databyjp

import plotly.express as px
fig = px.scatter(shot_blot_df,
                 title=f'{latest_day_str} - Playoff game shot profiles',
                 x="filt_avg", y="segment", size="shot_freq_x",
                 color="shot_acc_x", color_continuous_scale=px.colors.sequential.Blues,
                 facet_row="group",
                 template="plotly_white", width=1200, height=750,
                 labels={'filt_avg': 'Distance from the rim', 'segment': 'Sample size',
                         'pts_pct_x': 'Proportion of points', 'shot_ev': 'Expected<BR>points<BR>per shot'}
                 )

databyjp

import pandas as pd
import utils
from scipy.spatial.distance import cosine
from nba_api.stats.static import teams

shots_df = utils.load_shots_df()
gdf = utils.get_shot_dist_df(shots_df)

# FILTER GAMES FOR THE LATEST DAY
day_df = shots_df[shots_df["timeActual"].dt.date == shots_df["timeActual"].dt.date.max()]

databyjp

# Build more granular shot bins
filt_start = 0
filt_width = 2
filt_inc = 0.25
gdf_list = list()
for i in range(1 + int(30 * (1 / filt_inc))):
    filt_df = shots_df[(shots_df["shotDistance"] >= filt_start) & (shots_df["shotDistance"] < filt_start + filt_width)]
    gdf = filt_df.groupby("shot_zone").agg({"shot_made": "sum", "period": "count"})
    gdf = gdf.reset_index().rename({"period": "shot_atts"}, axis=1)
    gdf = gdf.assign(team="NBA")

databyjp

tm_ranks = gdf_tot.groupby("team").sum()["rel_pts"].sort_values().index.to_list()
fig = px.bar(gdf_tot, x="team", facet_row="shot_zone", y="shot_freq", color="rel_ev",
                 color_continuous_scale=px.colors.diverging.RdYlBu_r, color_continuous_midpoint=0,
                 category_orders={"team": tm_ranks})
fig.update_traces(marker=dict(line=dict(width=1, color='DarkSlateGrey')))
fig.show()

databyjp

# ========================================
# Now let's do the same for each team
# ========================================
shots_df["teamId"] = shots_df["teamId"].astype(int)
gdf_list = list()
for tm_id in shots_df.teamId.unique():
    tm_df = shots_df[shots_df.teamId == tm_id]
    tm_gdf = tm_df.groupby("shot_zone").agg({"shot_made": "sum", "period": "count"})
    tm_gdf = tm_gdf.reset_index().rename({"period": "shot_atts"}, axis=1)

databyjp

gdf = shots_df.groupby("shot_zone").agg({"shot_made": "sum", "period": "count"})
gdf = gdf.reset_index().rename({"period": "shot_atts"}, axis=1)
gdf = gdf.assign(team="NBA")
gdf = gdf.assign(shot_freq=gdf.shot_atts / gdf.shot_atts.sum())
gdf = gdf.assign(shot_acc=gdf.shot_made / gdf.shot_atts)

databyjp

import logging
import pandas as pd
import numpy as np
import utils
from nba_api.stats.static import players
from nba_api.stats.static import teams
import plotly.express as px

logger = logging.getLogger(__name__)

databyjp

def add_pbp_oncourt_columns(df):
    """
    Add on-court player columns to the play-by-play dataframe.
    Players based on substitution data and box-score data (for starters)
    :param df: PBP dataframe
    :return:
    """
    df = df.sort_values(["GAME_ID", "actionNumber"])
    df = df.reset_index(drop=True)

databyjp

def fetch_data_w_gameid(json_dir, gm_id, datatype="boxscore"):
    """
    Download a datafile based on gameID as downloaded from NBA API & saves to file
    :param json_dir: Directory for saving downloaded JSON
    :param gm_id: NBA game ID
    :param datatype: What data types to download - determines endpoint to use
    :return:
    """
    from nba_api.stats.endpoints import boxscoreadvancedv2
    from nba_api.live.nba.endpoints import playbyplay

databyjp

import os
srcdir = "dl_data/box_scores/json"
files = [f for f in os.listdir(srcdir) if f.endswith('.json') and not f.startswith('00')]
for f in files:
    os.rename(os.path.join(srcdir, f), os.path.join(srcdir, '00' + f))