",
      "epoch 186| loss: 0.97347 | train_accuracy: 0.53898 | valid_accuracy: 0.53684 |  0:13:25s\n",
      "epoch 187| loss: 0.97295 | train_accuracy: 0.53916 | valid_accuracy: 0.53421 |  0:13:30s\n",
      "epoch 188| loss: 0.97474 | train_accuracy: 0.54024 | valid_accuracy: 0.53947 |  0:13:34s\n",
      "epoch 189| loss: 0.97478 | train_accuracy: 0.54033 | valid_accuracy: 0.52368 |  0:13:38s\n",
      "epoch 190| loss: 0.9725  | train_accuracy: 0.5405  | valid_accuracy: 0.54474 |  0:13:42s\n",
      "epoch 191| loss: 0.97075 | train_accuracy: 0.53925 | valid_accuracy: 0.52105 |  0:13:47s\n",
      "epoch 192| loss: 0.97755 | train_accuracy: 0.53988 | valid_accuracy: 0.53421 |  0:13:51s\n",
      "epoch 193| loss: 0.97219 | train_accuracy: 0.53898 | valid_accuracy: 0.52368 |  0:13:55s\n",
      "epoch 194| loss: 0.97527 | train_accuracy: 0.53934 | valid_accuracy: 0.54211 |  0:13:59s\n",
      "epoch 195| loss: 0.97595 | train_accuracy: 0.53934 | valid_accuracy: 0.52895 |  0:14:03s\n",
      "epoch 196| loss: 0.97322 | train_accuracy: 0.54077 | valid_accuracy: 0.53947 |  0:14:08s\n",
      "epoch 197| loss: 0.9704  | train_accuracy: 0.54104 | valid_accuracy: 0.54211 |  0:14:12s\n",
      "epoch 198| loss: 0.97236 | train_accuracy: 0.53854 | valid_accuracy: 0.55    |  0:14:16s\n",
      "epoch 199| loss: 0.97018 | train_accuracy: 0.54077 | valid_accuracy: 0.53421 |  0:14:20s\n",
      "epoch 200| loss: 0.9727  | train_accuracy: 0.53916 | valid_accuracy: 0.53421 |  0:14:25s\n",
      "epoch 201| loss: 0.96901 | train_accuracy: 0.54059 | valid_accuracy: 0.55    |  0:14:29s\n",
      "epoch 202| loss: 0.97284 | train_accuracy: 0.53827 | valid_accuracy: 0.54737 |  0:14:33s\n",
      "epoch 203| loss: 0.96978 | train_accuracy: 0.54229 | valid_accuracy: 0.54474 |  0:14:37s\n",
      "epoch 204| loss: 0.9706  | train_accuracy: 0.54077 | valid_accuracy: 0.53947 |  0:14:41s\n",
      "epoch 205| loss: 0.974   | train_accuracy: 0.53988 | valid_accuracy: 0.55    |  0:14:46s\n",
      "epoch 206| loss: 0.96883 | train_accuracy: 0.54095 | valid_accuracy: 0.55263 |  0:14:50s\n",
      "epoch 207| loss: 0.97343 | train_accuracy: 0.54202 | valid_accuracy: 0.54211 |  0:14:54s\n",
      "epoch 208| loss: 0.97307 | train_accuracy: 0.5414  | valid_accuracy: 0.53947 |  0:14:58s\n",
      "epoch 209| loss: 0.97175 | train_accuracy: 0.54024 | valid_accuracy: 0.53421 |  0:15:02s\n",
      "epoch 210| loss: 0.97276 | train_accuracy: 0.54202 | valid_accuracy: 0.54474 |  0:15:07s\n",
      "epoch 211| loss: 0.96846 | train_accuracy: 0.54158 | valid_accuracy: 0.53947 |  0:15:11s\n",
      "epoch 212| loss: 0.97217 | train_accuracy: 0.54131 | valid_accuracy: 0.54211 |  0:15:15s\n",
      "epoch 213| loss: 0.96982 | train_accuracy: 0.54086 | valid_accuracy: 0.53421 |  0:15:19s\n",
      "epoch 214| loss: 0.97073 | train_accuracy: 0.54068 | valid_accuracy: 0.53947 |  0:15:23s\n",
      "epoch 215| loss: 0.97735 | train_accuracy: 0.53988 | valid_accuracy: 0.53947 |  0:15:28s\n",
      "epoch 216| loss: 0.971   | train_accuracy: 0.54417 | valid_accuracy: 0.54737 |  0:15:32s\n",
      "epoch 217| loss: 0.97229 | train_accuracy: 0.54399 | valid_accuracy: 0.54737 |  0:15:36s\n",
      "epoch 218| loss: 0.9705  | train_accuracy: 0.54363 | valid_accuracy: 0.55263 |  0:15:40s\n",
      "epoch 219| loss: 0.97426 | train_accuracy: 0.53943 | valid_accuracy: 0.54737 |  0:15:45s\n",
      "epoch 220| loss: 0.97373 | train_accuracy: 0.54158 | valid_accuracy: 0.55263 |  0:15:49s\n",
      "epoch 221| loss: 0.97256 | train_accuracy: 0.54185 | valid_accuracy: 0.54474 |  0:15:53s\n",
      "epoch 222| loss: 0.9737  | train_accuracy: 0.54363 | valid_accuracy: 0.55    |  0:15:57s\n",
      "epoch 223| loss: 0.96939 | train_accuracy: 0.54489 | valid_accuracy: 0.55263 |  0:16:02s\n",
      "epoch 224| loss: 0.97077 | train_accuracy: 0.54524 | valid_accuracy: 0.54737 |  0:16:06s\n",
      "epoch 225| loss: 0.97617 | train_accuracy: 0.54685 | valid_accuracy: 0.56053 |  0:16:10s\n",
      "epoch 226| loss: 0.96963 | train_accuracy: 0.54542 | valid_accuracy: 0.53421 |  0:16:14s\n",
      "epoch 227| loss: 0.97321 | train_accuracy: 0.54641 | valid_accuracy: 0.54211 |  0:16:18s\n",
      "epoch 228| loss: 0.97245 | train_accuracy: 0.54471 | valid_accuracy: 0.53421 |  0:16:23s\n",
      "epoch 229| loss: 0.9707  | train_accuracy: 0.54587 | valid_accuracy: 0.53947 |  0:16:27s\n",
      "epoch 230| loss: 0.96961 | train_accuracy: 0.54775 | valid_accuracy: 0.54737 |  0:16:32s\n",
      "epoch 231| loss: 0.96944 | train_accuracy: 0.54605 | valid_accuracy: 0.54211 |  0:16:36s\n",
      "epoch 232| loss: 0.96781 | train_accuracy: 0.54748 | valid_accuracy: 0.53684 |  0:16:40s\n",
      "epoch 233| loss: 0.96786 | train_accuracy: 0.5473  | valid_accuracy: 0.53421 |  0:16:45s\n",
      "epoch 234| loss: 0.96954 | train_accuracy: 0.54381 | valid_accuracy: 0.52895 |  0:16:49s\n",
      "epoch 235| loss: 0.969   | train_accuracy: 0.54363 | valid_accuracy: 0.53684 |  0:16:53s\n",
      "epoch 236| loss: 0.97168 | train_accuracy: 0.54614 | valid_accuracy: 0.53684 |  0:16:58s\n",
      "epoch 237| loss: 0.96776 | train_accuracy: 0.54506 | valid_accuracy: 0.53684 |  0:17:02s\n",
      "epoch 238| loss: 0.97061 | train_accuracy: 0.5456  | valid_accuracy: 0.54474 |  0:17:06s\n",
      "epoch 239| loss: 0.96821 | train_accuracy: 0.54596 | valid_accuracy: 0.53947 |  0:17:11s\n",
      "epoch 240| loss: 0.97156 | train_accuracy: 0.5448  | valid_accuracy: 0.53947 |  0:17:15s\n",
      "epoch 241| loss: 0.97124 | train_accuracy: 0.54426 | valid_accuracy: 0.53684 |  0:17:19s\n",
      "epoch 242| loss: 0.96768 | train_accuracy: 0.54489 | valid_accuracy: 0.53158 |  0:17:23s\n",
      "epoch 243| loss: 0.96657 | train_accuracy: 0.54909 | valid_accuracy: 0.52632 |  0:17:27s\n",
      "epoch 244| loss: 0.9683  | train_accuracy: 0.54632 | valid_accuracy: 0.52895 |  0:17:32s\n",
      "\n",
      "Early stopping occurred at epoch 244 with best_epoch = 144 and best_valid_accuracy = 0.56053\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/gustavofernandez/Desktop/iPre-resultados-deportivos/env/lib/python3.8/site-packages/pytorch_tabnet/callbacks.py:172: UserWarning: Best weights from best epoch are automatically used!\n",
      "  warnings.warn(wrn_msg)\n"
     ]
    }
   ],
   "source": [
    "\n",
    "clf = TabNetClassifier(\n",
    "    optimizer_fn=torch.optim.Adam,\n",
    "    optimizer_params=dict(lr=1e-3),\n",
    "    scheduler_params={\"step_size\":50,\n",
    "                        \"gamma\":0.9},\n",
    "    scheduler_fn=torch.optim.lr_scheduler.StepLR,\n",
    "    mask_type='entmax', # \"sparsemax\" \"entmax\"\n",
    "    )\n",
    "    \n",
    "clf.fit(X_train, y_train,\n",
    "    eval_set=[(X_train, y_train), (X_test, y_test)], \n",
    "    eval_name=['train', 'valid'], \n",
    "    eval_metric=['accuracy'], \n",
    "    max_epochs=1000 , patience=100, \n",
    "    batch_size=28, drop_last=False)\n",
    "\n",
    "pred = clf.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.00      0.00      0.00        87\n",
      "           1       0.58      0.88      0.70       184\n",
      "           2       0.51      0.48      0.50       109\n",
      "\n",
      "    accuracy                           0.56       380\n",
      "   macro avg       0.36      0.45      0.40       380\n",
      "weighted avg       0.43      0.56      0.48       380\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(\"classification report\")\n",
    "print(classification_report(y_test, pred))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Normalized confusion matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 500x500 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "        print(\"Normalized confusion matrix\")\n",
    "    else:\n",
    "        print('Confusion matrix, without normalization')\n",
    "\n",
    "\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=90)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    fmt = '.2f' if normalize else 'd'\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, format(cm[i, j], fmt),\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')\n",
    "\n",
    "cnf_matrix = confusion_matrix(y_test, pred)\n",
    "classes = np.unique(y_train)\n",
    "np.set_printoptions(precision=2)\n",
    "# Plot normalized confusion matrix\n",
    "plt.figure(figsize=(5,5))\n",
    "\n",
    "plot_confusion_matrix(cnf_matrix, classes=classes, normalize=True, title='Normalized confusion matrix')\n",
    "\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}