""" =================== Bin Statistic Sonar =================== StatsBomb has a great `blog `_ on the history of Sonars. Sonars show more information than heatmaps by introducing the angle of passes, shots or other events. The following examples show how to use the ``bin_statistic_sonar`` method to bin data by x/y coordinates and angles. More information is available on how to customize the plotted sonars in :ref:`sphx_glr_gallery_sonars_plot_sonar_grid.py` and :ref:`sphx_glr_gallery_sonars_plot_sonar.py`. """ import matplotlib.pyplot as plt import numpy as np from mplsoccer import Pitch, VerticalPitch, Sbopen ############################################################################## # Load the first game that Messi played as a false-9. parser = Sbopen() df = parser.event(69249)[0] # 0 index is the event file df = df[(df.type_name == 'Pass') & (df.team_name == 'Barcelona') & (~df.sub_type_name.isin(['Free Kick', 'Throw-in', 'Goal Kick', 'Kick Off', 'Corner']))].copy() ############################################################################## # Plot a Pass Sonar # ----------------- # Here, we calculate the angle and distance for each pass. # We then split the data into 6x4 grid cells. Within each grid cell, we # split the data into four equal segments of 90 degrees (360 / 4). # The defaults count the number of actions (passes) in each segment. pitch = Pitch() angle, distance = pitch.calculate_angle_and_distance(df.x, df.y, df.end_x, df.end_y) bs = pitch.bin_statistic_sonar(df.x, df.y, angle, bins=(6, 4, 4), # x, y, angle binning # center the first angle so it starts # at -45 degrees (90 / 2) rather than 0 degrees center=True) fig, ax = pitch.draw(figsize=(8, 5.5)) axs = pitch.sonar_grid(bs, width=15, fc='cornflowerblue', ec='black', ax=ax) ############################################################################## # Center argument # --------------- # You can either center the first slice around zero degrees (``center=True``) # or start the first segment at zero degrees (``center=False``). pitch = VerticalPitch() fig, axs = pitch.draw(figsize=(8, 6), nrows=1, ncols=2) angle, distance = pitch.calculate_angle_and_distance(df.x, df.y, df.end_x, df.end_y) bs_center = pitch.bin_statistic_sonar(df.x, df.y, angle, bins=(6, 4, 4), center=True) bs_not_center = pitch.bin_statistic_sonar(df.x, df.y, angle, bins=(6, 4, 4), center=False) axs_sonar = pitch.sonar_grid(bs_center, width=15, fc='cornflowerblue', ec='black', ax=axs[0]) axs_sonar = pitch.sonar_grid(bs_not_center, width=15, fc='cornflowerblue', ec='black', ax=axs[1]) text1 = pitch.text(60, 40, 'center=True', va='center', ha='center', fontsize=15, ax=axs[0]) text1 = pitch.text(60, 40, 'center=False', va='center', ha='center', fontsize=15, ax=axs[1]) ############################################################################## # Statistic # --------- # The default ``statistic='count'`` calculates counts in each segment. # You can also use the ``statistic`` and ``values`` arguments to calculate # other statistics. Here, we calculate the average pass distance # and plot this instead of the count of passes. # You can also normalize results between 0 to 1 with the ``normalize=True`` argument. pitch = Pitch() angle, distance = pitch.calculate_angle_and_distance(df.x, df.y, df.end_x, df.end_y) bs = pitch.bin_statistic_sonar(df.x, df.y, angle, # calculate the average distance # you can also calculate other statistics # such as std, median, sum, min and the max values=distance, statistic='mean', bins=(6, 4, 4), center=True) fig, ax = pitch.draw(figsize=(8, 5.5)) axs = pitch.sonar_grid(bs, width=15, fc='cornflowerblue', ec='black', ax=ax) ############################################################################## # Bins # ---- # In addition to integer values for ``bins``, you can use a sequence of angle edges. # The angle edges should be between zero and 2*pi (~6.283), i.e. the angles # in radians. You can convert from degrees to radians using numpy.radians. pitch = Pitch() angle, distance = pitch.calculate_angle_and_distance(df.x, df.y, df.end_x, df.end_y) x_bin = 3 # the bin argument can contain a mix of sequences and integers y_bin = pitch.dim.positional_y # I use cumsum so I can use widths rather than bin edges. # I convert to radians using numpy angle_bin = np.radians(np.array([0, 90, 45, 90, 90, 45])).cumsum() bs = pitch.bin_statistic_sonar(df.x, df.y, angle, bins=(x_bin, y_bin, angle_bin), center=True) fig, ax = pitch.draw(figsize=(8, 5.5)) axs = pitch.sonar_grid(bs, width=15, fc='cornflowerblue', ec='black', ax=ax) ############################################################################## # Binnumber # --------- # You can also get the bin numbers from the bin_statistic_sonar result. # Here, we use the ``binnumber`` to filter for the forward passes in the # final third and plot them as arrows. pitch = Pitch() angle, distance = pitch.calculate_angle_and_distance(df.x, df.y, df.end_x, df.end_y) bs = pitch.bin_statistic_sonar(df.x, df.y, angle, bins=(3, 1, 2), center=True) fig, ax = pitch.draw(figsize=(8, 5.5)) axs = pitch.sonar_grid(bs, width=15, fc='cornflowerblue', ec='black', ax=ax) mask = np.logical_and(np.logical_and(bs['binnumber'][0] == 2, # x in the final third bs['binnumber'][1] == 0), # only one y but here for completeness bs['binnumber'][2] == 0 # first angle ) arr = pitch.arrows(df[mask].x, df[mask].y, df[mask].end_x, df[mask].end_y, ax=ax) plt.show() # If you are using a Jupyter notebook you do not need this line