NFL Rush.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# NFL Game Win "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:36:45.500063Z",
     "start_time": "2020-01-06T16:36:39.758065Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "C:\\Users\\Solomonzhs\\Anaconda3\\lib\\site-packages\\tqdm\\std.py:648: FutureWarning: The Panel class is removed from pandas. Accessing it from the top-level namespace will also be removed in the next version\n",
      "  from pandas import Panel\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import matplotlib as mpl\n",
    "import datetime\n",
    "#from kaggle.competitions import nflrush\n",
    "from tqdm import tqdm\n",
    "from sklearn.ensemble import RandomForestRegressor\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import os\n",
    "import re\n",
    "import keras\n",
    "from sklearn.model_selection import KFold, RepeatedKFold\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "from sklearn.isotonic import IsotonicRegression\n",
    "from keras.utils import Sequence\n",
    "tqdm.pandas()\n",
    "\n",
    "from numba import jit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:37:32.588822Z",
     "start_time": "2020-01-06T16:37:32.578986Z"
    }
   },
   "outputs": [],
   "source": [
    "import random as rn\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "from keras import backend as K\n",
    "\n",
    "def init_seeds(seed):\n",
    "\n",
    "    # The below is necessary for starting Numpy generated random numbers\n",
    "    # in a well-defined initial state.\n",
    "\n",
    "    np.random.seed(seed)\n",
    "\n",
    "    # The below is necessary for starting core Python generated random numbers\n",
    "    # in a well-defined state.\n",
    "\n",
    "    rn.seed(seed)\n",
    "\n",
    "    # Force TensorFlow to use single thread.\n",
    "    # Multiple threads are a potential source of\n",
    "    # non-reproducible results.\n",
    "    # For further details, see: https://stackoverflow.com/questions/42022950/which-seeds-have-to-be-set-where-to-realize-100-reproducibility-of-training-res\n",
    "\n",
    "\n",
    "\n",
    "    # The below tf.set_random_seed() will make random number generation\n",
    "    # in the TensorFlow backend have a well-defined initial state.\n",
    "    # For further details, see: https://www.tensorflow.org/api_docs/python/tf/set_random_seed\n",
    "\n",
    "    tf.set_random_seed(seed)\n",
    "\n",
    "    sess = tf.Session(graph=tf.get_default_graph())\n",
    "    K.set_session(sess)\n",
    "    return sess"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:40:21.194932Z",
     "start_time": "2020-01-06T16:40:18.110851Z"
    }
   },
   "outputs": [],
   "source": [
    "train = pd.read_csv(r\"C:\\Users\\Solomonzhs\\Desktop\\Kaggle\\nfl-rush\\train.csv\", dtype={'WindSpeed': 'object'})\n",
    "train.shape\n",
    "results = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:40:29.816336Z",
     "start_time": "2020-01-06T16:40:29.811332Z"
    }
   },
   "outputs": [],
   "source": [
    "player_cols = ['X', 'Y', 'X_dir', 'Y_dir', 'X_S', 'Y_S', 'S', 'A', 'IsRusher', 'IsOnOffense']\n",
    "#player_cols = ['X', 'Y', 'X_dir', 'Y_dir', 'X_S', 'Y_S', 'S', 'A', 'Energy', 'IsRusher', 'IsOnOffense']\n",
    "play_cols = ['X_rusher', 'Y_rusher', 'YardLine_std']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:43:34.670219Z",
     "start_time": "2020-01-06T16:43:34.638861Z"
    }
   },
   "outputs": [],
   "source": [
    "def reorient(train, flip_left):\n",
    "    train['ToLeft'] = train.PlayDirection == \"left\"\n",
    "    #train['IsBallCarrier'] = train.NflId == train.NflIdRusher\n",
    "    \n",
    "    train.loc[train.VisitorTeamAbbr == \"ARI\", 'VisitorTeamAbbr'] = \"ARZ\"\n",
    "    train.loc[train.HomeTeamAbbr == \"ARI\", 'HomeTeamAbbr'] = \"ARZ\"\n",
    "    \n",
    "    train.loc[train.VisitorTeamAbbr == \"BAL\", 'VisitorTeamAbbr'] = \"BLT\"\n",
    "    train.loc[train.HomeTeamAbbr == \"BAL\", 'HomeTeamAbbr'] = \"BLT\"\n",
    "    \n",
    "    train.loc[train.VisitorTeamAbbr == \"CLE\", 'VisitorTeamAbbr'] = \"CLV\"\n",
    "    train.loc[train.HomeTeamAbbr == \"CLE\", 'HomeTeamAbbr'] = \"CLV\"\n",
    "    \n",
    "    train.loc[train.VisitorTeamAbbr == \"HOU\", 'VisitorTeamAbbr'] = \"HST\"\n",
    "    train.loc[train.HomeTeamAbbr == \"HOU\", 'HomeTeamAbbr'] = \"HST\"\n",
    "    \n",
    "    train['TeamOnOffense'] = \"home\"\n",
    "    train.loc[train.PossessionTeam != train.HomeTeamAbbr, 'TeamOnOffense'] = \"away\"\n",
    "    train['IsOnOffense'] = train.Team == train.TeamOnOffense # Is player on offense?\n",
    "    train['YardLine_std'] = 100 - train.YardLine\n",
    "    train.loc[train.FieldPosition.fillna('') == train.PossessionTeam,  \n",
    "          'YardLine_std'\n",
    "         ] = train.loc[train.FieldPosition.fillna('') == train.PossessionTeam,  \n",
    "          'YardLine']\n",
    "    train.loc[train.ToLeft, 'X'] = 120 - train.loc[train.ToLeft, 'X'] \n",
    "    train.loc[train.ToLeft, 'Y'] = 160/3 - train.loc[train.ToLeft, 'Y'] \n",
    "    train.loc[train.ToLeft, 'Orientation'] = np.mod(180 + train.loc[train.ToLeft, 'Orientation'], 360)\n",
    "    train['Dir'] = 90 - train.Dir\n",
    "    train.loc[train.ToLeft, 'Dir'] = np.mod(180 + train.loc[train.ToLeft, 'Dir'], 360)\n",
    "    train.loc[train.IsOnOffense, 'Dir'] = train.loc[train.IsOnOffense, 'Dir'].fillna(0).values\n",
    "    train.loc[~train.IsOnOffense, 'Dir'] = train.loc[~train.IsOnOffense, 'Dir'].fillna(180).values\n",
    "    \n",
    "    \n",
    "    train['IsRusher'] = train['NflId'] == train['NflIdRusher']\n",
    "    if flip_left:\n",
    "        df = train[train['IsRusher']].copy()\n",
    "        #df['left'] = df.Y < 160/6\n",
    "        df['left'] = df.Dir < 0\n",
    "        train = train.merge(df[['PlayId', 'left']], how='left', on='PlayId')\n",
    "        train['Y'] = train.Y\n",
    "        train.loc[train.left, 'Y'] = 160/3 - train.loc[train.left, 'Y']\n",
    "        train['Dir'] = train.Dir\n",
    "        train.loc[train.left, 'Dir'] = np.mod( - train.loc[train.left, 'Dir'], 360)\n",
    "        train.drop('left', axis=1, inplace=True)\n",
    "        \n",
    "    \n",
    "    train['X_dir'] = np.cos( (np.pi / 180) * train .Dir)\n",
    "    train['Y_dir'] = np.sin( (np.pi / 180) * train.Dir)\n",
    "    train['X_S'] = train.X_dir * train.S\n",
    "    train['Y_S'] = train.Y_dir * train.S\n",
    "    train['X_A'] = train.X_dir * train.A\n",
    "    train['Y_A'] = train.Y_dir * train.A\n",
    "    \n",
    "    train['PlayerHeight'] = train['PlayerHeight'].apply(lambda x: 12*int(x.split('-')[0])+int(x.split('-')[1]))\n",
    "    train['BMI'] = (train['PlayerWeight'] * 703) / (train['PlayerHeight'] ** 2)\n",
    "    train['Energy'] = train['PlayerWeight'] * (train['S'] ** 2) \n",
    "    train.loc[train['Season'] == 2017, 'S'] = (train['S'][train['Season'] == 2017] - 2.4355) / 1.2930 * 1.4551 + 2.7570\n",
    "    train['time_step'] = 0.0\n",
    "    train = train.sort_values(by=['PlayId', 'IsOnOffense', 'IsRusher', 'Y']).reset_index(drop=True)\n",
    "    return  train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:43:43.064734Z",
     "start_time": "2020-01-06T16:43:43.058733Z"
    }
   },
   "outputs": [],
   "source": [
    "def add_features(train):\n",
    "        \n",
    "    df = train[train.IsRusher][['PlayId', 'time_step', 'X', 'Y']].copy()\n",
    "    df.columns = ['PlayId', 'time_step', 'X_rusher', 'Y_rusher']\n",
    "    train = train.merge(df, how='left', on=['PlayId', 'time_step'])\n",
    "    #train.loc[~train.IsRusher, 'X'] = train.loc[~train.IsRusher, 'X'] - train.loc[~train.IsRusher, 'X_rusher']\n",
    "    #train.loc[~train.IsRusher, 'Y'] = train.loc[~train.IsRusher, 'Y'] - train.loc[~train.IsRusher, 'Y_rusher']\n",
    "    train.X -= train.X_rusher\n",
    "    train.Y -= train.Y_rusher\n",
    "    \n",
    "    #train.drop(['Orientation', 'Dir', 'TeamOnOffense', 'YardLine', 'left'], axis=1, inplace=True)\n",
    "    return  train\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:43:53.709250Z",
     "start_time": "2020-01-06T16:43:53.697288Z"
    }
   },
   "outputs": [],
   "source": [
    "def time_forward(train, time_step):\n",
    "    train = train.copy()\n",
    "    train['X'] = train['X'] + time_step * train['X_S'] + 0.5 * time_step**2 * train['X_A']\n",
    "    train['Y'] = train['Y'] + time_step * train['Y_S'] + 0.5 * time_step**2 * train['Y_A']\n",
    "    x_s = np.clip(train['X_S'] + time_step * train['X_A'], -14, 14)\n",
    "    y_s = np.clip(train['Y_S'] + time_step * train['Y_A'], -14, 14)\n",
    "    if 1:\n",
    "        train['X_A'] = np.clip(train['X_A'] * train['X_S'] / (x_s), -10, 10)\n",
    "        train['Y_A'] = np.clip(train['Y_A'] * train['Y_S'] / (y_s), -10, 10)\n",
    "    train['X_S'] = x_s\n",
    "    train['Y_S'] = y_s\n",
    "    train['time_step'] = time_step\n",
    "    return train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:44:05.637666Z",
     "start_time": "2020-01-06T16:43:59.687101Z"
    }
   },
   "outputs": [],
   "source": [
    "train_reorient = reorient(train.copy(), flip_left=True)\n",
    "len_train_reorient = len(train_reorient) // 22"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:44:11.619808Z",
     "start_time": "2020-01-06T16:44:11.358806Z"
    }
   },
   "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>GameId</th>\n",
       "      <th>PlayId</th>\n",
       "      <th>Team</th>\n",
       "      <th>X</th>\n",
       "      <th>Y</th>\n",
       "      <th>S</th>\n",
       "      <th>A</th>\n",
       "      <th>Dis</th>\n",
       "      <th>Orientation</th>\n",
       "      <th>Dir</th>\n",
       "      <th>...</th>\n",
       "      <th>IsRusher</th>\n",
       "      <th>X_dir</th>\n",
       "      <th>Y_dir</th>\n",
       "      <th>X_S</th>\n",
       "      <th>Y_S</th>\n",
       "      <th>X_A</th>\n",
       "      <th>Y_A</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Energy</th>\n",
       "      <th>time_step</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>46.65</td>\n",
       "      <td>14.503333</td>\n",
       "      <td>5.136589</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.51</td>\n",
       "      <td>255.47</td>\n",
       "      <td>79.16</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>0.188067</td>\n",
       "      <td>0.982156</td>\n",
       "      <td>0.855705</td>\n",
       "      <td>4.468811</td>\n",
       "      <td>0.142931</td>\n",
       "      <td>0.746439</td>\n",
       "      <td>26.715085</td>\n",
       "      <td>4078.3925</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>50.68</td>\n",
       "      <td>17.913333</td>\n",
       "      <td>2.064336</td>\n",
       "      <td>2.43</td>\n",
       "      <td>0.16</td>\n",
       "      <td>192.63</td>\n",
       "      <td>105.69</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>-0.270432</td>\n",
       "      <td>0.962739</td>\n",
       "      <td>-0.492187</td>\n",
       "      <td>1.752185</td>\n",
       "      <td>-0.657151</td>\n",
       "      <td>2.339456</td>\n",
       "      <td>27.935571</td>\n",
       "      <td>682.3544</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>46.09</td>\n",
       "      <td>18.493333</td>\n",
       "      <td>1.918039</td>\n",
       "      <td>1.13</td>\n",
       "      <td>0.40</td>\n",
       "      <td>261.99</td>\n",
       "      <td>92.82</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>-0.049198</td>\n",
       "      <td>0.998789</td>\n",
       "      <td>-0.083145</td>\n",
       "      <td>1.687953</td>\n",
       "      <td>-0.055594</td>\n",
       "      <td>1.128632</td>\n",
       "      <td>28.749228</td>\n",
       "      <td>605.4932</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>46.00</td>\n",
       "      <td>20.133333</td>\n",
       "      <td>1.389116</td>\n",
       "      <td>0.59</td>\n",
       "      <td>0.31</td>\n",
       "      <td>183.01</td>\n",
       "      <td>67.27</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>0.386389</td>\n",
       "      <td>0.922336</td>\n",
       "      <td>0.471395</td>\n",
       "      <td>1.125250</td>\n",
       "      <td>0.227970</td>\n",
       "      <td>0.544178</td>\n",
       "      <td>33.744000</td>\n",
       "      <td>401.8680</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>45.33</td>\n",
       "      <td>20.693333</td>\n",
       "      <td>0.488822</td>\n",
       "      <td>1.35</td>\n",
       "      <td>0.01</td>\n",
       "      <td>207.61</td>\n",
       "      <td>71.30</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>0.320613</td>\n",
       "      <td>0.947210</td>\n",
       "      <td>0.134657</td>\n",
       "      <td>0.397828</td>\n",
       "      <td>0.432828</td>\n",
       "      <td>1.278734</td>\n",
       "      <td>35.993600</td>\n",
       "      <td>50.8032</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>509757</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>84.57</td>\n",
       "      <td>28.963333</td>\n",
       "      <td>4.360000</td>\n",
       "      <td>1.79</td>\n",
       "      <td>0.47</td>\n",
       "      <td>148.08</td>\n",
       "      <td>93.34</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>-0.058261</td>\n",
       "      <td>0.998301</td>\n",
       "      <td>-0.254018</td>\n",
       "      <td>4.352594</td>\n",
       "      <td>-0.104287</td>\n",
       "      <td>1.786959</td>\n",
       "      <td>36.975674</td>\n",
       "      <td>6083.0720</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>509758</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>85.68</td>\n",
       "      <td>29.023333</td>\n",
       "      <td>1.040000</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.10</td>\n",
       "      <td>43.66</td>\n",
       "      <td>236.53</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>-0.551500</td>\n",
       "      <td>-0.834175</td>\n",
       "      <td>-0.573560</td>\n",
       "      <td>-0.867542</td>\n",
       "      <td>-0.193025</td>\n",
       "      <td>-0.291961</td>\n",
       "      <td>38.385135</td>\n",
       "      <td>323.3984</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>509759</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>86.77</td>\n",
       "      <td>29.133333</td>\n",
       "      <td>2.140000</td>\n",
       "      <td>2.12</td>\n",
       "      <td>0.22</td>\n",
       "      <td>44.55</td>\n",
       "      <td>285.31</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>0.264041</td>\n",
       "      <td>-0.964511</td>\n",
       "      <td>0.565049</td>\n",
       "      <td>-2.064054</td>\n",
       "      <td>0.559768</td>\n",
       "      <td>-2.044764</td>\n",
       "      <td>30.158284</td>\n",
       "      <td>1195.2756</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>509760</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>86.76</td>\n",
       "      <td>31.313333</td>\n",
       "      <td>2.580000</td>\n",
       "      <td>0.74</td>\n",
       "      <td>0.26</td>\n",
       "      <td>65.37</td>\n",
       "      <td>311.26</td>\n",
       "      <td>...</td>\n",
       "      <td>False</td>\n",
       "      <td>0.659477</td>\n",
       "      <td>-0.751725</td>\n",
       "      <td>1.701451</td>\n",
       "      <td>-1.939450</td>\n",
       "      <td>0.488013</td>\n",
       "      <td>-0.556276</td>\n",
       "      <td>30.427632</td>\n",
       "      <td>1664.1000</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>509761</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>80.80</td>\n",
       "      <td>26.983333</td>\n",
       "      <td>4.870000</td>\n",
       "      <td>4.10</td>\n",
       "      <td>0.45</td>\n",
       "      <td>135.44</td>\n",
       "      <td>28.24</td>\n",
       "      <td>...</td>\n",
       "      <td>True</td>\n",
       "      <td>0.880973</td>\n",
       "      <td>0.473166</td>\n",
       "      <td>4.290340</td>\n",
       "      <td>2.304318</td>\n",
       "      <td>3.611991</td>\n",
       "      <td>1.939980</td>\n",
       "      <td>30.959333</td>\n",
       "      <td>5265.1518</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>509762 rows × 63 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            GameId          PlayId  Team      X          Y         S     A  \\\n",
       "0       2017090700  20170907000118  away  46.65  14.503333  5.136589  0.76   \n",
       "1       2017090700  20170907000118  away  50.68  17.913333  2.064336  2.43   \n",
       "2       2017090700  20170907000118  away  46.09  18.493333  1.918039  1.13   \n",
       "3       2017090700  20170907000118  away  46.00  20.133333  1.389116  0.59   \n",
       "4       2017090700  20170907000118  away  45.33  20.693333  0.488822  1.35   \n",
       "...            ...             ...   ...    ...        ...       ...   ...   \n",
       "509757  2018123015  20181230154157  home  84.57  28.963333  4.360000  1.79   \n",
       "509758  2018123015  20181230154157  home  85.68  29.023333  1.040000  0.35   \n",
       "509759  2018123015  20181230154157  home  86.77  29.133333  2.140000  2.12   \n",
       "509760  2018123015  20181230154157  home  86.76  31.313333  2.580000  0.74   \n",
       "509761  2018123015  20181230154157  home  80.80  26.983333  4.870000  4.10   \n",
       "\n",
       "         Dis  Orientation     Dir  ...  IsRusher     X_dir     Y_dir  \\\n",
       "0       0.51       255.47   79.16  ...     False  0.188067  0.982156   \n",
       "1       0.16       192.63  105.69  ...     False -0.270432  0.962739   \n",
       "2       0.40       261.99   92.82  ...     False -0.049198  0.998789   \n",
       "3       0.31       183.01   67.27  ...     False  0.386389  0.922336   \n",
       "4       0.01       207.61   71.30  ...     False  0.320613  0.947210   \n",
       "...      ...          ...     ...  ...       ...       ...       ...   \n",
       "509757  0.47       148.08   93.34  ...     False -0.058261  0.998301   \n",
       "509758  0.10        43.66  236.53  ...     False -0.551500 -0.834175   \n",
       "509759  0.22        44.55  285.31  ...     False  0.264041 -0.964511   \n",
       "509760  0.26        65.37  311.26  ...     False  0.659477 -0.751725   \n",
       "509761  0.45       135.44   28.24  ...      True  0.880973  0.473166   \n",
       "\n",
       "             X_S       Y_S       X_A       Y_A        BMI     Energy  \\\n",
       "0       0.855705  4.468811  0.142931  0.746439  26.715085  4078.3925   \n",
       "1      -0.492187  1.752185 -0.657151  2.339456  27.935571   682.3544   \n",
       "2      -0.083145  1.687953 -0.055594  1.128632  28.749228   605.4932   \n",
       "3       0.471395  1.125250  0.227970  0.544178  33.744000   401.8680   \n",
       "4       0.134657  0.397828  0.432828  1.278734  35.993600    50.8032   \n",
       "...          ...       ...       ...       ...        ...        ...   \n",
       "509757 -0.254018  4.352594 -0.104287  1.786959  36.975674  6083.0720   \n",
       "509758 -0.573560 -0.867542 -0.193025 -0.291961  38.385135   323.3984   \n",
       "509759  0.565049 -2.064054  0.559768 -2.044764  30.158284  1195.2756   \n",
       "509760  1.701451 -1.939450  0.488013 -0.556276  30.427632  1664.1000   \n",
       "509761  4.290340  2.304318  3.611991  1.939980  30.959333  5265.1518   \n",
       "\n",
       "        time_step  \n",
       "0             0.0  \n",
       "1             0.0  \n",
       "2             0.0  \n",
       "3             0.0  \n",
       "4             0.0  \n",
       "...           ...  \n",
       "509757        0.0  \n",
       "509758        0.0  \n",
       "509759        0.0  \n",
       "509760        0.0  \n",
       "509761        0.0  \n",
       "\n",
       "[509762 rows x 63 columns]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_reorient"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:46:10.756035Z",
     "start_time": "2020-01-06T16:46:10.751067Z"
    }
   },
   "outputs": [],
   "source": [
    "def augment_data(train_reorient, time_steps):\n",
    "    train_reorient = pd.concat(time_forward(train_reorient, time_step) for time_step in time_steps)\n",
    "    train_reorient = add_features(train_reorient)\n",
    "    return train_reorient"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:46:29.707007Z",
     "start_time": "2020-01-06T16:46:19.500719Z"
    }
   },
   "outputs": [],
   "source": [
    "time_steps = [0, 0.3, 0.6]\n",
    "train_reorient = augment_data(train_reorient, time_steps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:46:55.635788Z",
     "start_time": "2020-01-06T16:46:54.840243Z"
    }
   },
   "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>GameId</th>\n",
       "      <th>PlayId</th>\n",
       "      <th>Team</th>\n",
       "      <th>X</th>\n",
       "      <th>Y</th>\n",
       "      <th>S</th>\n",
       "      <th>A</th>\n",
       "      <th>Dis</th>\n",
       "      <th>Orientation</th>\n",
       "      <th>Dir</th>\n",
       "      <th>...</th>\n",
       "      <th>Y_dir</th>\n",
       "      <th>X_S</th>\n",
       "      <th>Y_S</th>\n",
       "      <th>X_A</th>\n",
       "      <th>Y_A</th>\n",
       "      <th>BMI</th>\n",
       "      <th>Energy</th>\n",
       "      <th>time_step</th>\n",
       "      <th>X_rusher</th>\n",
       "      <th>Y_rusher</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>5.400000</td>\n",
       "      <td>-8.300000</td>\n",
       "      <td>5.136589</td>\n",
       "      <td>0.76</td>\n",
       "      <td>0.51</td>\n",
       "      <td>255.47</td>\n",
       "      <td>79.16</td>\n",
       "      <td>...</td>\n",
       "      <td>0.982156</td>\n",
       "      <td>0.855705</td>\n",
       "      <td>4.468811</td>\n",
       "      <td>0.142931</td>\n",
       "      <td>0.746439</td>\n",
       "      <td>26.715085</td>\n",
       "      <td>4078.3925</td>\n",
       "      <td>0.0</td>\n",
       "      <td>41.250000</td>\n",
       "      <td>22.803333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>9.430000</td>\n",
       "      <td>-4.890000</td>\n",
       "      <td>2.064336</td>\n",
       "      <td>2.43</td>\n",
       "      <td>0.16</td>\n",
       "      <td>192.63</td>\n",
       "      <td>105.69</td>\n",
       "      <td>...</td>\n",
       "      <td>0.962739</td>\n",
       "      <td>-0.492187</td>\n",
       "      <td>1.752185</td>\n",
       "      <td>-0.657151</td>\n",
       "      <td>2.339456</td>\n",
       "      <td>27.935571</td>\n",
       "      <td>682.3544</td>\n",
       "      <td>0.0</td>\n",
       "      <td>41.250000</td>\n",
       "      <td>22.803333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>4.840000</td>\n",
       "      <td>-4.310000</td>\n",
       "      <td>1.918039</td>\n",
       "      <td>1.13</td>\n",
       "      <td>0.40</td>\n",
       "      <td>261.99</td>\n",
       "      <td>92.82</td>\n",
       "      <td>...</td>\n",
       "      <td>0.998789</td>\n",
       "      <td>-0.083145</td>\n",
       "      <td>1.687953</td>\n",
       "      <td>-0.055594</td>\n",
       "      <td>1.128632</td>\n",
       "      <td>28.749228</td>\n",
       "      <td>605.4932</td>\n",
       "      <td>0.0</td>\n",
       "      <td>41.250000</td>\n",
       "      <td>22.803333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>4.750000</td>\n",
       "      <td>-2.670000</td>\n",
       "      <td>1.389116</td>\n",
       "      <td>0.59</td>\n",
       "      <td>0.31</td>\n",
       "      <td>183.01</td>\n",
       "      <td>67.27</td>\n",
       "      <td>...</td>\n",
       "      <td>0.922336</td>\n",
       "      <td>0.471395</td>\n",
       "      <td>1.125250</td>\n",
       "      <td>0.227970</td>\n",
       "      <td>0.544178</td>\n",
       "      <td>33.744000</td>\n",
       "      <td>401.8680</td>\n",
       "      <td>0.0</td>\n",
       "      <td>41.250000</td>\n",
       "      <td>22.803333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>2017090700</td>\n",
       "      <td>20170907000118</td>\n",
       "      <td>away</td>\n",
       "      <td>4.080000</td>\n",
       "      <td>-2.110000</td>\n",
       "      <td>0.488822</td>\n",
       "      <td>1.35</td>\n",
       "      <td>0.01</td>\n",
       "      <td>207.61</td>\n",
       "      <td>71.30</td>\n",
       "      <td>...</td>\n",
       "      <td>0.947210</td>\n",
       "      <td>0.134657</td>\n",
       "      <td>0.397828</td>\n",
       "      <td>0.432828</td>\n",
       "      <td>1.278734</td>\n",
       "      <td>35.993600</td>\n",
       "      <td>50.8032</td>\n",
       "      <td>0.0</td>\n",
       "      <td>41.250000</td>\n",
       "      <td>22.803333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1529281</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>0.374455</td>\n",
       "      <td>3.181422</td>\n",
       "      <td>4.360000</td>\n",
       "      <td>1.79</td>\n",
       "      <td>0.47</td>\n",
       "      <td>148.08</td>\n",
       "      <td>93.34</td>\n",
       "      <td>...</td>\n",
       "      <td>0.998301</td>\n",
       "      <td>-0.316590</td>\n",
       "      <td>5.424770</td>\n",
       "      <td>-0.083675</td>\n",
       "      <td>1.433777</td>\n",
       "      <td>36.975674</td>\n",
       "      <td>6083.0720</td>\n",
       "      <td>0.6</td>\n",
       "      <td>84.024362</td>\n",
       "      <td>28.715121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1529282</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>1.276757</td>\n",
       "      <td>-0.264865</td>\n",
       "      <td>1.040000</td>\n",
       "      <td>0.35</td>\n",
       "      <td>0.10</td>\n",
       "      <td>43.66</td>\n",
       "      <td>236.53</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.834175</td>\n",
       "      <td>-0.689375</td>\n",
       "      <td>-1.042718</td>\n",
       "      <td>-0.160597</td>\n",
       "      <td>-0.242912</td>\n",
       "      <td>38.385135</td>\n",
       "      <td>323.3984</td>\n",
       "      <td>0.6</td>\n",
       "      <td>84.024362</td>\n",
       "      <td>28.715121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1529283</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>3.185425</td>\n",
       "      <td>-1.188277</td>\n",
       "      <td>2.140000</td>\n",
       "      <td>2.12</td>\n",
       "      <td>0.22</td>\n",
       "      <td>44.55</td>\n",
       "      <td>285.31</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.964511</td>\n",
       "      <td>0.900909</td>\n",
       "      <td>-3.290913</td>\n",
       "      <td>0.351085</td>\n",
       "      <td>-1.282472</td>\n",
       "      <td>30.158284</td>\n",
       "      <td>1195.2756</td>\n",
       "      <td>0.6</td>\n",
       "      <td>84.024362</td>\n",
       "      <td>28.715121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1529284</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>3.844350</td>\n",
       "      <td>1.334413</td>\n",
       "      <td>2.580000</td>\n",
       "      <td>0.74</td>\n",
       "      <td>0.26</td>\n",
       "      <td>65.37</td>\n",
       "      <td>311.26</td>\n",
       "      <td>...</td>\n",
       "      <td>-0.751725</td>\n",
       "      <td>1.994259</td>\n",
       "      <td>-2.273216</td>\n",
       "      <td>0.416360</td>\n",
       "      <td>-0.474601</td>\n",
       "      <td>30.427632</td>\n",
       "      <td>1664.1000</td>\n",
       "      <td>0.6</td>\n",
       "      <td>84.024362</td>\n",
       "      <td>28.715121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1529285</td>\n",
       "      <td>2018123015</td>\n",
       "      <td>20181230154157</td>\n",
       "      <td>home</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>4.870000</td>\n",
       "      <td>4.10</td>\n",
       "      <td>0.45</td>\n",
       "      <td>135.44</td>\n",
       "      <td>28.24</td>\n",
       "      <td>...</td>\n",
       "      <td>0.473166</td>\n",
       "      <td>6.457535</td>\n",
       "      <td>3.468306</td>\n",
       "      <td>2.399781</td>\n",
       "      <td>1.288909</td>\n",
       "      <td>30.959333</td>\n",
       "      <td>5265.1518</td>\n",
       "      <td>0.6</td>\n",
       "      <td>84.024362</td>\n",
       "      <td>28.715121</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1529286 rows × 65 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             GameId          PlayId  Team         X         Y         S     A  \\\n",
       "0        2017090700  20170907000118  away  5.400000 -8.300000  5.136589  0.76   \n",
       "1        2017090700  20170907000118  away  9.430000 -4.890000  2.064336  2.43   \n",
       "2        2017090700  20170907000118  away  4.840000 -4.310000  1.918039  1.13   \n",
       "3        2017090700  20170907000118  away  4.750000 -2.670000  1.389116  0.59   \n",
       "4        2017090700  20170907000118  away  4.080000 -2.110000  0.488822  1.35   \n",
       "...             ...             ...   ...       ...       ...       ...   ...   \n",
       "1529281  2018123015  20181230154157  home  0.374455  3.181422  4.360000  1.79   \n",
       "1529282  2018123015  20181230154157  home  1.276757 -0.264865  1.040000  0.35   \n",
       "1529283  2018123015  20181230154157  home  3.185425 -1.188277  2.140000  2.12   \n",
       "1529284  2018123015  20181230154157  home  3.844350  1.334413  2.580000  0.74   \n",
       "1529285  2018123015  20181230154157  home  0.000000  0.000000  4.870000  4.10   \n",
       "\n",
       "          Dis  Orientation     Dir  ...     Y_dir       X_S       Y_S  \\\n",
       "0        0.51       255.47   79.16  ...  0.982156  0.855705  4.468811   \n",
       "1        0.16       192.63  105.69  ...  0.962739 -0.492187  1.752185   \n",
       "2        0.40       261.99   92.82  ...  0.998789 -0.083145  1.687953   \n",
       "3        0.31       183.01   67.27  ...  0.922336  0.471395  1.125250   \n",
       "4        0.01       207.61   71.30  ...  0.947210  0.134657  0.397828   \n",
       "...       ...          ...     ...  ...       ...       ...       ...   \n",
       "1529281  0.47       148.08   93.34  ...  0.998301 -0.316590  5.424770   \n",
       "1529282  0.10        43.66  236.53  ... -0.834175 -0.689375 -1.042718   \n",
       "1529283  0.22        44.55  285.31  ... -0.964511  0.900909 -3.290913   \n",
       "1529284  0.26        65.37  311.26  ... -0.751725  1.994259 -2.273216   \n",
       "1529285  0.45       135.44   28.24  ...  0.473166  6.457535  3.468306   \n",
       "\n",
       "              X_A       Y_A        BMI     Energy time_step   X_rusher  \\\n",
       "0        0.142931  0.746439  26.715085  4078.3925       0.0  41.250000   \n",
       "1       -0.657151  2.339456  27.935571   682.3544       0.0  41.250000   \n",
       "2       -0.055594  1.128632  28.749228   605.4932       0.0  41.250000   \n",
       "3        0.227970  0.544178  33.744000   401.8680       0.0  41.250000   \n",
       "4        0.432828  1.278734  35.993600    50.8032       0.0  41.250000   \n",
       "...           ...       ...        ...        ...       ...        ...   \n",
       "1529281 -0.083675  1.433777  36.975674  6083.0720       0.6  84.024362   \n",
       "1529282 -0.160597 -0.242912  38.385135   323.3984       0.6  84.024362   \n",
       "1529283  0.351085 -1.282472  30.158284  1195.2756       0.6  84.024362   \n",
       "1529284  0.416360 -0.474601  30.427632  1664.1000       0.6  84.024362   \n",
       "1529285  2.399781  1.288909  30.959333  5265.1518       0.6  84.024362   \n",
       "\n",
       "          Y_rusher  \n",
       "0        22.803333  \n",
       "1        22.803333  \n",
       "2        22.803333  \n",
       "3        22.803333  \n",
       "4        22.803333  \n",
       "...            ...  \n",
       "1529281  28.715121  \n",
       "1529282  28.715121  \n",
       "1529283  28.715121  \n",
       "1529284  28.715121  \n",
       "1529285  28.715121  \n",
       "\n",
       "[1529286 rows x 65 columns]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_reorient"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:47:18.694708Z",
     "start_time": "2020-01-06T16:47:17.640642Z"
    }
   },
   "outputs": [],
   "source": [
    "df = train_reorient[player_cols]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:47:21.787909Z",
     "start_time": "2020-01-06T16:47:21.781943Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1529286, 10)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:50:15.481828Z",
     "start_time": "2020-01-06T16:50:15.454311Z"
    }
   },
   "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>X</th>\n",
       "      <th>Y</th>\n",
       "      <th>X_dir</th>\n",
       "      <th>Y_dir</th>\n",
       "      <th>X_S</th>\n",
       "      <th>Y_S</th>\n",
       "      <th>S</th>\n",
       "      <th>A</th>\n",
       "      <th>IsRusher</th>\n",
       "      <th>IsOnOffense</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>5.40</td>\n",
       "      <td>-8.30</td>\n",
       "      <td>0.188067</td>\n",
       "      <td>0.982156</td>\n",
       "      <td>0.855705</td>\n",
       "      <td>4.468811</td>\n",
       "      <td>5.136589</td>\n",
       "      <td>0.76</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>9.43</td>\n",
       "      <td>-4.89</td>\n",
       "      <td>-0.270432</td>\n",
       "      <td>0.962739</td>\n",
       "      <td>-0.492187</td>\n",
       "      <td>1.752185</td>\n",
       "      <td>2.064336</td>\n",
       "      <td>2.43</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>4.84</td>\n",
       "      <td>-4.31</td>\n",
       "      <td>-0.049198</td>\n",
       "      <td>0.998789</td>\n",
       "      <td>-0.083145</td>\n",
       "      <td>1.687953</td>\n",
       "      <td>1.918039</td>\n",
       "      <td>1.13</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>4.75</td>\n",
       "      <td>-2.67</td>\n",
       "      <td>0.386389</td>\n",
       "      <td>0.922336</td>\n",
       "      <td>0.471395</td>\n",
       "      <td>1.125250</td>\n",
       "      <td>1.389116</td>\n",
       "      <td>0.59</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>4.08</td>\n",
       "      <td>-2.11</td>\n",
       "      <td>0.320613</td>\n",
       "      <td>0.947210</td>\n",
       "      <td>0.134657</td>\n",
       "      <td>0.397828</td>\n",
       "      <td>0.488822</td>\n",
       "      <td>1.35</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>4.60</td>\n",
       "      <td>1.63</td>\n",
       "      <td>0.997391</td>\n",
       "      <td>-0.072194</td>\n",
       "      <td>0.718121</td>\n",
       "      <td>-0.051980</td>\n",
       "      <td>0.826432</td>\n",
       "      <td>0.73</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>7.29</td>\n",
       "      <td>2.83</td>\n",
       "      <td>-0.962975</td>\n",
       "      <td>0.269592</td>\n",
       "      <td>-0.404449</td>\n",
       "      <td>0.113229</td>\n",
       "      <td>0.488822</td>\n",
       "      <td>0.54</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>22.12</td>\n",
       "      <td>3.63</td>\n",
       "      <td>-0.822243</td>\n",
       "      <td>-0.569136</td>\n",
       "      <td>-0.213783</td>\n",
       "      <td>-0.147975</td>\n",
       "      <td>0.308763</td>\n",
       "      <td>1.86</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>3.69</td>\n",
       "      <td>6.53</td>\n",
       "      <td>-0.996179</td>\n",
       "      <td>0.087330</td>\n",
       "      <td>-1.006141</td>\n",
       "      <td>0.088203</td>\n",
       "      <td>1.152789</td>\n",
       "      <td>0.32</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>5.38</td>\n",
       "      <td>11.80</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>-0.000698</td>\n",
       "      <td>1.240000</td>\n",
       "      <td>-0.000866</td>\n",
       "      <td>1.411623</td>\n",
       "      <td>0.74</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>4.64</td>\n",
       "      <td>13.89</td>\n",
       "      <td>0.607514</td>\n",
       "      <td>-0.794309</td>\n",
       "      <td>0.674341</td>\n",
       "      <td>-0.881683</td>\n",
       "      <td>1.265326</td>\n",
       "      <td>0.83</td>\n",
       "      <td>False</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>2.28</td>\n",
       "      <td>-6.38</td>\n",
       "      <td>0.260336</td>\n",
       "      <td>0.965518</td>\n",
       "      <td>1.340730</td>\n",
       "      <td>4.972418</td>\n",
       "      <td>5.811810</td>\n",
       "      <td>0.77</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>3.97</td>\n",
       "      <td>-2.68</td>\n",
       "      <td>0.586372</td>\n",
       "      <td>0.810042</td>\n",
       "      <td>1.002697</td>\n",
       "      <td>1.385171</td>\n",
       "      <td>1.940546</td>\n",
       "      <td>0.82</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>3.32</td>\n",
       "      <td>-1.88</td>\n",
       "      <td>0.678929</td>\n",
       "      <td>0.734204</td>\n",
       "      <td>1.018393</td>\n",
       "      <td>1.101306</td>\n",
       "      <td>1.704219</td>\n",
       "      <td>1.36</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>4.15</td>\n",
       "      <td>-1.35</td>\n",
       "      <td>0.998906</td>\n",
       "      <td>0.046758</td>\n",
       "      <td>1.857966</td>\n",
       "      <td>0.086969</td>\n",
       "      <td>2.109351</td>\n",
       "      <td>1.51</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>-1.01</td>\n",
       "      <td>1.04</td>\n",
       "      <td>-0.934453</td>\n",
       "      <td>0.356086</td>\n",
       "      <td>-0.784941</td>\n",
       "      <td>0.299112</td>\n",
       "      <td>0.961476</td>\n",
       "      <td>1.22</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>4.17</td>\n",
       "      <td>1.41</td>\n",
       "      <td>0.883357</td>\n",
       "      <td>0.468701</td>\n",
       "      <td>0.883357</td>\n",
       "      <td>0.468701</td>\n",
       "      <td>1.141535</td>\n",
       "      <td>0.69</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>4.17</td>\n",
       "      <td>2.06</td>\n",
       "      <td>0.968886</td>\n",
       "      <td>0.247506</td>\n",
       "      <td>1.647107</td>\n",
       "      <td>0.420761</td>\n",
       "      <td>1.929292</td>\n",
       "      <td>1.55</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>2.85</td>\n",
       "      <td>5.41</td>\n",
       "      <td>0.318132</td>\n",
       "      <td>0.948046</td>\n",
       "      <td>0.439022</td>\n",
       "      <td>1.308304</td>\n",
       "      <td>1.569175</td>\n",
       "      <td>0.80</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>4.05</td>\n",
       "      <td>11.34</td>\n",
       "      <td>0.988964</td>\n",
       "      <td>-0.148155</td>\n",
       "      <td>2.076825</td>\n",
       "      <td>-0.311125</td>\n",
       "      <td>2.379439</td>\n",
       "      <td>1.48</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>2.93</td>\n",
       "      <td>12.97</td>\n",
       "      <td>0.969659</td>\n",
       "      <td>-0.244461</td>\n",
       "      <td>2.230216</td>\n",
       "      <td>-0.562261</td>\n",
       "      <td>2.604513</td>\n",
       "      <td>1.39</td>\n",
       "      <td>False</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.00</td>\n",
       "      <td>0.911690</td>\n",
       "      <td>0.410878</td>\n",
       "      <td>3.309436</td>\n",
       "      <td>1.491487</td>\n",
       "      <td>4.101251</td>\n",
       "      <td>3.35</td>\n",
       "      <td>True</td>\n",
       "      <td>True</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        X      Y     X_dir     Y_dir       X_S       Y_S         S     A  \\\n",
       "0    5.40  -8.30  0.188067  0.982156  0.855705  4.468811  5.136589  0.76   \n",
       "1    9.43  -4.89 -0.270432  0.962739 -0.492187  1.752185  2.064336  2.43   \n",
       "2    4.84  -4.31 -0.049198  0.998789 -0.083145  1.687953  1.918039  1.13   \n",
       "3    4.75  -2.67  0.386389  0.922336  0.471395  1.125250  1.389116  0.59   \n",
       "4    4.08  -2.11  0.320613  0.947210  0.134657  0.397828  0.488822  1.35   \n",
       "5    4.60   1.63  0.997391 -0.072194  0.718121 -0.051980  0.826432  0.73   \n",
       "6    7.29   2.83 -0.962975  0.269592 -0.404449  0.113229  0.488822  0.54   \n",
       "7   22.12   3.63 -0.822243 -0.569136 -0.213783 -0.147975  0.308763  1.86   \n",
       "8    3.69   6.53 -0.996179  0.087330 -1.006141  0.088203  1.152789  0.32   \n",
       "9    5.38  11.80  1.000000 -0.000698  1.240000 -0.000866  1.411623  0.74   \n",
       "10   4.64  13.89  0.607514 -0.794309  0.674341 -0.881683  1.265326  0.83   \n",
       "11   2.28  -6.38  0.260336  0.965518  1.340730  4.972418  5.811810  0.77   \n",
       "12   3.97  -2.68  0.586372  0.810042  1.002697  1.385171  1.940546  0.82   \n",
       "13   3.32  -1.88  0.678929  0.734204  1.018393  1.101306  1.704219  1.36   \n",
       "14   4.15  -1.35  0.998906  0.046758  1.857966  0.086969  2.109351  1.51   \n",
       "15  -1.01   1.04 -0.934453  0.356086 -0.784941  0.299112  0.961476  1.22   \n",
       "16   4.17   1.41  0.883357  0.468701  0.883357  0.468701  1.141535  0.69   \n",
       "17   4.17   2.06  0.968886  0.247506  1.647107  0.420761  1.929292  1.55   \n",
       "18   2.85   5.41  0.318132  0.948046  0.439022  1.308304  1.569175  0.80   \n",
       "19   4.05  11.34  0.988964 -0.148155  2.076825 -0.311125  2.379439  1.48   \n",
       "20   2.93  12.97  0.969659 -0.244461  2.230216 -0.562261  2.604513  1.39   \n",
       "21   0.00   0.00  0.911690  0.410878  3.309436  1.491487  4.101251  3.35   \n",
       "\n",
       "    IsRusher  IsOnOffense  \n",
       "0      False        False  \n",
       "1      False        False  \n",
       "2      False        False  \n",
       "3      False        False  \n",
       "4      False        False  \n",
       "5      False        False  \n",
       "6      False        False  \n",
       "7      False        False  \n",
       "8      False        False  \n",
       "9      False        False  \n",
       "10     False        False  \n",
       "11     False         True  \n",
       "12     False         True  \n",
       "13     False         True  \n",
       "14     False         True  \n",
       "15     False         True  \n",
       "16     False         True  \n",
       "17     False         True  \n",
       "18     False         True  \n",
       "19     False         True  \n",
       "20     False         True  \n",
       "21      True         True  "
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(n=22)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:48:09.760784Z",
     "start_time": "2020-01-06T16:48:09.620790Z"
    }
   },
   "outputs": [],
   "source": [
    "len_2017 = train[train.Season == 2017].shape[0] // 22"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:49:38.786725Z",
     "start_time": "2020-01-06T16:49:38.774189Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11900"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len_2017"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:52:09.405142Z",
     "start_time": "2020-01-06T16:52:09.386139Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 5.4       , -8.3       ,  0.18806703, ...,  0.76      ,\n",
       "         0.        ,  0.        ],\n",
       "       [ 9.43      , -4.89      , -0.2704324 , ...,  2.43      ,\n",
       "         0.        ,  0.        ],\n",
       "       [ 4.84      , -4.31      , -0.04919842, ...,  1.13      ,\n",
       "         0.        ,  0.        ],\n",
       "       ...,\n",
       "       [ 3.185425  , -1.1882774 ,  0.2640414 , ...,  2.12      ,\n",
       "         0.        ,  1.        ],\n",
       "       [ 3.8443503 ,  1.3344132 ,  0.65947706, ...,  0.74      ,\n",
       "         0.        ,  1.        ],\n",
       "       [ 0.        ,  0.        ,  0.88097334, ...,  4.1       ,\n",
       "         1.        ,  1.        ]], dtype=float32)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_players"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:53:00.847331Z",
     "start_time": "2020-01-06T16:52:59.435284Z"
    }
   },
   "outputs": [],
   "source": [
    "ss_players = StandardScaler()\n",
    "X_players = df.values.astype('float32') \n",
    "X_players = ss_players.fit_transform(X_players)\n",
    "X_players = X_players.reshape((-1, 22, len(player_cols)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:53:16.453139Z",
     "start_time": "2020-01-06T16:53:16.446119Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_dmat(X, Y):\n",
    "    dmat = np.zeros((22, 22))\n",
    "    for i in range(22):\n",
    "        for j in range(i+1, 22):\n",
    "            d = np.sqrt((X[i] - X[j])**2 + (Y[i] - Y[j])**2)\n",
    "            dmat[i, j] = d\n",
    "            dmat[j, i] = d\n",
    "    dmat = dmat.reshape((1, 22, 22))\n",
    "    return dmat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:54:24.709916Z",
     "start_time": "2020-01-06T16:53:29.914608Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(69513, 22, 22)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X,Y = train_reorient.X.values, train_reorient.Y.values\n",
    "dmats = [get_dmat(X[i:i+22], Y[i:i+22]) for i in range(0, train_reorient.shape[0], 22)]\n",
    "dmats = np.vstack(dmats).reshape((-1, 22, 22))\n",
    "dmats.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:57:48.898459Z",
     "start_time": "2020-01-06T16:57:47.147978Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "inv_dmats = 1. / (1e-2 + dmats)**2\n",
    "inv_dmats /= inv_dmats.sum(axis=2, keepdims=True)\n",
    "_ = plt.hist(inv_dmats.ravel(), bins=100, log=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:57:48.988505Z",
     "start_time": "2020-01-06T16:57:48.900454Z"
    }
   },
   "outputs": [],
   "source": [
    "X_play = train_reorient[play_cols].values[::22]\n",
    "ss_play = StandardScaler()\n",
    "X_play = ss_play.fit_transform(X_play)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:57:55.347051Z",
     "start_time": "2020-01-06T16:57:55.286566Z"
    }
   },
   "outputs": [],
   "source": [
    "del train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:59:35.099910Z",
     "start_time": "2020-01-06T16:59:35.073255Z"
    }
   },
   "outputs": [],
   "source": [
    "from keras import backend as K\n",
    "\n",
    "\n",
    "class LayerNormalization(keras.layers.Layer):\n",
    "\n",
    "    def __init__(self,\n",
    "                 center=True,\n",
    "                 scale=True,\n",
    "                 epsilon=None,\n",
    "                 gamma_initializer='ones',\n",
    "                 beta_initializer='zeros',\n",
    "                 gamma_regularizer=None,\n",
    "                 beta_regularizer=None,\n",
    "                 gamma_constraint=None,\n",
    "                 beta_constraint=None,\n",
    "                 **kwargs):\n",
    "        \"\"\"Layer normalization layer\n",
    "                See: [Layer Normalization](https://arxiv.org/pdf/1607.06450.pdf)\n",
    "        :param center: Add an offset parameter if it is True.\n",
    "        :param scale: Add a scale parameter if it is True.\n",
    "        :param epsilon: Epsilon for calculating variance.\n",
    "        :param gamma_initializer: Initializer for the gamma weight.\n",
    "        :param beta_initializer: Initializer for the beta weight.\n",
    "        :param gamma_regularizer: Optional regularizer for the gamma weight.\n",
    "        :param beta_regularizer: Optional regularizer for the beta weight.\n",
    "        :param gamma_constraint: Optional constraint for the gamma weight.\n",
    "        :param beta_constraint: Optional constraint for the beta weight.\n",
    "        :param kwargs:\n",
    "        \"\"\"\n",
    "        super(LayerNormalization, self).__init__(**kwargs)\n",
    "        self.supports_masking = True\n",
    "        self.center = center\n",
    "        self.scale = scale\n",
    "        if epsilon is None:\n",
    "            epsilon = K.epsilon() * K.epsilon()\n",
    "        self.epsilon = epsilon\n",
    "        self.gamma_initializer = keras.initializers.get(gamma_initializer)\n",
    "        self.beta_initializer = keras.initializers.get(beta_initializer)\n",
    "        self.gamma_regularizer = keras.regularizers.get(gamma_regularizer)\n",
    "        self.beta_regularizer = keras.regularizers.get(beta_regularizer)\n",
    "        self.gamma_constraint = keras.constraints.get(gamma_constraint)\n",
    "        self.beta_constraint = keras.constraints.get(beta_constraint)\n",
    "        self.gamma, self.beta = None, None\n",
    "\n",
    "        \n",
    "    def get_config(self):\n",
    "        config = {\n",
    "            'center': self.center,\n",
    "            'scale': self.scale,\n",
    "            'epsilon': self.epsilon,\n",
    "            'gamma_initializer': keras.initializers.serialize(self.gamma_initializer),\n",
    "            'beta_initializer': keras.initializers.serialize(self.beta_initializer),\n",
    "            'gamma_regularizer': keras.regularizers.serialize(self.gamma_regularizer),\n",
    "            'beta_regularizer': keras.regularizers.serialize(self.beta_regularizer),\n",
    "            'gamma_constraint': keras.constraints.serialize(self.gamma_constraint),\n",
    "            'beta_constraint': keras.constraints.serialize(self.beta_constraint),\n",
    "        }\n",
    "        base_config = super(LayerNormalization, self).get_config()\n",
    "        return dict(list(base_config.items()) + list(config.items()))\n",
    "\n",
    "    def compute_output_shape(self, input_shape):\n",
    "        return input_shape\n",
    "\n",
    "    def compute_mask(self, inputs, input_mask=None):\n",
    "        return input_mask\n",
    "\n",
    "    def build(self, input_shape):\n",
    "        shape = input_shape[-1:]\n",
    "        if self.scale:\n",
    "            self.gamma = self.add_weight(\n",
    "                shape=shape,\n",
    "                initializer=self.gamma_initializer,\n",
    "                regularizer=self.gamma_regularizer,\n",
    "                constraint=self.gamma_constraint,\n",
    "                name='gamma',\n",
    "            )\n",
    "        if self.center:\n",
    "            self.beta = self.add_weight(\n",
    "                shape=shape,\n",
    "                initializer=self.beta_initializer,\n",
    "                regularizer=self.beta_regularizer,\n",
    "                constraint=self.beta_constraint,\n",
    "                name='beta',\n",
    "            )\n",
    "        super(LayerNormalization, self).build(input_shape)\n",
    "        \n",
    "    def call(self, inputs, training=None):\n",
    "        mean = K.mean(inputs, axis=-1, keepdims=True)\n",
    "        variance = K.mean(K.square(inputs - mean), axis=-1, keepdims=True)\n",
    "        std = K.sqrt(variance + self.epsilon)\n",
    "        outputs = (inputs - mean) / std\n",
    "        if self.scale:\n",
    "            outputs *= self.gamma\n",
    "        if self.center:\n",
    "            outputs += self.beta\n",
    "        return outputs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T16:59:44.825669Z",
     "start_time": "2020-01-06T16:59:44.668598Z"
    }
   },
   "outputs": [],
   "source": [
    "y_mae = train_reorient['Yards'].values[::22]\n",
    "y_mean = np.median(y_mae)\n",
    "y_crps = np.zeros(shape=(y_mae.shape[0], 199))\n",
    "for i,yard in enumerate(y_mae):\n",
    "    y_crps[i, yard+99:] = 1\n",
    "    \n",
    "yardline = train_reorient['YardLine_std'].values[::22]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:02.755546Z",
     "start_time": "2020-01-06T17:00:02.749544Z"
    }
   },
   "outputs": [],
   "source": [
    "def crps(y_true, y_pred):\n",
    "    stops = np.arange(-99, 100)\n",
    "    unit_steps = stops >= y_true.reshape(-1, 1)\n",
    "    crps = np.mean((y_pred - unit_steps)**2)\n",
    "    return crps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:08.951547Z",
     "start_time": "2020-01-06T17:00:08.728525Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.014500742984789266"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "baseline = np.histogram(y_mae, bins = range(-100, 100, 1))[0].cumsum()/len(y_mae)\n",
    "crps(y_mae, baseline)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:14.509335Z",
     "start_time": "2020-01-06T17:00:14.503332Z"
    }
   },
   "outputs": [],
   "source": [
    "def nondecreasing(x):\n",
    "    X_ir = np.arange(199).astype('float64')\n",
    "    ir = IsotonicRegression(0, 1)\n",
    "    x = ir.fit_transform(X_ir, x.astype('float64'))        \n",
    "    return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:23.841415Z",
     "start_time": "2020-01-06T17:00:23.825410Z"
    }
   },
   "outputs": [],
   "source": [
    "from keras.models import Model, load_model\n",
    "from keras.layers import Input, BatchNormalization, Activation, Add, Multiply, Dot\n",
    "from keras.layers import Embedding, Permute, Reshape, GaussianNoise\n",
    "from keras.layers.core import Dropout, Lambda, Dense, Flatten\n",
    "from keras.layers.convolutional import Conv2D, Conv2DTranspose, UpSampling2D, Conv1D\n",
    "from keras.layers.pooling import GlobalMaxPooling1D, GlobalAveragePooling2D, AveragePooling2D\n",
    "from keras.layers.merge import concatenate, Concatenate\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau, LearningRateScheduler\n",
    "from keras.optimizers import Adam, SGD, Nadam\n",
    "from keras import backend as K\n",
    "\n",
    "from keras.engine.topology import Layer\n",
    "import tensorflow as tf\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:33.350854Z",
     "start_time": "2020-01-06T17:00:33.339552Z"
    }
   },
   "outputs": [],
   "source": [
    "class ScaleLayer(Layer):\n",
    "\n",
    "    def __init__(self, output_dim, **kwargs):\n",
    "        self.output_dim = output_dim\n",
    "        super(ScaleLayer, self).__init__(**kwargs)\n",
    "\n",
    "    def build(self, input_shape):\n",
    "        super(ScaleLayer, self).build(input_shape) \n",
    "\n",
    "    def call(self, x):\n",
    "        xx = K.arange(-99, 100, dtype=tf.float32)\n",
    "        mu = y_mean + tf.reshape(x[:, 0], (-1, 1))\n",
    "        sigma_minus = tf.identity(K.exp(0.5 * tf.reshape(x[:, 1], (-1, 1))), name=\"sigma\")\n",
    "        sigma_plus = tf.identity(K.exp(0.5 * tf.reshape(x[:, 2], (-1, 1))), name=\"sigma\")\n",
    "        xx = tf.subtract(xx, mu)\n",
    "        pcf = tf.where(xx >= 0, tf.divide (xx, sigma_plus),  tf.divide (xx, sigma_minus))\n",
    "        return pcf\n",
    "\n",
    "    def compute_output_shape(self, input_shape):\n",
    "        return (input_shape[0], self.output_dim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2020-01-06T17:00:45.435544Z",
     "start_time": "2020-01-06T17:00:45.428544Z"
    }
   },
   "outputs": [],
   "source": [
    "def message_mult(n, e):\n",
    "    res = Lambda(lambda c: K.batch_dot(c[0], c[1]), name=\"mul_1\")([e, n])\n",
    "    res = Lambda(lambda res: K.sum(res, axis=-1))(res)\n",
    "    res = Lambda(lambda res: K.sum(res, axis=-1, keepdims=True))(res)\n",
    "    return res\n",
    "\n",
    "def rusher_dist_mult(dist, players):\n",
    "    res = Lambda(lambda c: c[0] * c[1])([dist, players])\n",
    "    return res\n",
    "  \n",
    "def dist_mult(dist, players):\n",
    "    res = Lambda(lambda c: K.batch_dot(c[0], c[1]))([dist, players])\n",
    "    return res\n"
   ]
  }
 ],
 "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": "3.7.4"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "208.594px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}