Modeling- Baseline.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Baseline Modeling\n",
    "- Robert Shaw\n",
    "- CS109a Project: Data Driven March Madness\n",
    "\n",
    "In this file, we fit a simple head to head model to see how well we do in predicting games. This head to head model will eventually be used to predict an entire bracket down the road."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load in Our Datasets and Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import march_madness_classes as mmc\n",
    "import march_madness_games as mmg\n",
    "import march_madness_models as mmm\n",
    "\n",
    "# import libraries\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "teams = pd.read_csv(\"datasets/kaggle_data/Teams.csv\")\n",
    "seeds = pd.read_csv(\"datasets/kaggle_data/TourneySeeds.csv\")\n",
    "slots = pd.read_csv(\"datasets/kaggle_data/TourneySlots.csv\")\n",
    "tourney_data = pd.read_csv(\"datasets/kaggle_data/TourneyCompactResults.csv\")\n",
    "regular_data = pd.read_csv(\"datasets/kaggle_data/RegularSeasonCompactResults.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "tourney_arr = mmg.filter_into_seasons(tourney_data)\n",
    "regular_arr = mmg.filter_into_seasons(regular_data)\n",
    "seeds_arr = mmg.filter_into_seasons(seeds)\n",
    "slots_arr = mmg.filter_into_seasons(slots)\n",
    "\n",
    "seeds_1985 = seeds_arr[0]\n",
    "slots_1985 = slots_arr[0]\n",
    "tourney_1985 = tourney_arr[0]\n",
    "regular_1985 = regular_arr[0]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Transform the Stationary Distibution Data Into One DataFrame/CSV File\n",
    "- Improve our database."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "stationary_arr = []\n",
    "for i in range(1985, 2017):\n",
    "    df = pd.read_csv(\"datasets/our_data/markov_data/{}_stationary_distribution\".format(i), index_col = 0)\n",
    "    df.columns = [i]\n",
    "    stationary_arr.append(df.transpose()) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# concat into a \n",
    "stationary_df = pd.concat(stationary_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# save as a csv\n",
    "stationary_df.to_csv(\"datasets/our_data/stationary\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1985    1.0\n",
       "1986    1.0\n",
       "1987    1.0\n",
       "1988    1.0\n",
       "1989    1.0\n",
       "1990    1.0\n",
       "1991    1.0\n",
       "1992    1.0\n",
       "1993    1.0\n",
       "1994    1.0\n",
       "1995    1.0\n",
       "1996    1.0\n",
       "1997    1.0\n",
       "1998    1.0\n",
       "1999    1.0\n",
       "2000    1.0\n",
       "2001    1.0\n",
       "2002    1.0\n",
       "2003    1.0\n",
       "2004    1.0\n",
       "2005    1.0\n",
       "2006    1.0\n",
       "2007    1.0\n",
       "2008    1.0\n",
       "2009    1.0\n",
       "2010    1.0\n",
       "2011    1.0\n",
       "2012    1.0\n",
       "2013    1.0\n",
       "2014    1.0\n",
       "2015    1.0\n",
       "2016    1.0\n",
       "dtype: float64"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# everything appears to be working if they sum to one\n",
    "stationary_df.sum(axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Get Predictors For Our Baseline Model\n",
    "- Pull from our database."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "stationary = pd.read_csv(\"datasets/our_data/stationary\", index_col =0)\n",
    "avg_points_against =  pd.read_csv(\"datasets/our_data/avg_points_against\", index_col =0)\n",
    "avg_points_for = pd.read_csv(\"datasets/our_data/avg_points_for\", index_col =0)\n",
    "away_wins = pd.read_csv(\"datasets/our_data/away_wins\", index_col =0)\n",
    "bad_losses = pd.read_csv(\"datasets/our_data/bad_losses\", index_col =0)\n",
    "consistency= pd.read_csv(\"datasets/our_data/consistency\", index_col =0)\n",
    "dominance= pd.read_csv(\"datasets/our_data/dominance\", index_col =0)\n",
    "good_wins_matrix= pd.read_csv(\"datasets/our_data/good_wins_matrix\", index_col =0)\n",
    "rpi= pd.read_csv(\"datasets/our_data/rpi\", index_col =0)\n",
    "tough_wins= pd.read_csv(\"datasets/our_data/tough_wins\", index_col =0)\n",
    "win_percentage= pd.read_csv(\"datasets/our_data/win_percentage\", index_col =0)\n",
    "win_percentage_vs_tourney_teams_matrix= pd.read_csv(\"datasets/our_data/win_percentage_vs_tourney_teams_matrix\", index_col =0)\n",
    "wins_vs_tourney_teams= pd.read_csv(\"datasets/our_data/wins_vs_tourney_teams\", index_col =0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Functions to go into our March Madness Games Module\n",
    "- These have been moved to march_madness_games.py\n",
    "- These functions have been explained in march_madness_games.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# extracting from the dataframe\n",
    "def get_predictor(team_id, year, df):\n",
    "    return df.loc[year, str(team_id)]\n",
    "\n",
    "def get_predictor_dif(team_id_1, team_id_2, year, df):\n",
    "    return df.loc[year, str(team_id_1)] - df.loc[year, str(team_id_2)]\n",
    "\n",
    "def get_predictors(team_id, year, df_arr):\n",
    "    row = np.zeros(len(df_arr))\n",
    "    i = 0\n",
    "    for df in df_arr:\n",
    "        row[i] = get_predictor(team_id, year, df)\n",
    "        i = i + 1\n",
    "    return row\n",
    "\n",
    "def get_predictors_dif(team_id_1, team_id_2, year, df_arr):\n",
    "    row = np.zeros(len(df_arr))\n",
    "    i = 0\n",
    "    for df in df_arr:\n",
    "        row[i] = float(get_predictor_dif(team_id_1, team_id_2, year, df))\n",
    "        i = i + 1\n",
    "    return row"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# function to extract the y_values of the team with the min index winning\n",
    "def extract_response(tourney_game_df):\n",
    "    \n",
    "    # response for a given year\n",
    "    min_index_win = np.zeros(tourney_game_df.shape[0])\n",
    "    i = 0\n",
    "    for index, game in tourney_game_df.iterrows():\n",
    "        if int(game[\"Prediction\"]) == min(int(game[\"Strongseed Team\"]), int(game[\"Weakseed Team\"])):\n",
    "            min_index_win[i] = 1 \n",
    "        i = i + 1\n",
    "           \n",
    "    return min_index_win\n",
    "\n",
    "# function to extract the y_values of the team with the min index winning\n",
    "def extract_predictors(tourney_game_df, predictor_list, predictor_dfs, year):\n",
    "    # buffer to hold our values\n",
    "    pred_matrix = np.zeros((tourney_game_df.shape[0], len(predictor_list)))\n",
    "    \n",
    "    # fill predictor matrix\n",
    "    for i in range(tourney_game_df.shape[0]):   \n",
    "        # min and max index teams\n",
    "        min_index_team = min(int(tourney_game_df.loc[i, \"Strongseed Team\"]), int(tourney_game_df.loc[i, \"Weakseed Team\"]))\n",
    "        max_index_team = max(int(tourney_game_df.loc[i, \"Strongseed Team\"]), int(tourney_game_df.loc[i, \"Weakseed Team\"]))                  \n",
    "\n",
    "        # fill matrix\n",
    "        pred_matrix[i,  0] = min_index_team\n",
    "        pred_matrix[i,  1] = max_index_team\n",
    "        pred_matrix[i, 2:] = get_predictors_dif(min_index_team, max_index_team, year, predictor_dfs)\n",
    "\n",
    "    # gen dataframe                       \n",
    "    pred_df = pd.DataFrame(data = pred_matrix, columns = predictor_list)\n",
    "                           \n",
    "    return pred_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def get_tourney_results(seeds, slots, raw_data):\n",
    "    tourney = mmc.Tournament(seeds, slots, mmm.ActualTournament(raw_data))\n",
    "    return tourney.entire_bracket\n",
    "\n",
    "# get single years worth of games\n",
    "def generate_single_year_of_games(year, seed_list, slot_list, tourney_data, predictors, predictor_dfs):\n",
    "    # get results of the games\n",
    "    tourney_results = get_tourney_results(seed_list, slot_list, tourney_data)\n",
    "    \n",
    "    # get predictors\n",
    "    pred_df = extract_predictors(tourney_results, predictors, predictor_dfs, year)\n",
    "    \n",
    "    # get response\n",
    "    resp_arr = extract_response(tourney_results)\n",
    "    \n",
    "    return pred_df, resp_arr\n",
    "\n",
    "def generate_multiple_years_of_games(years, seed_list_arr, slot_list_arr, tourney_data_arr, predictors, predictor_dfs):\n",
    "    min_year = 1985\n",
    "\n",
    "    preds = pd.DataFrame({})\n",
    "    resps = np.array([])\n",
    "    \n",
    "    for year in years:\n",
    "        year_index = int(year) - min_year\n",
    "        # generate 1 year of data\n",
    "        pred_df, resp_arr = generate_single_year_of_games(year, \n",
    "                                                          seed_list_arr[year_index], \n",
    "                                                          slot_list_arr[year_index], \n",
    "                                                          tourney_data_arr[year_index],\n",
    "                                                          predictors,\n",
    "                                                          predictor_dfs)\n",
    "        print year\n",
    "        # add to list we are keeping \n",
    "        preds = pd.concat([preds, pred_df])\n",
    "        resps = np.concatenate((resps, resp_arr))\n",
    "        \n",
    "    return preds, resps\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "column_names = [\"min_index_id\", \"max_index_id\", \"wins_vs_tourney_teams dif\", \"stationary dif\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "tourney_arr = mmg.filter_into_seasons(tourney_data)\n",
    "regular_arr = mmg.filter_into_seasons(regular_data)\n",
    "seeds_arr = mmg.filter_into_seasons(seeds)\n",
    "slots_arr = mmg.filter_into_seasons(slots)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "predictor_dfs = [wins_vs_tourney_teams, stationary]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1985\n",
      "1986\n",
      "1987\n",
      "1988\n",
      "1989\n",
      "1990\n",
      "1991\n",
      "1992\n",
      "1993\n",
      "1994\n",
      "1995\n",
      "1996\n",
      "1997\n",
      "1998\n",
      "1999\n",
      "2000\n"
     ]
    }
   ],
   "source": [
    "pred, resp = generate_multiple_years_of_games(range(1985, 2001), \n",
    "                                              seeds_arr, \n",
    "                                              slots_arr, \n",
    "                                              tourney_arr, \n",
    "                                              column_names, \n",
    "                                              predictor_dfs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "pred_df = pred.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "resp_df = pd.DataFrame(data=resp, columns=[\"min_index_win\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "RangeIndex(start=0, stop=1008, step=1)\n",
      "RangeIndex(start=0, stop=1008, step=1)\n"
     ]
    }
   ],
   "source": [
    "print pred_df.index\n",
    "print resp_df.index"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Try Fitting a Head to Head Model\n",
    "\n",
    "#### NOTE: We now use a windowing approach for train test split ---> see Model Selection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.cross_validation import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_index, cross_index = train_test_split(pred_df.index, test_size = .25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "train_x = pred_df.loc[train_index]\n",
    "train_y = resp_df.loc[train_index]\n",
    "cross_x = pred_df.loc[cross_index]\n",
    "cross_y = resp_df.loc[cross_index]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression as LogReg\n",
    "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA\n",
    "from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis as QDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "logistic_model = LogReg(C=100)\n",
    "lda_model = LDA()\n",
    "qda_model = QDA()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.74603174603174605"
      ]
     },
     "execution_count": 66,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = logistic_model\n",
    "model.fit(train_x.iloc[:, 2:], train_y.values.T[0])\n",
    "model.score(cross_x.iloc[:, 2:], cross_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.73809523809523814"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = lda_model\n",
    "model.fit(train_x.iloc[:, 2:], train_y.values.T[0])\n",
    "model.score(cross_x.iloc[:, 2:], cross_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>min_index_id</th>\n",
       "      <th>max_index_id</th>\n",
       "      <th>wins_vs_tourney_teams dif</th>\n",
       "      <th>stationary dif</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>632</th>\n",
       "      <td>1330.0</td>\n",
       "      <td>1437.0</td>\n",
       "      <td>-6.0</td>\n",
       "      <td>-0.002784</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61</th>\n",
       "      <td>1272.0</td>\n",
       "      <td>1437.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.001299</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>941</th>\n",
       "      <td>1163.0</td>\n",
       "      <td>1211.0</td>\n",
       "      <td>8.0</td>\n",
       "      <td>0.003702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>216</th>\n",
       "      <td>1283.0</td>\n",
       "      <td>1424.0</td>\n",
       "      <td>-2.0</td>\n",
       "      <td>-0.003136</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>907</th>\n",
       "      <td>1108.0</td>\n",
       "      <td>1390.0</td>\n",
       "      <td>-9.0</td>\n",
       "      <td>-0.005476</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     min_index_id  max_index_id  wins_vs_tourney_teams dif  stationary dif\n",
       "632        1330.0        1437.0                       -6.0       -0.002784\n",
       "61         1272.0        1437.0                        0.0        0.001299\n",
       "941        1163.0        1211.0                        8.0        0.003702\n",
       "216        1283.0        1424.0                       -2.0       -0.003136\n",
       "907        1108.0        1390.0                       -9.0       -0.005476"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_x.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x11cd75c50>"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgMAAAFVCAYAAABy5OwgAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nEEIIscL9PxPK8KX0zRY9AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x11cd432d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(8, 5))\n",
    "ax = fig.add_subplot(111)\n",
    "\n",
    "ax.scatter(train_x.iloc[train_y.values.T[0] == 1, 3], train_x.iloc[train_y.values.T[0] == 1, 2], color=\"g\", alpha=.5, label=\"Min Index Win\")\n",
    "ax.scatter(train_x.iloc[train_y.values.T[0] == 0, 3], train_x.iloc[train_y.values.T[0] == 0, 2], color=\"r\", alpha=.5, label=\"Min Index Loss\")\n",
    "ax.legend()\n",
    "\n",
    "ax.set_title(\"Distribution of $\\pi_1-\\pi_2$ and Dominance dif\")\n",
    "ax.set_xlabel(\"$\\pi$ dif\")\n",
    "ax.set_ylabel(\"dominance dif\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### These are now at march_madness_models.py / march_madness_classes.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# see march madness models\n",
    "class ModelPredictor(object):\n",
    "    # init function\n",
    "    def __init__(self, model, dfs_arr, year):\n",
    "        self.model = model\n",
    "        self.dfs_arr = dfs_arr\n",
    "        self.year = year\n",
    "        return\n",
    "    \n",
    "    # head to head predicitons\n",
    "    def predict(self, team_1, team_2):\n",
    "        team_1 = int(team_1)\n",
    "        team_2 = int(team_2)\n",
    "        \n",
    "        # min and max index\n",
    "        min_index_team = min(team_1, team_2)\n",
    "        max_index_team = max(team_1, team_2)\n",
    "        \n",
    "        # get the x values\n",
    "        row = get_predictors_dif(min_index_team, max_index_team, self.year, self.dfs_arr)\n",
    "\n",
    "        # predict under model\n",
    "        y_hat = model.predict(row.reshape(1,-1))\n",
    "        \n",
    "        if y_hat == 1:\n",
    "            return min_index_team\n",
    "        else:\n",
    "            return max_index_team\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "year = 1992\n",
    "\n",
    "seeds = seeds_arr[year - 1985]\n",
    "slots = slots_arr[year - 1985]\n",
    "resul = tourney_arr[year - 1985]\n",
    "\n",
    "tourney_actual = mmc.Tournament(seeds, slots, mmc.ActualTournament(resul))\n",
    "tourney_actual.simulate_tournament()\n",
    "\n",
    "tourney_top_seed = mmc.Tournament(seeds, slots, mmc.BasicPredictor())\n",
    "tourney_top_seed.simulate_tournament()\n",
    "\n",
    "tourney_model = mmc.Tournament(seeds, slots, ModelPredictor(lda_model, predictor_dfs, year))\n",
    "tourney_model.simulate_tournament()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Points  : 1320\n",
      "\n",
      "Total Accuracy: 46 / 63 = 0.730158730159\n",
      "R1    Accuracy: 26 / 32 = 0.8125\n",
      "R2    Accuracy: 11 / 16 = 0.6875\n",
      "R3    Accuracy: 5 / 8 = 0.625\n",
      "R4    Accuracy: 2 / 4 = 0.5\n",
      "R5    Accuracy: 1 / 2 = 0.5\n",
      "R6    Accuracy: 1 / 1 = 1.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(1320, 0.7301587301587301)"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tourney_model.score_model(tourney_actual, print_res=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Points  : 1220\n",
      "\n",
      "Total Accuracy: 43 / 63 = 0.68253968254\n",
      "R1    Accuracy: 24 / 32 = 0.75\n",
      "R2    Accuracy: 11 / 16 = 0.6875\n",
      "R3    Accuracy: 5 / 8 = 0.625\n",
      "R4    Accuracy: 1 / 4 = 0.25\n",
      "R5    Accuracy: 1 / 2 = 0.5\n",
      "R6    Accuracy: 1 / 1 = 1.0\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(1220, 0.6825396825396826)"
      ]
     },
     "execution_count": 196,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tourney_top_seed.score_model(tourney_actual, print_res=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 140,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "score_mod = np.zeros(len(range(1985, 2001)))\n",
    "accur_mod = np.zeros(len(range(1985, 2001)))\n",
    "\n",
    "score_top = np.zeros(len(range(1985, 2001)))\n",
    "accur_top = np.zeros(len(range(1985, 2001)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "i = 0\n",
    "for year in range(1985, 2001):\n",
    "    seeds = seeds_arr[year - 1985]\n",
    "    slots = slots_arr[year - 1985]\n",
    "    resul = tourney_arr[year - 1985]\n",
    "\n",
    "    tourney_actual = mmc.Tournament(seeds, slots, mmc.ActualTournament(resul))\n",
    "    tourney_actual.simulate_tournament()\n",
    "\n",
    "    tourney_top_seed = mmc.Tournament(seeds, slots, mmc.BasicPredictor())\n",
    "    tourney_top_seed.simulate_tournament()\n",
    "\n",
    "    tourney_model = mmc.Tournament(seeds, slots, ModelPredictor(lda_model, predictor_dfs, year))\n",
    "    tourney_model.simulate_tournament()\n",
    "    \n",
    "    score_mod[i], accur_mod[i] = tourney_model.score_model(tourney_actual, print_res=False)\n",
    "    score_top[i], accur_top[i] = tourney_top_seed.score_model(tourney_actual, print_res=False)\n",
    "    \n",
    "    i = i + 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 142,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "920.0\n",
      "0.650793650794\n"
     ]
    }
   ],
   "source": [
    "print np.mean(score_mod)\n",
    "print np.mean(accur_mod)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 143,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "890.0\n",
      "0.64880952381\n"
     ]
    }
   ],
   "source": [
    "print np.mean(score_top)\n",
    "print np.mean(accur_top)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}