WNCAA_2018_Predictions.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 460,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import math\n",
    "from sklearn import cross_validation, linear_model\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "import csv\n",
    "import random\n",
    "import numpy as np\n",
    "import matplotlib as plt\n",
    "import seaborn as sns\n",
    "% matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "I used code adapted from this Kaggle user: https://www.kaggle.com/mineshjethva/2018-logistics-lb-public-1-5\n",
    "I added features based on spreads of assists, personal foals, and defensive rebounds.\n",
    "The model already incorporated seed differences and score spreads."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 461,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1344, 3)\n",
      "   Season Seed  TeamID\n",
      "0    1998  W01    3330\n",
      "1    1998  W02    3163\n",
      "2    1998  W03    3112\n",
      "3    1998  W04    3301\n",
      "4    1998  W05    3272\n",
      "(46342, 34)\n",
      "   Season  DayNum  WTeamID  WScore  LTeamID  LScore WLoc  NumOT  WFGM  WFGA  \\\n",
      "0    2010      11     3103      63     3237      49    H      0    23    54   \n",
      "1    2010      11     3104      73     3399      68    N      0    26    62   \n",
      "2    2010      11     3110      71     3224      59    A      0    29    62   \n",
      "3    2010      11     3111      63     3267      58    A      0    27    52   \n",
      "4    2010      11     3119      74     3447      70    H      1    30    74   \n",
      "\n",
      "  ...   LFGA3  LFTM  LFTA  LOR  LDR  LAst  LTO  LStl  LBlk  LPF  \n",
      "0 ...      13     6    10   11   27    11   23     7     6   19  \n",
      "1 ...      21    14    27   14   26     7   20     4     2   27  \n",
      "2 ...      14    19    23   17   23     8   15     6     0   15  \n",
      "3 ...      26    16    25   22   22    15   11    14     5   14  \n",
      "4 ...      17    11    21   21   32    12   14     4     2   14  \n",
      "\n",
      "[5 rows x 34 columns]\n"
     ]
    }
   ],
   "source": [
    "\n",
    "#the seed information\n",
    "df_seeds = pd.read_csv('WNCAATourneySeeds2018.csv')\n",
    "print(df_seeds.shape)\n",
    "print(df_seeds.head())\n",
    "#print(df_seeds.Season.value_counts())\n",
    "\n",
    "#tour information\n",
    "df_tour = pd.read_csv('WRegularSeasonDetailedResults.csv')\n",
    "print(df_tour.shape)\n",
    "print(df_tour.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 462,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4398, 36)\n",
      "   Season  DayNum  WTeamID  WScore  LTeamID  LScore WLoc  NumOT  WFGM  WFGA  \\\n",
      "0    2010      11     3173      77     3277      74    H      0    26    65   \n",
      "1    2010      17     3323      68     3277      67    A      0    22    65   \n",
      "2    2010      34     3231      68     3277      63    A      0    19    60   \n",
      "3    2010      56     3458      62     3277      54    H      0    20    46   \n",
      "4    2010      68     3326      65     3277      62    A      0    18    52   \n",
      "\n",
      "   ...    LFTA  LOR  LDR  LAst  LTO  LStl  LBlk  LPF  WSeed  LSeed  \n",
      "0  ...      26   17   28    16   22     7     9   29    8.0      5  \n",
      "1  ...      22   14   31    13   21     6     9   22    2.0      5  \n",
      "2  ...      27   27   24    10   10     6    10   24    NaN      5  \n",
      "3  ...      17   10   20    11   26     7     3   21    7.0      5  \n",
      "4  ...      26   14   23     9   17     9     4   20    2.0      5  \n",
      "\n",
      "[5 rows x 36 columns]\n"
     ]
    }
   ],
   "source": [
    "df_seeds['seed_int'] = df_seeds['Seed'].apply( lambda x : int(x[1:3]) )\n",
    "\n",
    "df_winseeds = df_seeds.loc[:, ['TeamID', 'Season', 'seed_int']].rename(columns={'TeamID':'WTeamID', 'seed_int':'WSeed'})\n",
    "df_lossseeds = df_seeds.loc[:, ['TeamID', 'Season', 'seed_int']].rename(columns={'TeamID':'LTeamID', 'seed_int':'LSeed'})\n",
    "df_dummy = pd.merge(left=df_tour, right=df_winseeds, how='left', on=['Season', 'WTeamID'])\n",
    "df_concat = pd.merge(left=df_dummy, right=df_lossseeds, on=['Season', 'LTeamID'])\n",
    "print(df_concat.shape)\n",
    "print(df_concat.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 463,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(4398, 37)\n",
      "   Season  DayNum  WTeamID  WScore  LTeamID  LScore WLoc  NumOT  WFGM  WFGA  \\\n",
      "0    2010      11     3173      77     3277      74    H      0    26    65   \n",
      "1    2010      17     3323      68     3277      67    A      0    22    65   \n",
      "2    2010      34     3231      68     3277      63    A      0    19    60   \n",
      "3    2010      56     3458      62     3277      54    H      0    20    46   \n",
      "4    2010      68     3326      65     3277      62    A      0    18    52   \n",
      "\n",
      "     ...     LOR  LDR  LAst  LTO  LStl  LBlk  LPF  WSeed  LSeed  DiffSeed  \n",
      "0    ...      17   28    16   22     7     9   29    8.0      5         1  \n",
      "1    ...      14   31    13   21     6     9   22    2.0      5         1  \n",
      "2    ...      27   24    10   10     6    10   24    NaN      5         1  \n",
      "3    ...      10   20    11   26     7     3   21    7.0      5         1  \n",
      "4    ...      14   23     9   17     9     4   20    2.0      5         1  \n",
      "\n",
      "[5 rows x 37 columns]\n",
      "2011    501\n",
      "2014    499\n",
      "2017    499\n",
      "2010    497\n",
      "2013    491\n",
      "2012    491\n",
      "2018    477\n",
      "2015    475\n",
      "2016    468\n",
      "Name: Season, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "df_concat['DiffSeed'] = df_concat[['LSeed', 'WSeed']].apply(lambda x : 0 if x[0] == x[1] else 1, axis = 1)\n",
    "print(df_concat.shape)\n",
    "print(df_concat.head())\n",
    "print(df_concat.Season.value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 464,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10080, 2)\n",
      "               ID  Pred\n",
      "0  2014_3103_3107   0.5\n",
      "1  2014_3103_3113   0.5\n",
      "2  2014_3103_3119   0.5\n",
      "3  2014_3103_3124   0.5\n",
      "4  2014_3103_3140   0.5\n"
     ]
    }
   ],
   "source": [
    "df_sample_sub1 = pd.read_csv('WSampleSubmissionStage1.csv')\n",
    "\n",
    "#prepares sample submission\n",
    "df_sample_sub2 = pd.read_csv('WSampleSubmissionStage2.csv')\n",
    "\n",
    "df_sample_sub=pd.concat([df_sample_sub1, df_sample_sub2])\n",
    "\n",
    "print(df_sample_sub.shape)\n",
    "print(df_sample_sub.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 465,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10080, 5)\n",
      "               ID  Pred  Season  TeamID1  TeamID2\n",
      "0  2014_3103_3107   0.5    2014     3103     3107\n",
      "1  2014_3103_3113   0.5    2014     3103     3113\n",
      "2  2014_3103_3119   0.5    2014     3103     3119\n",
      "3  2014_3103_3124   0.5    2014     3103     3124\n",
      "4  2014_3103_3140   0.5    2014     3103     3140\n"
     ]
    }
   ],
   "source": [
    "df_sample_sub['Season'] = df_sample_sub['ID'].apply(lambda x : int(x.split('_')[0]) )\n",
    "df_sample_sub['TeamID1'] = df_sample_sub['ID'].apply(lambda x : int(x.split('_')[1]) )\n",
    "df_sample_sub['TeamID2'] = df_sample_sub['ID'].apply(lambda x : int(x.split('_')[2]) )\n",
    "print(df_sample_sub.shape)\n",
    "print(df_sample_sub.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 466,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(8796, 27)\n",
      "   DayNum  DiffSeed  NumOT  Result  Season  Team1_Ast  Team1_DR  Team1_PF  \\\n",
      "0      11         1      0     1.0    2010         14        31        22   \n",
      "1      17         1      0     1.0    2010         12        29        20   \n",
      "2      34         1      0     1.0    2010          6        26        22   \n",
      "3      56         1      0     1.0    2010         12        16        20   \n",
      "4      68         1      0     1.0    2010          8        27        20   \n",
      "\n",
      "   Team1_Score  Team2_Ast    ...     Score_Pct  DR_Ratio  DR_Total    DR_Pct  \\\n",
      "0           77         16    ...      0.509934  1.107143        59  0.525424   \n",
      "1           68         13    ...      0.503704  0.935484        60  0.483333   \n",
      "2           68         10    ...      0.519084  1.083333        50  0.520000   \n",
      "3           62         11    ...      0.534483  0.800000        36  0.444444   \n",
      "4           65          9    ...      0.511811  1.173913        50  0.540000   \n",
      "\n",
      "   Ast_Ratio  Ast_Total   Ast_Pct  PF_Ratio  PF_Total    PF_Pct  \n",
      "0   0.875000         30  0.466667  0.758621        51  0.431373  \n",
      "1   0.923077         25  0.480000  0.909091        42  0.476190  \n",
      "2   0.600000         16  0.375000  0.916667        46  0.478261  \n",
      "3   1.090909         23  0.521739  0.952381        41  0.487805  \n",
      "4   0.888889         17  0.470588  1.000000        40  0.500000  \n",
      "\n",
      "[5 rows x 27 columns]\n"
     ]
    }
   ],
   "source": [
    "winners = df_concat.rename( columns = { 'WTeamID' : 'TeamID1', \n",
    "                                                       'LTeamID' : 'TeamID2',\n",
    "                                                      'WScore' : 'Team1_Score',\n",
    "                                                      'LScore' : 'Team2_Score',\n",
    "                                                      'WAst': 'Team1_Ast',\n",
    "                                                      'LAst': 'Team2_Ast',\n",
    "                                                      'WDR': 'Team1_DR',\n",
    "                                                      'LDR': 'Team2_DR',\n",
    "                                                      'WPF': 'Team1_PF',\n",
    "                                                      'LPF': 'Team2_PF'}).drop(['WSeed', 'LSeed', 'WLoc', 'LBlk', 'LFGA', 'LFGA3', 'LFGM', 'LFGM3',\n",
    "       'LFTA', 'LFTM', 'LOR', 'LStl', 'LTO', 'WBlk', 'WFGA', 'WFGA3', 'WFGM', 'WFGM3',\n",
    "       'WFTA', 'WFTM', 'WOR', 'WStl', 'WTO'], axis = 1)\n",
    "winners['Result'] = 1.0\n",
    "\n",
    "losers = df_concat.rename( columns = { 'WTeamID' : 'TeamID2', \n",
    "                                                       'LTeamID' : 'TeamID1',\n",
    "                                                      'WScore' : 'Team2_Score',\n",
    "                                                      'LScore' : 'Team1_Score',\n",
    "                                                    'WAst': 'Team2_Ast',\n",
    "                                                      'LAst': 'Team1_Ast',\n",
    "                                                      'WDR': 'Team2_DR',\n",
    "                                                      'LDR': 'Team1_DR',\n",
    "                                                     'WPF': 'Team2_PF',\n",
    "                                                      'LPF': 'Team1_PF',}).drop(['WSeed', 'LSeed', 'WLoc', 'LBlk', 'LFGA', 'LFGA3', 'LFGM', 'LFGM3',\n",
    "       'LFTA', 'LFTM', 'LOR', 'LStl', 'LTO', 'WBlk', 'WFGA', 'WFGA3', 'WFGM', 'WFGM3',\n",
    "       'WFTA', 'WFTM', 'WOR', 'WStl', 'WTO'], axis = 1)\n",
    "\n",
    "losers['Result'] = 0.0\n",
    "\n",
    "train = pd.concat( [winners, losers], axis = 0).reset_index(drop = True)\n",
    "\n",
    "train['Score_Ratio'] = train['Team1_Score'] / train['Team2_Score']\n",
    "train['Score_Total'] = train['Team1_Score'] + train['Team2_Score']\n",
    "train['Score_Pct'] = train['Team1_Score'] / train['Score_Total']\n",
    "\n",
    "train['DR_Ratio'] = train['Team1_DR'] / train['Team2_DR']\n",
    "train['DR_Total'] = train['Team1_DR'] + train['Team2_DR']\n",
    "train['DR_Pct'] = train['Team1_DR'] / train['DR_Total']\n",
    "\n",
    "train['Ast_Ratio'] = train['Team1_Ast'] / train['Team2_Ast']\n",
    "train['Ast_Total'] = train['Team1_Ast'] + train['Team2_Ast']\n",
    "train['Ast_Pct'] = train['Team1_Ast'] / train['Ast_Total']\n",
    "\n",
    "train['PF_Ratio'] = train['Team1_PF'] / train['Team2_PF']\n",
    "train['PF_Total'] = train['Team1_PF'] + train['Team2_PF']\n",
    "train['PF_Pct'] = train['Team1_PF'] / train['PF_Total']\n",
    "\n",
    "print(train.shape)\n",
    "print(train.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 467,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['DayNum', 'DiffSeed', 'NumOT', 'Result', 'Season', 'Team1_Ast',\n",
       "       'Team1_DR', 'Team1_PF', 'Team1_Score', 'Team2_Ast', 'Team2_DR',\n",
       "       'Team2_PF', 'Team2_Score', 'TeamID1', 'TeamID2', 'Score_Ratio',\n",
       "       'Score_Total', 'Score_Pct', 'DR_Ratio', 'DR_Total', 'DR_Pct',\n",
       "       'Ast_Ratio', 'Ast_Total', 'Ast_Pct', 'PF_Ratio', 'PF_Total', 'PF_Pct'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 467,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 468,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DayNum         0\n",
       "DiffSeed       0\n",
       "NumOT          0\n",
       "Result         0\n",
       "Season         0\n",
       "Team1_Ast      0\n",
       "Team1_DR       0\n",
       "Team1_PF       0\n",
       "Team1_Score    0\n",
       "Team2_Ast      0\n",
       "Team2_DR       0\n",
       "Team2_PF       0\n",
       "Team2_Score    0\n",
       "TeamID1        0\n",
       "TeamID2        0\n",
       "Score_Ratio    0\n",
       "Score_Total    0\n",
       "Score_Pct      0\n",
       "DR_Ratio       0\n",
       "DR_Total       0\n",
       "DR_Pct         0\n",
       "Ast_Ratio      0\n",
       "Ast_Total      0\n",
       "Ast_Pct        0\n",
       "PF_Ratio       0\n",
       "PF_Total       0\n",
       "PF_Pct         0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 468,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 469,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    8796.000000\n",
       "mean        1.026322\n",
       "std         0.241386\n",
       "min         0.242991\n",
       "25%         0.867647\n",
       "50%         1.000049\n",
       "75%         1.152542\n",
       "max         4.115385\n",
       "Name: Score_Ratio, dtype: float64"
      ]
     },
     "execution_count": 469,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train.Score_Ratio.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 470,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "years = [2014, 2015, 2016, 2017,2018]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 471,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_inner = pd.merge( train.loc[ train['Season'].isin(years), : ].reset_index(drop = True), \n",
    "         df_sample_sub.drop(['ID', 'Pred'], axis = 1), \n",
    "         on = ['Season', 'TeamID1', 'TeamID2'], how = 'inner' )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 472,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DayNum         0\n",
       "DiffSeed       0\n",
       "NumOT          0\n",
       "Result         0\n",
       "Season         0\n",
       "Team1_Ast      0\n",
       "Team1_DR       0\n",
       "Team1_PF       0\n",
       "Team1_Score    0\n",
       "Team2_Ast      0\n",
       "Team2_DR       0\n",
       "Team2_PF       0\n",
       "Team2_Score    0\n",
       "TeamID1        0\n",
       "TeamID2        0\n",
       "Score_Ratio    0\n",
       "Score_Total    0\n",
       "Score_Pct      0\n",
       "DR_Ratio       0\n",
       "DR_Total       0\n",
       "DR_Pct         0\n",
       "Ast_Ratio      0\n",
       "Ast_Total      0\n",
       "Ast_Pct        0\n",
       "PF_Ratio       0\n",
       "PF_Total       0\n",
       "PF_Pct         0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 472,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_inner.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 473,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1380, 27)"
      ]
     },
     "execution_count": 473,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_inner.head()\n",
    "train_test_inner.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 474,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['DayNum', 'DiffSeed', 'NumOT', 'Result', 'Season', 'Team1_Ast',\n",
       "       'Team1_DR', 'Team1_PF', 'Team1_Score', 'Team2_Ast', 'Team2_DR',\n",
       "       'Team2_PF', 'Team2_Score', 'TeamID1', 'TeamID2', 'Score_Ratio',\n",
       "       'Score_Total', 'Score_Pct', 'DR_Ratio', 'DR_Total', 'DR_Pct',\n",
       "       'Ast_Ratio', 'Ast_Total', 'Ast_Pct', 'PF_Ratio', 'PF_Total', 'PF_Pct'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 474,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_inner.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 475,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Season</th>\n",
       "      <th>TeamID1</th>\n",
       "      <th>NumOT1</th>\n",
       "      <th>TeamID2</th>\n",
       "      <th>NumOT2</th>\n",
       "      <th>NumOT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>3103</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3143</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>3103</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3163</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>3103</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3169</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>3103</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3173</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>3103</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3181</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Season  TeamID1  NumOT1  TeamID2  NumOT2  NumOT\n",
       "0    2014     3103     0.0     3143     0.0      0\n",
       "1    2014     3103     0.0     3163     0.0      0\n",
       "2    2014     3103     0.0     3169     0.0      0\n",
       "3    2014     3103     0.0     3173     0.0      0\n",
       "4    2014     3103     0.0     3181     0.0      0"
      ]
     },
     "execution_count": 475,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#From the inner join, we will create data per team id to estimate the parameters we are missing that are independent of the year. Essentially, we are trying to estimate the average behavior of the team across the year.\n",
    "\n",
    "team1d_num_ot = train_test_inner.groupby(['Season', 'TeamID1'])['NumOT'].median().reset_index()\\\n",
    ".set_index('Season').rename(columns = {'NumOT' : 'NumOT1'})\n",
    "team2d_num_ot = train_test_inner.groupby(['Season', 'TeamID2'])['NumOT'].median().reset_index()\\\n",
    ".set_index('Season').rename(columns = {'NumOT' : 'NumOT2'})\n",
    "\n",
    "num_ot = team1d_num_ot.join(team2d_num_ot).reset_index()\n",
    "\n",
    "#sum the number of ot calls and subtract by one to prevent overcounting\n",
    "num_ot['NumOT'] = num_ot[['NumOT1', 'NumOT2']].apply(lambda x : round( x.sum() ), axis = 1 )\n",
    "\n",
    "num_ot.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 476,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#Here we look at the comparable statistics. For the TeamID2 column, we would consider the inverse of the ratio, and 1 minus the score attempt percentage.\n",
    "\n",
    "def geo_mean( x ):\n",
    "    return np.exp( np.mean(np.log(x)) )\n",
    "\n",
    "def harm_mean( x ):\n",
    "    return np.mean( x ** -1.0 ) ** -1.0\n",
    "\n",
    "team1d_score_spread = train_test_inner.groupby(['Season', 'TeamID1'])[['Score_Ratio', 'Score_Pct']]\\\n",
    ".agg({ 'Score_Ratio': geo_mean, 'Score_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'Score_Ratio' : 'Score_Ratio1', 'Score_Pct' : 'Score_Pct1'})\n",
    "\n",
    "team2d_score_spread = train_test_inner.groupby(['Season', 'TeamID2'])[['Score_Ratio', 'Score_Pct']]\\\n",
    ".agg({ 'Score_Ratio': geo_mean, 'Score_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'Score_Ratio' : 'Score_Ratio2', 'Score_Pct' : 'Score_Pct2'})\n",
    "\n",
    "\n",
    "score_spread = team1d_score_spread.join(team2d_score_spread).reset_index()\n",
    "\n",
    "#geometric mean of score ratio of team 1 and inverse of team 2\n",
    "score_spread['Score_Ratio'] = score_spread[['Score_Ratio1', 'Score_Ratio2']].apply(lambda x : ( x[0] * ( x[1] ** -1.0) ), axis = 1 ) ** 0.5\n",
    "\n",
    "#harmonic mean of score pct\n",
    "score_spread['Score_Pct'] = score_spread[['Score_Pct1', 'Score_Pct2']].apply(lambda x : 0.5*( x[0] ** -1.0 ) + 0.5*( 1.0 - x[1] ) ** -1.0, axis = 1 ) ** -1.0\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 477,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "team1d_ast_spread = train_test_inner.groupby(['Season', 'TeamID1'])[['Ast_Ratio', 'Ast_Pct']]\\\n",
    ".agg({ 'Ast_Ratio': geo_mean, 'Ast_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'Ast_Ratio' : 'Ast_Ratio1', 'Ast_Pct' : 'Ast_Pct1'})\n",
    "\n",
    "team2d_ast_spread = train_test_inner.groupby(['Season', 'TeamID2'])[['Ast_Ratio', 'Ast_Pct']]\\\n",
    ".agg({ 'Ast_Ratio': geo_mean, 'Ast_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'Ast_Ratio' : 'Ast_Ratio2', 'Ast_Pct' : 'Ast_Pct2'})\n",
    "\n",
    "\n",
    "ast_spread = team1d_ast_spread.join(team2d_ast_spread).reset_index()\n",
    "\n",
    "#geometric mean of score ratio of team 1 and inverse of team 2\n",
    "score_spread['Ast_Ratio'] = ast_spread[['Ast_Ratio1', 'Ast_Ratio2']].apply(lambda x : ( x[0] * ( x[1] ** -1.0) ), axis = 1 ) ** 0.5\n",
    "\n",
    "#harmonic mean of score pct\n",
    "score_spread['Ast_Pct'] = ast_spread[['Ast_Pct1', 'Ast_Pct2']].apply(lambda x : 0.5*( x[0] ** -1.0 ) + 0.5*( 1.0 - x[1] ) ** -1.0, axis = 1 ) ** -1.0\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 478,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "team1d_DR_spread = train_test_inner.groupby(['Season', 'TeamID1'])[['DR_Ratio', 'DR_Pct']]\\\n",
    ".agg({ 'DR_Ratio': geo_mean, 'DR_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'DR_Ratio' : 'DR_Ratio1', 'DR_Pct' : 'DR_Pct1'})\n",
    "\n",
    "team2d_DR_spread = train_test_inner.groupby(['Season', 'TeamID2'])[['DR_Ratio', 'DR_Pct']]\\\n",
    ".agg({ 'DR_Ratio': geo_mean, 'DR_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'DR_Ratio' : 'DR_Ratio2', 'DR_Pct' : 'DR_Pct2'})\n",
    "\n",
    "\n",
    "DR_spread = team1d_DR_spread.join(team2d_DR_spread).reset_index()\n",
    "\n",
    "#geometric mean of score ratio of team 1 and inverse of team 2\n",
    "score_spread['DR_Ratio'] = DR_spread[['DR_Ratio1', 'DR_Ratio2']].apply(lambda x : ( x[0] * ( x[1] ** -1.0) ), axis = 1 ) ** 0.5\n",
    "#harmonic mean of score pct\n",
    "score_spread['DR_Pct'] = DR_spread[['DR_Pct1', 'DR_Pct2']].apply(lambda x : 0.5*( x[0] ** -1.0 ) + 0.5*( 1.0 - x[1] ) ** -1.0, axis = 1 ) ** -1.0\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 479,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "team1d_PF_spread = train_test_inner.groupby(['Season', 'TeamID1'])[['PF_Ratio', 'PF_Pct']]\\\n",
    ".agg({ 'PF_Ratio': geo_mean, 'PF_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'PF_Ratio' : 'PF_Ratio1', 'PF_Pct' : 'PF_Pct1'})\n",
    "\n",
    "team2d_PF_spread = train_test_inner.groupby(['Season', 'TeamID2'])[['PF_Ratio', 'PF_Pct']]\\\n",
    ".agg({ 'PF_Ratio': geo_mean, 'PF_Pct' : harm_mean}).reset_index()\\\n",
    ".set_index('Season').rename(columns = {'PF_Ratio' : 'PF_Ratio2', 'PF_Pct' : 'PF_Pct2'})\n",
    "\n",
    "\n",
    "PF_spread = team1d_PF_spread.join(team2d_PF_spread).reset_index()\n",
    "\n",
    "#geometric mean of score ratio of team 1 and inverse of team 2\n",
    "score_spread['PF_Ratio'] = PF_spread[['PF_Ratio1', 'PF_Ratio2']].apply(lambda x : ( x[0] * ( x[1] ** -1.0) ), axis = 1 ) ** 0.5\n",
    "\n",
    "#harmonic mean of score pct\n",
    "score_spread['PF_Pct'] = PF_spread[['PF_Pct1', 'PF_Pct2']].apply(lambda x : 0.5*( x[0] ** -1.0 ) + 0.5*( 1.0 - x[1] ) ** -1.0, axis = 1 ) ** -1.0\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 480,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Season          0\n",
       "TeamID1         0\n",
       "Score_Ratio1    0\n",
       "Score_Pct1      0\n",
       "TeamID2         0\n",
       "Score_Ratio2    0\n",
       "Score_Pct2      0\n",
       "Score_Ratio     0\n",
       "Score_Pct       0\n",
       "Ast_Ratio       0\n",
       "Ast_Pct         0\n",
       "DR_Ratio        0\n",
       "DR_Pct          0\n",
       "PF_Ratio        0\n",
       "PF_Pct          0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 480,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "score_spread.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 481,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['DayNum', 'DiffSeed', 'NumOT', 'Result', 'Season', 'Team1_Ast',\n",
       "       'Team1_DR', 'Team1_PF', 'Team1_Score', 'Team2_Ast', 'Team2_DR',\n",
       "       'Team2_PF', 'Team2_Score', 'TeamID1', 'TeamID2', 'Score_Ratio',\n",
       "       'Score_Total', 'Score_Pct', 'DR_Ratio', 'DR_Total', 'DR_Pct',\n",
       "       'Ast_Ratio', 'Ast_Total', 'Ast_Pct', 'PF_Ratio', 'PF_Total', 'PF_Pct'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 481,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_inner.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 482,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Season         0\n",
       "NumOT          0\n",
       "Score_Ratio    0\n",
       "Score_Pct      0\n",
       "Ast_Ratio      0\n",
       "Ast_Pct        0\n",
       "DR_Ratio       0\n",
       "DR_Pct         0\n",
       "PF_Ratio       0\n",
       "PF_Pct         0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 482,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train = train_test_inner.loc[:, ['Season', 'NumOT', 'Score_Ratio', 'Score_Pct', 'Ast_Ratio',\n",
    "       'Ast_Pct', 'DR_Ratio', 'DR_Pct', 'PF_Ratio', 'PF_Pct']] \n",
    "train_labels = train_test_inner['Result']\n",
    "X_train.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 483,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_outer = pd.merge( train.loc[ train['Season'].isin(years), : ].reset_index(drop = True), \n",
    "         df_sample_sub.drop(['ID', 'Pred'], axis = 1), \n",
    "         on = ['Season', 'TeamID1', 'TeamID2'], how = 'outer' )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 484,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DayNum         9043\n",
       "DiffSeed       9043\n",
       "NumOT          9043\n",
       "Result         9043\n",
       "Season            0\n",
       "Team1_Ast      9043\n",
       "Team1_DR       9043\n",
       "Team1_PF       9043\n",
       "Team1_Score    9043\n",
       "Team2_Ast      9043\n",
       "Team2_DR       9043\n",
       "Team2_PF       9043\n",
       "Team2_Score    9043\n",
       "TeamID1           0\n",
       "TeamID2           0\n",
       "Score_Ratio    9043\n",
       "Score_Total    9043\n",
       "Score_Pct      9043\n",
       "DR_Ratio       9043\n",
       "DR_Total       9043\n",
       "DR_Pct         9043\n",
       "Ast_Ratio      9043\n",
       "Ast_Total      9043\n",
       "Ast_Pct        9043\n",
       "PF_Ratio       9043\n",
       "PF_Total       9043\n",
       "PF_Pct         9043\n",
       "dtype: int64"
      ]
     },
     "execution_count": 484,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_outer.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 485,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_outer = train_test_outer.loc[ train_test_outer['Result'].isnull(), \n",
    "                                        ['TeamID1', 'TeamID2', 'Season']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 486,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TeamID1    0\n",
       "TeamID2    0\n",
       "Season     0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 486,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_outer.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 487,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_missing = pd.merge( pd.merge(score_spread.loc[:, ['TeamID1', 'TeamID2', 'Season', 'Score_Ratio', 'Score_Pct', 'Ast_Ratio',\n",
    "       'Ast_Pct', 'DR_Ratio', 'DR_Pct', 'PF_Ratio', 'PF_Pct']], \n",
    "                   train_test_outer, on = ['TeamID1', 'TeamID2', 'Season']),\n",
    "         num_ot.loc[:, ['TeamID1', 'TeamID2', 'Season', 'NumOT']],\n",
    "         on = ['TeamID1', 'TeamID2', 'Season'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 488,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "TeamID1        0\n",
       "TeamID2        0\n",
       "Season         0\n",
       "Score_Ratio    0\n",
       "Score_Pct      0\n",
       "Ast_Ratio      0\n",
       "Ast_Pct        0\n",
       "DR_Ratio       0\n",
       "DR_Pct         0\n",
       "PF_Ratio       0\n",
       "PF_Pct         0\n",
       "NumOT          0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 488,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_missing.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 489,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>TeamID1</th>\n",
       "      <th>TeamID2</th>\n",
       "      <th>Season</th>\n",
       "      <th>Score_Ratio</th>\n",
       "      <th>Score_Pct</th>\n",
       "      <th>Ast_Ratio</th>\n",
       "      <th>Ast_Pct</th>\n",
       "      <th>DR_Ratio</th>\n",
       "      <th>DR_Pct</th>\n",
       "      <th>PF_Ratio</th>\n",
       "      <th>PF_Pct</th>\n",
       "      <th>NumOT</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3103</td>\n",
       "      <td>3143</td>\n",
       "      <td>2014</td>\n",
       "      <td>0.937046</td>\n",
       "      <td>0.484342</td>\n",
       "      <td>0.889416</td>\n",
       "      <td>0.447515</td>\n",
       "      <td>1.000711</td>\n",
       "      <td>0.495040</td>\n",
       "      <td>1.057927</td>\n",
       "      <td>0.523009</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3103</td>\n",
       "      <td>3163</td>\n",
       "      <td>2014</td>\n",
       "      <td>1.114925</td>\n",
       "      <td>0.520949</td>\n",
       "      <td>1.340258</td>\n",
       "      <td>0.540817</td>\n",
       "      <td>1.058948</td>\n",
       "      <td>0.507087</td>\n",
       "      <td>0.895009</td>\n",
       "      <td>0.468967</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3103</td>\n",
       "      <td>3169</td>\n",
       "      <td>2014</td>\n",
       "      <td>0.836696</td>\n",
       "      <td>0.455329</td>\n",
       "      <td>0.914889</td>\n",
       "      <td>0.448542</td>\n",
       "      <td>0.682533</td>\n",
       "      <td>0.393040</td>\n",
       "      <td>0.985273</td>\n",
       "      <td>0.494603</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3103</td>\n",
       "      <td>3181</td>\n",
       "      <td>2014</td>\n",
       "      <td>0.985393</td>\n",
       "      <td>0.499427</td>\n",
       "      <td>1.061666</td>\n",
       "      <td>0.500333</td>\n",
       "      <td>1.044097</td>\n",
       "      <td>0.510465</td>\n",
       "      <td>0.958931</td>\n",
       "      <td>0.491312</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>3103</td>\n",
       "      <td>3199</td>\n",
       "      <td>2014</td>\n",
       "      <td>0.935506</td>\n",
       "      <td>0.483282</td>\n",
       "      <td>0.982900</td>\n",
       "      <td>0.473371</td>\n",
       "      <td>1.036946</td>\n",
       "      <td>0.503235</td>\n",
       "      <td>1.135509</td>\n",
       "      <td>0.531591</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   TeamID1  TeamID2  Season  Score_Ratio  Score_Pct  Ast_Ratio   Ast_Pct  \\\n",
       "0     3103     3143    2014     0.937046   0.484342   0.889416  0.447515   \n",
       "1     3103     3163    2014     1.114925   0.520949   1.340258  0.540817   \n",
       "2     3103     3169    2014     0.836696   0.455329   0.914889  0.448542   \n",
       "3     3103     3181    2014     0.985393   0.499427   1.061666  0.500333   \n",
       "4     3103     3199    2014     0.935506   0.483282   0.982900  0.473371   \n",
       "\n",
       "   DR_Ratio    DR_Pct  PF_Ratio    PF_Pct  NumOT  \n",
       "0  1.000711  0.495040  1.057927  0.523009      0  \n",
       "1  1.058948  0.507087  0.895009  0.468967      0  \n",
       "2  0.682533  0.393040  0.985273  0.494603      0  \n",
       "3  1.044097  0.510465  0.958931  0.491312      0  \n",
       "4  1.036946  0.503235  1.135509  0.531591      0  "
      ]
     },
     "execution_count": 489,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_missing.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 490,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#We scale our data for our logistic regression, and make sure our categorical variables are properly processed.\n",
    "\n",
    "X_test = train_test_missing.loc[:, ['Season', 'NumOT', 'Score_Ratio', 'Score_Pct', 'Ast_Ratio', 'DR_Ratio', 'PF_Ratio', 'Ast_Pct', 'DR_Pct', 'PF_Pct']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 491,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Season</th>\n",
       "      <th>NumOT</th>\n",
       "      <th>Score_Ratio</th>\n",
       "      <th>Score_Pct</th>\n",
       "      <th>Ast_Ratio</th>\n",
       "      <th>DR_Ratio</th>\n",
       "      <th>PF_Ratio</th>\n",
       "      <th>Ast_Pct</th>\n",
       "      <th>DR_Pct</th>\n",
       "      <th>PF_Pct</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>2014</td>\n",
       "      <td>0</td>\n",
       "      <td>0.937046</td>\n",
       "      <td>0.484342</td>\n",
       "      <td>0.889416</td>\n",
       "      <td>1.000711</td>\n",
       "      <td>1.057927</td>\n",
       "      <td>0.447515</td>\n",
       "      <td>0.495040</td>\n",
       "      <td>0.523009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2014</td>\n",
       "      <td>0</td>\n",
       "      <td>1.114925</td>\n",
       "      <td>0.520949</td>\n",
       "      <td>1.340258</td>\n",
       "      <td>1.058948</td>\n",
       "      <td>0.895009</td>\n",
       "      <td>0.540817</td>\n",
       "      <td>0.507087</td>\n",
       "      <td>0.468967</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2014</td>\n",
       "      <td>0</td>\n",
       "      <td>0.836696</td>\n",
       "      <td>0.455329</td>\n",
       "      <td>0.914889</td>\n",
       "      <td>0.682533</td>\n",
       "      <td>0.985273</td>\n",
       "      <td>0.448542</td>\n",
       "      <td>0.393040</td>\n",
       "      <td>0.494603</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2014</td>\n",
       "      <td>0</td>\n",
       "      <td>0.985393</td>\n",
       "      <td>0.499427</td>\n",
       "      <td>1.061666</td>\n",
       "      <td>1.044097</td>\n",
       "      <td>0.958931</td>\n",
       "      <td>0.500333</td>\n",
       "      <td>0.510465</td>\n",
       "      <td>0.491312</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>2014</td>\n",
       "      <td>0</td>\n",
       "      <td>0.935506</td>\n",
       "      <td>0.483282</td>\n",
       "      <td>0.982900</td>\n",
       "      <td>1.036946</td>\n",
       "      <td>1.135509</td>\n",
       "      <td>0.473371</td>\n",
       "      <td>0.503235</td>\n",
       "      <td>0.531591</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Season  NumOT  Score_Ratio  Score_Pct  Ast_Ratio  DR_Ratio  PF_Ratio  \\\n",
       "0    2014      0     0.937046   0.484342   0.889416  1.000711  1.057927   \n",
       "1    2014      0     1.114925   0.520949   1.340258  1.058948  0.895009   \n",
       "2    2014      0     0.836696   0.455329   0.914889  0.682533  0.985273   \n",
       "3    2014      0     0.985393   0.499427   1.061666  1.044097  0.958931   \n",
       "4    2014      0     0.935506   0.483282   0.982900  1.036946  1.135509   \n",
       "\n",
       "    Ast_Pct    DR_Pct    PF_Pct  \n",
       "0  0.447515  0.495040  0.523009  \n",
       "1  0.540817  0.507087  0.468967  \n",
       "2  0.448542  0.393040  0.494603  \n",
       "3  0.500333  0.510465  0.491312  \n",
       "4  0.473371  0.503235  0.531591  "
      ]
     },
     "execution_count": 491,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 492,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "n = X_train.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 493,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_merge = pd.concat( [X_train, X_test], axis = 0 ).reset_index(drop = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 494,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_merge = pd.concat( [pd.get_dummies( train_test_merge['Season'].astype(object) ), \n",
    "            train_test_merge.drop('Season', axis = 1) ], axis = 1 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 495,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_merge = pd.concat( [pd.get_dummies( train_test_merge['NumOT'].astype(object) ), \n",
    "            train_test_merge.drop('NumOT', axis = 1) ], axis = 1 )\n",
    "\n",
    "X_train = train_test_merge.loc[:(n - 1), :].reset_index(drop = True)\n",
    "X_test = train_test_merge.loc[n:, :].reset_index(drop = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 496,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0              0\n",
       "1              0\n",
       "2              0\n",
       "3              0\n",
       "4              0\n",
       "2014           0\n",
       "2015           0\n",
       "2016           0\n",
       "2017           0\n",
       "2018           0\n",
       "Ast_Pct        0\n",
       "Ast_Ratio      0\n",
       "DR_Pct         0\n",
       "DR_Ratio       0\n",
       "PF_Pct         0\n",
       "PF_Ratio       0\n",
       "Score_Pct      0\n",
       "Score_Ratio    0\n",
       "dtype: int64"
      ]
     },
     "execution_count": 496,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_test_merge.isnull().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 497,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "x_max = X_train.max()\n",
    "x_min = X_train.min()\n",
    "\n",
    "X_train = ( X_train - x_min ) / ( x_max - x_min + 1e-14)\n",
    "X_test = ( X_test - x_min ) / ( x_max - x_min + 1e-14)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 498,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(1380, 18)\n",
      "     0    1    2    3    4  2014  2015  2016  2017  2018   Ast_Pct  Ast_Ratio  \\\n",
      "0  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.565777   0.114526   \n",
      "1  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.416327   0.066122   \n",
      "2  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.595918   0.127697   \n",
      "3  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.560653   0.112430   \n",
      "4  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.522449   0.097959   \n",
      "\n",
      "     DR_Pct  DR_Ratio    PF_Pct  PF_Ratio  Score_Pct  Score_Ratio  \n",
      "0  0.740524  0.436087  0.521062  0.249917   0.618894     0.369421  \n",
      "1  0.482878  0.201931  0.413239  0.177417   0.574964     0.327960  \n",
      "2  0.535398  0.237955  0.442943  0.195827   0.541489     0.298757  \n",
      "3  0.575221  0.268437  0.762301  0.495495   0.613417     0.364042  \n",
      "4  0.604078  0.292501  0.436374  0.191662   0.587111     0.339051  \n",
      "(1380,)\n",
      "0    1.0\n",
      "1    1.0\n",
      "2    1.0\n",
      "3    1.0\n",
      "4    1.0\n",
      "Name: Result, dtype: float64\n",
      "0.0    693\n",
      "1.0    687\n",
      "Name: Result, dtype: int64\n",
      "(7579, 18)\n",
      "     0    1    2    3    4  2014  2015  2016  2017  2018   Ast_Pct  Ast_Ratio  \\\n",
      "0  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.420162   0.074882   \n",
      "1  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.536694   0.119046   \n",
      "2  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.421444   0.077377   \n",
      "3  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.486131   0.091755   \n",
      "4  1.0  0.0  0.0  0.0  0.0   1.0   0.0   0.0   0.0   0.0  0.452456   0.084039   \n",
      "\n",
      "     DR_Pct  DR_Ratio    PF_Pct  PF_Ratio  Score_Pct  Score_Ratio  \n",
      "0  0.433260  0.177186  0.539026  0.250714   0.471098     0.242259  \n",
      "1  0.455648  0.193163  0.436785  0.195919   0.548708     0.316329  \n",
      "2  0.243703  0.089898  0.485284  0.226278   0.409589     0.200472  \n",
      "3  0.461926  0.189089  0.479058  0.217418   0.503080     0.262391  \n",
      "4  0.448490  0.187127  0.555263  0.276808   0.468850     0.241617  \n"
     ]
    }
   ],
   "source": [
    "print(X_train.shape)\n",
    "print(X_train.head())\n",
    "\n",
    "print(train_labels.shape)\n",
    "print(train_labels.head())\n",
    "print(train_labels.value_counts())\n",
    "\n",
    "print(X_test.shape)\n",
    "print(X_test.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 499,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LogisticRegressionCV(Cs=8, class_weight=None, cv=5, dual=False,\n",
       "           fit_intercept=True, intercept_scaling=1.0, max_iter=100,\n",
       "           multi_class='ovr', n_jobs=4, penalty='l2', random_state=None,\n",
       "           refit=True, scoring='neg_log_loss', solver='lbfgs', tol=0.0001,\n",
       "           verbose=0)"
      ]
     },
     "execution_count": 499,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LogisticRegressionCV\n",
    "\n",
    "log_clf = LogisticRegressionCV(cv = 5,Cs=8,n_jobs=4,scoring=\"neg_log_loss\")\n",
    "\n",
    "log_clf.fit( X_train, train_labels )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 500,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[-0.69319232 -0.69384506 -0.69426906 -0.65223014 -0.5802668  -0.3883429\n",
      "  -0.18465013 -0.04041747]\n",
      " [-0.69377894 -0.70137892 -0.74050641 -0.67751376 -0.56304702 -0.42552432\n",
      "  -0.28420844 -0.22979473]\n",
      " [-0.69301696 -0.69152739 -0.66928477 -0.57173751 -0.42656562 -0.26914993\n",
      "  -0.18042533 -0.05072779]\n",
      " [-0.69217085 -0.68015817 -0.57593277 -0.36652923 -0.19334575 -0.10206284\n",
      "  -0.06793893 -0.07886935]\n",
      " [-0.69201616 -0.67810423 -0.55805588 -0.35199471 -0.21179919 -0.08065747\n",
      "  -0.03913599 -0.0389182 ]]\n"
     ]
    }
   ],
   "source": [
    "print(log_clf.scores_[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 501,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAD8CAYAAACfF6SlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x164ce8884e0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.plot(log_clf.scores_[1])\n",
    "# plt.ylabel('some numbers')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 502,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_inner['Pred1'] = log_clf.predict_proba(X_train)[:,1]\n",
    "train_test_missing['Pred1'] = log_clf.predict_proba(X_test)[:,1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 503,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sub = pd.merge(df_sample_sub, \n",
    "                         pd.concat( [train_test_missing.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']],\n",
    "                                     train_test_inner.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']] ],\n",
    "                                   axis = 0).reset_index(drop = True),\n",
    "                  on = ['Season', 'TeamID1', 'TeamID2'], how = 'outer')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 504,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10423, 6)\n",
      "               ID  Pred  Season  TeamID1  TeamID2  Pred1\n",
      "0  2014_3103_3107   0.5    2014     3103     3107    NaN\n",
      "1  2014_3103_3113   0.5    2014     3103     3113    NaN\n",
      "2  2014_3103_3119   0.5    2014     3103     3119    NaN\n",
      "3  2014_3103_3124   0.5    2014     3103     3124    NaN\n",
      "4  2014_3103_3140   0.5    2014     3103     3140    NaN\n"
     ]
    }
   ],
   "source": [
    "print(sub.shape)\n",
    "print(sub.head())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 505,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#We get the 'average' probability of success for each team\n",
    "\n",
    "team1_probs = sub.groupby('TeamID1')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "team2_probs = sub.groupby('TeamID2')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "#Any missing value for the prediciton will be imputed with the product of the probabilities calculated above. We assume these are independent events.\n",
    "\n",
    "sub['Pred'] = sub[['TeamID1', 'TeamID2','Pred1']]\\\n",
    ".apply(lambda x : team1_probs.get(x[0]) * ( 1 - team2_probs.get(x[1]) ) if np.isnan(x[2]) else x[2], \n",
    "       axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 506,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2016, 2)\n"
     ]
    }
   ],
   "source": [
    "sub.ID.value_counts()\n",
    "sub=sub.groupby('ID', as_index=False).agg({\"Pred\": \"mean\"})\n",
    "sub.ID.value_counts()\n",
    "sub2018=sub.loc[sub['ID'].isin(df_sample_sub2.ID)]\n",
    "print(sub2018.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 507,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ID</th>\n",
       "      <th>Pred</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>8064</th>\n",
       "      <td>2018_3110_3113</td>\n",
       "      <td>9.348854e-43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8065</th>\n",
       "      <td>2018_3110_3114</td>\n",
       "      <td>9.348854e-43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8066</th>\n",
       "      <td>2018_3110_3124</td>\n",
       "      <td>9.348854e-43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8067</th>\n",
       "      <td>2018_3110_3125</td>\n",
       "      <td>9.348854e-43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8068</th>\n",
       "      <td>2018_3110_3129</td>\n",
       "      <td>9.348854e-43</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  ID          Pred\n",
       "8064  2018_3110_3113  9.348854e-43\n",
       "8065  2018_3110_3114  9.348854e-43\n",
       "8066  2018_3110_3124  9.348854e-43\n",
       "8067  2018_3110_3125  9.348854e-43\n",
       "8068  2018_3110_3129  9.348854e-43"
      ]
     },
     "execution_count": 507,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 508,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.36850240102023485"
      ]
     },
     "execution_count": 508,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018.Pred.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 509,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    2.016000e+03\n",
       "mean     3.685024e-01\n",
       "std      4.511544e-01\n",
       "min      2.331184e-50\n",
       "25%      6.869387e-12\n",
       "50%      5.945273e-04\n",
       "75%      9.984556e-01\n",
       "max      1.000000e+00\n",
       "Name: Pred, dtype: float64"
      ]
     },
     "execution_count": 509,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018.Pred.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 510,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x164cdd3c6d8>"
      ]
     },
     "execution_count": 510,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAD8CAYAAABgmUMCAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nkqReDA5JUi//H9r9qH35kzHeAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x164cddd8710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sub2018['Pred'].plot.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 511,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#sub2018[['ID', 'Pred']].to_csv('sub_2018_all_only18.csv', index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 512,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import lightgbm as lgb\n",
    "d_train = lgb.Dataset(X_train, label=train_labels)\n",
    "\n",
    "params = {}\n",
    "params['learning_rate'] = 0.003\n",
    "params['boosting_type'] = 'gbdt'\n",
    "params['objective'] = 'binary'\n",
    "params['metric'] = 'binary_logloss'\n",
    "params['sub_feature'] = 0.5\n",
    "params['num_leaves'] = 10\n",
    "params['min_data'] = 50\n",
    "params['max_depth'] = 10\n",
    "clf = lgb.train(params, d_train, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 513,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_inner['Pred1'] = clf.predict(X_train)\n",
    "train_test_missing['Pred1'] = clf.predict(X_test)\n",
    "\n",
    "sub2 = pd.merge(df_sample_sub, pd.concat( [train_test_missing.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']],\n",
    "                                     train_test_inner.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']] ],\n",
    "                                   axis = 0).reset_index(drop = True),\n",
    "                  on = ['Season', 'TeamID1', 'TeamID2'], how = 'outer')\n",
    "\n",
    "#We get the 'average' probability of success for each team\n",
    "\n",
    "team1_probs = sub2.groupby('TeamID1')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "team2_probs = sub2.groupby('TeamID2')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "#Any missing value for the prediciton will be imputed with the product of the probabilities calculated above. We assume these are independent events.\n",
    "\n",
    "sub['Pred'] = sub2[['TeamID1', 'TeamID2','Pred1']]\\\n",
    ".apply(lambda x : team1_probs.get(x[0]) * ( 1 - team2_probs.get(x[1]) ) if np.isnan(x[2]) else x[2], \n",
    "       axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 514,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2016, 2)\n"
     ]
    }
   ],
   "source": [
    "sub.ID.value_counts()\n",
    "sub=sub.groupby('ID', as_index=False).agg({\"Pred\": \"mean\"})\n",
    "sub.ID.value_counts()\n",
    "sub2018_lgb=sub.loc[sub['ID'].isin(df_sample_sub2.ID)]\n",
    "print(sub2018.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 515,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style>\n",
       "    .dataframe thead tr:only-child th {\n",
       "        text-align: right;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: left;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ID</th>\n",
       "      <th>Pred</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>8064</th>\n",
       "      <td>2018_3110_3113</td>\n",
       "      <td>0.377013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8065</th>\n",
       "      <td>2018_3110_3114</td>\n",
       "      <td>0.420593</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8066</th>\n",
       "      <td>2018_3110_3124</td>\n",
       "      <td>0.396654</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8067</th>\n",
       "      <td>2018_3110_3125</td>\n",
       "      <td>0.396582</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8068</th>\n",
       "      <td>2018_3110_3129</td>\n",
       "      <td>0.386071</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  ID      Pred\n",
       "8064  2018_3110_3113  0.377013\n",
       "8065  2018_3110_3114  0.420593\n",
       "8066  2018_3110_3124  0.396654\n",
       "8067  2018_3110_3125  0.396582\n",
       "8068  2018_3110_3129  0.386071"
      ]
     },
     "execution_count": 515,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018_lgb.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 516,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.4484592580025936"
      ]
     },
     "execution_count": 516,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018_lgb.Pred.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 517,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "#sub2018_lgb[['ID', 'Pred']].to_csv('sub_2018_lgb.csv', index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 518,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x164cee4fa58>"
      ]
     },
     "execution_count": 518,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAD8CAYAAABthzNFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nJHUyICRJnQwISVInA0KS1On/A4EJRMoM53VRAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x164cddd8d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sub2018_lgb['Pred'].plot.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 519,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "count    2016.000000\n",
       "mean        0.448459\n",
       "std         0.125140\n",
       "min         0.176626\n",
       "25%         0.376549\n",
       "50%         0.397124\n",
       "75%         0.585225\n",
       "max         0.627403\n",
       "Name: Pred, dtype: float64"
      ]
     },
     "execution_count": 519,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018_lgb.Pred.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 520,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "params = {}\n",
    "params['learning_rate'] = 0.003\n",
    "params['boosting_type'] = 'gbdt'\n",
    "params['objective'] = 'binary'\n",
    "params['metric'] = 'binary_logloss'\n",
    "params['sub_feature'] = 0.5\n",
    "params['num_leaves'] = 50\n",
    "params['min_data'] = 200\n",
    "params['max_depth'] = 50\n",
    "clf = lgb.train(params, d_train, 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 521,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_test_inner['Pred1'] = clf.predict(X_train)\n",
    "train_test_missing['Pred1'] = clf.predict(X_test)\n",
    "\n",
    "sub = pd.merge(df_sample_sub, pd.concat( [train_test_missing.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']],\n",
    "                                     train_test_inner.loc[:, ['Season', 'TeamID1', 'TeamID2', 'Pred1']] ],\n",
    "                                   axis = 0).reset_index(drop = True),\n",
    "                  on = ['Season', 'TeamID1', 'TeamID2'], how = 'outer')\n",
    "\n",
    "#We get the 'average' probability of success for each team\n",
    "\n",
    "team1_probs = sub.groupby('TeamID1')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "team2_probs = sub.groupby('TeamID2')['Pred1'].apply(lambda x : (x ** -1.0).mean() ** -1.0 ).fillna(0.5).to_dict()\n",
    "#Any missing value for the prediciton will be imputed with the product of the probabilities calculated above. We assume these are independent events.\n",
    "\n",
    "sub['Pred'] = sub[['TeamID1', 'TeamID2','Pred1']]\\\n",
    ".apply(lambda x : team1_probs.get(x[0]) * ( 1 - team2_probs.get(x[1]) ) if np.isnan(x[2]) else x[2], \n",
    "       axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 522,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2016, 2)\n"
     ]
    }
   ],
   "source": [
    "sub.ID.value_counts()\n",
    "sub=sub.groupby('ID', as_index=False).agg({\"Pred\": \"mean\"})\n",
    "sub.ID.value_counts()\n",
    "sub2018_lgb3=sub.loc[sub['ID'].isin(df_sample_sub2.ID)]\n",
    "print(sub2018.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 523,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.4508362721958463"
      ]
     },
     "execution_count": 523,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sub2018_lgb3.Pred.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 524,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x164cee32748>"
      ]
     },
     "execution_count": 524,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAD8CAYAAABthzNFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nugwISVKXASFJ6jIgJEldBoQkqev/A9q1QhMZHuUFAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x164cee6a048>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sub2018_lgb3['Pred'].plot.hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 526,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "sub2018_lgb[['ID', 'Pred']].to_csv('sub_2018_lgb3.csv', index = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 528,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting binarytree\n",
      "  Downloading binarytree-3.0.1.tar.gz\n",
      "Building wheels for collected packages: binarytree\n",
      "  Running setup.py bdist_wheel for binarytree: started\n",
      "  Running setup.py bdist_wheel for binarytree: finished with status 'done'\n",
      "  Stored in directory: C:\\Users\\MissFlave\\AppData\\Local\\pip\\Cache\\wheels\\6f\\af\\aa\\27c4aa57282254c29bf429bd6a54ef30b70a41fd1e11a00539\n",
      "Successfully built binarytree\n",
      "Installing collected packages: binarytree\n",
      "Successfully installed binarytree-3.0.1\n",
      "Collecting bracketeer\n",
      "  Downloading bracketeer-0.1.7.tar.gz\n",
      "Requirement already satisfied: pandas in c:\\users\\missflave\\anaconda\\lib\\site-packages (from bracketeer)\n",
      "Requirement already satisfied: binarytree in c:\\users\\missflave\\anaconda\\lib\\site-packages (from bracketeer)\n",
      "Requirement already satisfied: matplotlib in c:\\users\\missflave\\anaconda\\lib\\site-packages (from bracketeer)\n",
      "Requirement already satisfied: Pillow in c:\\users\\missflave\\anaconda\\lib\\site-packages (from bracketeer)\n",
      "Requirement already satisfied: python-dateutil>=2 in c:\\users\\missflave\\anaconda\\lib\\site-packages (from pandas->bracketeer)\n",
      "Requirement already satisfied: pytz>=2011k in c:\\users\\missflave\\anaconda\\lib\\site-packages (from pandas->bracketeer)\n",
      "Requirement already satisfied: numpy>=1.7.0 in c:\\users\\missflave\\anaconda\\lib\\site-packages (from pandas->bracketeer)\n",
      "Requirement already satisfied: six>=1.10 in c:\\users\\missflave\\anaconda\\lib\\site-packages (from matplotlib->bracketeer)\n",
      "Requirement already satisfied: cycler>=0.10 in c:\\users\\missflave\\anaconda\\lib\\site-packages (from matplotlib->bracketeer)\n",
      "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in c:\\users\\missflave\\anaconda\\lib\\site-packages (from matplotlib->bracketeer)\n",
      "Requirement already satisfied: olefile in c:\\users\\missflave\\anaconda\\lib\\site-packages (from Pillow->bracketeer)\n",
      "Building wheels for collected packages: bracketeer\n",
      "  Running setup.py bdist_wheel for bracketeer: started\n",
      "  Running setup.py bdist_wheel for bracketeer: finished with status 'done'\n",
      "  Stored in directory: C:\\Users\\MissFlave\\AppData\\Local\\pip\\Cache\\wheels\\b3\\83\\ad\\71c4245864d3b61f2f93df4620912261d76148d5eeeae526e3\n",
      "Successfully built bracketeer\n",
      "Installing collected packages: bracketeer\n",
      "Successfully installed bracketeer-0.1.7\n"
     ]
    }
   ],
   "source": [
    "!pip install binarytree\n",
    "!pip install bracketeer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 536,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'season'",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2441\u001b[0m             \u001b[1;32mtry\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2442\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2443\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'season'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[1;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-536-e27f02f0fd24>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      5\u001b[0m         \u001b[0mseedsPath\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'./WStage2DataFiles/WNCAATourneySeeds.csv'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      6\u001b[0m         \u001b[0msubmissionPath\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'sub_2018_lgb3.csv'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 7\u001b[1;33m         \u001b[0mslotsPath\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'./WStage2DataFiles/WNCAATourneySlots.csv'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      8\u001b[0m )\n\u001b[0;32m      9\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\bracketeer\\__init__.py\u001b[0m in \u001b[0;36mbuild_bracket\u001b[1;34m(outputPath, teamsPath, seedsPath, slotsPath, submissionPath, year)\u001b[0m\n\u001b[0;32m     78\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     79\u001b[0m     \u001b[1;31m# Create bracket tree from slot data.  Create seed to slot mapping as well.\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 80\u001b[1;33m     \u001b[0ms\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mslots_df\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mslots_df\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'season'\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m==\u001b[0m \u001b[0myear\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     81\u001b[0m     \u001b[0mseed_slot_map\u001b[0m \u001b[1;33m=\u001b[0m \u001b[1;33m{\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m \u001b[1;34m'R6CH'\u001b[0m\u001b[1;33m}\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     82\u001b[0m     \u001b[0mbkt\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mextNode\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   1962\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1963\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1964\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1965\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1966\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_getitem_column\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m_getitem_column\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   1969\u001b[0m         \u001b[1;31m# get column\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1970\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mis_unique\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1971\u001b[1;33m             \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_item_cache\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1972\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1973\u001b[0m         \u001b[1;31m# duplicate columns & possible reduce dimensionality\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\generic.py\u001b[0m in \u001b[0;36m_get_item_cache\u001b[1;34m(self, item)\u001b[0m\n\u001b[0;32m   1643\u001b[0m         \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mcache\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1644\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mres\u001b[0m \u001b[1;32mis\u001b[0m \u001b[1;32mNone\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 1645\u001b[1;33m             \u001b[0mvalues\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_data\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   1646\u001b[0m             \u001b[0mres\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_box_item_values\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mvalues\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   1647\u001b[0m             \u001b[0mcache\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m]\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mres\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\internals.py\u001b[0m in \u001b[0;36mget\u001b[1;34m(self, item, fastpath)\u001b[0m\n\u001b[0;32m   3588\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3589\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0misnull\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 3590\u001b[1;33m                 \u001b[0mloc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mitem\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   3591\u001b[0m             \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   3592\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0marange\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0misnull\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda\\lib\\site-packages\\pandas\\core\\indexes\\base.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method, tolerance)\u001b[0m\n\u001b[0;32m   2442\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2443\u001b[0m             \u001b[1;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2444\u001b[1;33m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2445\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2446\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;32mpandas\\_libs\\hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[1;34m()\u001b[0m\n",
      "\u001b[1;31mKeyError\u001b[0m: 'season'"
     ]
    }
   ],
   "source": [
    "from bracketeer import build_bracket\n",
    "b = build_bracket(\n",
    "        outputPath='output.png',\n",
    "        teamsPath='./WStage2DataFiles/WTeams.csv',\n",
    "        seedsPath='./WStage2DataFiles/WNCAATourneySeeds.csv',\n",
    "        submissionPath='sub_2018_lgb3.csv',\n",
    "        slotsPath='./WStage2DataFiles/WNCAATourneySlots.csv',\n",
    ")\n",
    "\n",
    "from IPython.display import Image\n",
    "Image(\"output.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}