Skip to content

Instantly share code, notes, and snippets.

@903124

903124/EPA_calculation.ipynb

Last active June 3, 2020 08:29
Show Gist options
  • Star 2 You must be signed in to star a gist
  • Fork 2 You must be signed in to fork a gist
  • Save 903124/d304f76688b0699497a35b61b6d1e267 to your computer and use it in GitHub Desktop.
Save 903124/d304f76688b0699497a35b61b6d1e267 to your computer and use it in GitHub Desktop.
Download ZIP
Display the source blob
Display the rendered blob
Raw
EPA_calculation.ipynb
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
@tzane
Copy link

tzane commented Jan 21, 2018

This is the only open-source contribution I have been able to find showing EPA calculation using the nfldb data source. Thanks for posting this! For whatever it's worth this simplified code produces the same result but might be easier to read:

import csv
import numpy as np
import nfldb

EPA_observed = np.zeros(100)
EPA_play = np.zeros(100)

db = nfldb.connect()
q = nfldb.Query(db) 
games = q.game(season_year=[2016], season_type='Regular').as_games()

def yardstr_to_num(yard_text):
    
    yard_split = yard_text.split()
    pos_indicate = yard_split[0]
    
    if pos_indicate == 'OWN':
		yardline_from_str = int(yard_split[1])
		return yardline_from_str
    elif pos_indicate == 'OPP':
		yardline_from_str = 100 - int(yard_split[1])
		return yardline_from_str
    else:
        yardline_from_str = 50
        return yardline_from_str
    
def iterate_plays(nfldb_obj, deduct_punts=False):
    for game in nfldb_obj:
        for drive in game.drives:
            if drive.result == 'Field Goal' or drive.result == 'Punt' or drive.result == 'Touchdown' or drive.result == 'Missed FG' or drive.result == 'Safety' or drive.result == 'Fumble, Safety' or drive.result == 'Interception' or drive.result == 'Fumble':
                for play in drive.plays:
                    if (play.passing_att == 1 or play.rushing_att == 1 or play.passing_sk == 1) and int(play.down) == 1:
					
						yardlinefromstr	= yardstr_to_num(str(play.yardline))
						
						if deduct_punts:
							end_field_fromstr = yardstr_to_num(str(drive.end_field))					
							
						if int(play.down) == 1:
							if drive.result == 'Field Goal':
								EPA_play[yardlinefromstr-1] += 1
								EPA_observed[yardlinefromstr-1] += 3
							if drive.result == 'Missed FG':
								EPA_play[yardlinefromstr-1] += 1
							if drive.result == 'Interception':
								EPA_play[yardlinefromstr-1] += 1
								if deduct_punts:
									EPA_observed[yardlinefromstr-1] -= temp_EPA[100-end_field_fromstr]
							if drive.result == 'Fumble':
								EPA_play[yardlinefromstr-1] += 1
								if deduct_punts:
									EPA_observed[yardlinefromstr-1] -= temp_EPA[100-end_field_fromstr]
							if drive.result == 'Punt':
								EPA_play[yardlinefromstr-1] += 1
								if deduct_punts:
									EPA_observed[yardlinefromstr-1] -= temp_EPA[100 - int(-0.0116 * end_field_fromstr * end_field_fromstr + 1.5343 * end_field_fromstr + 37.91) - 1]
							if drive.result == 'Safety' or drive.result == 'Fumble, Safety':
								EPA_play[yardlinefromstr-1] += 1
								EPA_observed[yardlinefromstr-1] -= 2
							if drive.result == 'Touchdown':
								EPA_play[yardlinefromstr-1] += 1
								EPA_observed[yardlinefromstr-1] += 7
	
iterate_plays(games)

for i in range(4):
    for j in range(99):
	    if EPA_play[j] > 0:
		    EPA_observed[j] = float(EPA_observed[j] / EPA_play[j])
			
    temp_EPA = EPA_observed
    EPA_observed = np.zeros(100)
    EPA_play = np.zeros(100)
	
    iterate_plays(games, deduct_punts=True)
	
for j in range(99):
	    if EPA_play[j] > 0:
		    EPA_observed[j] = float(EPA_observed[j] / EPA_play[j])

cof = np.polyfit(np.linspace(1,99,num=99),EPA_observed[0:99],5)

x = np.linspace(1,99,num=99)

EPA_observed_smooth = cof[0]*x**5 + cof[1]*x**4 + cof[2]*x**3 + cof[3]*x**2 + cof[4]*x + cof[5]

team_list = np.array(['ARI','ATL','BAL','BUF','CAR','CHI','CIN','CLE','DAL','DEN','DET','GB','HOU','IND','JAX','KC','LA','MIA','MIN','NE','NO'
                 ,'NYG','NYJ','OAK','PHI','PIT','SD','SEA','SF','TB','TEN','WAS'])	

week_array = [i+1 for i in range(17)]
week = 17
EPA_team = np.zeros(32)

next_drive_start_yardline = 50

epa_result = np.empty(0)

for game in games:
    for drive in game.drives:
        
        if str(game.home_team) == str(drive.pos_team):
            opp_team = str(game.away_team)
        else:
            opp_team = str(game.home_team)
		
        yardlinefromstr = yardstr_to_num(str(drive.start_field))
		
        EP_start = EPA_observed_smooth[yardlinefromstr-1]

        end_field_fromstr = yardstr_to_num(str(drive.end_field))
		
        if drive.result == 'End of Game' or drive.result == 'End of Half':
		    EP_end = EP_start
			
        if drive.result == 'Missed FG' or drive.result == 'Interception' or drive.result == 'Fumble' or drive.result == 'Downs' or drive.result == 'Blocked FG' or drive.result == 'Blocked Punt' or drive.result == 'Blocked FG, Downs' or drive.result == 'Blocked Punt, Downs':
		    EP_end = -EPA_observed_smooth[100-end_field_fromstr-1]
			
        if drive.result == 'Punt':
            EP_end = -EPA_observed_smooth[next_drive_start_yardline-1]
			
        if drive.result == 'Touchdown':
            EP_end = 7
		
        if drive.result == 'Field Goal':
		    EP_end = 3
			
        if drive.result == 'Fumble, Safety' or drive.result == 'Safety':
		    EP_end = -2 - EPA_observed_smooth[next_drive_start_yardline-1]
			
        epa_result = np.append(epa_result,(EP_end-EP_start))

        if drive.pos_team == 'JAC':
            drive.pos_team = 'JAX'
        if opp_team == 'JAC':
            opp_team = 'JAX'		

        EPA_team[np.where(team_list==drive.pos_team)[0][0]] += ( EP_end - EP_start )

        EPA_team[np.where(team_list==opp_team)[0][0]] -= ( EP_end - EP_start )    

        next_drive_start_yardline = yardlinefromstr
		
bye_week = [9,11,8,10,7,9,9,13,7,11,10,4,9,10,5,5,8,8,6,9,5,8,11,10,4,8,11,5,8,6,13,9]

Estimated_wins = np.zeros(32)

for i in range(32):
    
    if week < bye_week[i]:
        game_count = week
    else:
        game_count = week - 1
      
    EPA_team[i] = EPA_team[i] / game_count *16
    Estimated_wins[i] = 16*(1/(1+(2.7128**(-(0.0085*EPA_team[i])))))
    
    print(team_list[i],EPA_team[i])

np.savetxt('EPA_2016.csv',EPA_team,delimiter=',')

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment