distance.py

def distance(pid, yr):
    # Computes the Mahalanobis distance for a given player to all other player.

    # Get player data
    player = orig_data[(orig_data.playerID == pid) & (orig_data.yearID == yr)][cols]
    sid = player.index.astype(int)[0]

    print('Comparing: {} (id: {})'.format(pid, sid))

    # Mask invalid values in the player vector
    pvec = np.ma.masked_invalid(np.array(player))

    min_player = None
    min_val = None

    for i in range(len(data)):
        # Get the ith player season
        cdata = data.iloc[i]

        # Ignore the current player season
        if cdata.name == sid:
            continue

        # Mask invalid values
        cvec = np.ma.masked_invalid(np.array(cdata))

        # Find difference between x and y
        delta = pvec - cvec

        # Find Mahalanobis distance
        dist = np.sqrt(np.einsum('nj,jk,nk->n', delta, invcov, delta))[0]

        # Check to see if current distance is smallest, if so, keep it.
        if min_id == None or min_val > dist:
            min_player = batting_data.iloc[i]
            min_val = dist

    # Print out the most similar season
    print('Most similar: dist: {}\n{}'.format(min_val, min_player))