import pandas as pd
data = pd.read_csv('karnataka-2013-results.csv')
data.head()

pd.set_option('display.line_width', 200)
pd.set_option('display.max_columns', 20)

data['Votes'].sum()

data[data['Votes'] == data['Votes'].max()]

data['Party'].value_counts()

data['Rank'] = data.groupby('District')['Votes'].rank(ascending=False)
data.head()

data[data['Rank'] == 1]['Party'].value_counts()

data[data['Rank'] == 1].set_index('District').head()

rank1 = data[data['Rank'] == 1].set_index('District')
rank2 = data[data['Rank'] == 2].set_index('District')
pd.merge(rank1, rank2, left_index=True, right_index=True, suffixes=['1', '2']).head()

margins = pd.merge(rank1, rank2, left_index=True, right_index=True, suffixes=['1', '2'])
margins['Margin'] = margins['Votes1'] - margins['Votes2']

margins.sort('Margin', ascending=False).head()

margins['Votes'] = data.groupby('District')['Votes'].sum()
margins['% Margin'] = margins['Margin'].astype(float) / margins['Votes']
margins.sort('% Margin', ascending=False).head()

margins[margins['Votes1'].astype(float) / margins['Votes'] > 0.5].head()

margins[margins['Votes1'].astype(float) / margins['Votes'] > 0.5]['Party1'].value_counts()

majority = pd.DataFrame({
    'Absolute majority': margins[margins['Votes1'].astype(float) / margins['Votes'] > 0.5]['Party1'].value_counts(),
    'Winning party': data[data['Rank'] == 1]['Party'].value_counts(),
})
majority['Ratio'] = majority['Absolute majority'] / majority['Winning party']
majority.dropna().sort('Ratio', ascending=False)

data['District'].apply(lambda district: margins.ix[district]).head()

data['% votes'] = data['Votes'].astype(float) / data['District'].apply(lambda district: margins['Votes'][district])
data.head()

sum(data['% votes'] < 1. / 6)

len(data)

data[data['% votes'] < 1. / 6].sort('% votes').head(50)