nflgame-python-for-american-football

Title: NFLGame - Python for American Football Date: 2013-03-14 23:37 Author: Michael Category: Python Tags: iPython, NFL, Analysis Slug: nflgame-python-for-american-football Status: published

NFLgame is a Python package that allows programmatic access to NFL data. It is a super fun way to get stats about your favorite NFL player or team. My plan is more sinister as I attempt to leverage this in my quest to win a fantasy football league.

The Github repo is located here.

Download this Notebook


:::python
import nflgame

Lets have a look at some basic stats for the last 3 games of the NFL Regular season

First we’ll have a look at the top 25 rushers for weeks 15, 16, 17. Awesome that we get summary stats right out of the box, right?


:::python
games = nflgame.games(2012, week=[15, 16, 17])
players = nflgame.combine(games)
for p in players.rushing().sort("rushing_yds").limit(25):
    print p, p.rushing_yds

A.Peterson 497 A.Morris 378 M.Lynch 324 D.Williams 322 C.Spiller 300 J.Charles 289 A.Foster 276 M.Forte 260 V.Ballard 252 K.Moreno 240 B.Pierce 232 D.Martin 220 S.Jackson 206 C.Johnson 206 D.McFadden 200 D.Murray 197 R.Bush 195 M.Ingram 182 S.Ridley 181 S.Greene 180 R.Wilson 179 F.Gore 179 M.Leshoure 158 R.Rice 150 D.Wilson 147

Next, we’ll look at the top 25 running backs and the total amount of touchdowns for weeks 15, 16, 17 using the players variable we created above.


:::python
for p in players.touchdowns().sort("rushing_tds").limit(25):
        print p, p.rushing_tds

A.Morris 6 M.Tolbert 5 R.Wilson 4 C.Wells 3 M.Forte 2 R.Grant 2 M.Turner 2 M.Ingram 2 A.Peterson 2 T.Richardson 2 K.Moreno 2 J.Hester 2 M.Leshoure 2 D.Williams 2 M.Lynch 2 S.Ridley 2 D.Woodhead 2 C.Johnson 2 S.Greene 2 D.Lewis 1 A.Dalton 1 B.Green-Ellis 1 A.Rodgers 1 D.Harris 1 D.Wilson 1

Fumbles are bad in fantasy football, umkay. Lets see the top 25 fumblers for the same weeks…


:::python
for p in players.fumbles().sort("fumbles_tot").limit(25):
    print p, p.fumbles_tot

P.Rivers 4 C.Kaepernick 4 R.Tannehill 3 M.Sanchez 3 N.Foles 2 A.Dalton 2 J.Cutler 2 M.Colston 2 J.Freeman 2 D.Amendola 2 C.Ponder 2 C.Henne 2 T.Holliday 2 M.Stafford 2 R.Fitzpatrick 2 D.Murray 2 A.Brown 2 D.Moore 2 J.McKnight 2 T.Yates 2 C.Harbor 1 J.Maclin 1 C.Thornton 1 A.Green 1 M.Forte 1

Now lets have a go with a basic ‘player on player’ comparison and see if we can dig up any help from the past

In my first week of the 2012 Fantasy Football season I had the RB match up between Marshawn Lynch and Frank Gore. My team started M. “Beast Mode” Lynch and I came up a bit short… could the outcome have been different with some analysis? I had other RB’s on the bench.

M. Lynch, player ID: 00-0025399

F. Gore, player ID: 00-0023500

The player can be looked up by name or player ID. Iterate through the list of player in for loop and print p.playerid or print p.name … or just look in players.json and use control-f , then start typing a players name to find all the available fields for each player.


 :::python
 week1 = nflgame.games(2012, week=1)
    players = nflgame.combine(week1)
    for p in players.rushing().sort("rushing_yds").limit(25):
        if p.playerid == "00-0025399" or p.playerid == "00-0023500":
            print "--------"
            print p.player, p.formatted_stats()

    print "-------"

Frank Gore (RB, SF) rushing_lngtd: 23, rushing_tds: 1, rushing_twopta: 0, rushing_lng: 23, rushing_yds: 112, rushing_att: 16, rushing_twoptm: 0, receiving_tds: 0, receiving_lng: 1, receiving_rec: 1, receiving_twopta: 0, receiving_yds: 1, receiving_lngtd: 0, receiving_twoptm: 0

Marshawn Lynch (RB, SEA) rushing_lngtd: 0, rushing_tds: 0, rushing_twopta: 0, rushing_lng: 11, rushing_yds: 85, rushing_att: 21, rushing_twoptm: 0, receiving_tds: 0, receiving_lng: 7, receiving_rec: 2, receiving_twopta: 0, receiving_yds: 12, receiving_lngtd: 0, receiving_twoptm: 0, fumbles_trcv: 0, fumbles_tot: 1, fumbles_rcv: 0, fumbles_yds: 0, fumbles_lost: 0

Lets have a look at three, first game, match-ups between the 49’ers and the Seahawks.


 :::python
 game12 = nflgame.one(2012, 7, "SF", "SEA")
    print game12.nice_score()
    print '*****'
    print "2012 Away team: %s" % (game12.away)
    # convert the named tuple to dict and iterate with items
    for i, v in game12.stats_away._asdict().items():
        print i, v
    print '-------'
    print "2012 Home team: %s" %game11.home
    # convert the named tuple to dict and iterate with items
    for i, v in game11.stats_home._asdict().items():
        print i, v

    print '-------'

SF (13) vs. SEA (6) ***** 2012 Away team: SEA first_downs 13 total_yds 251 passing_yds 115 rushing_yds 136 penalty_cnt 3 penalty_yds 20 turnovers 1 punt_cnt 4 punt_yds 194 punt_avg 31 pos_time 27:59 ——- 2012 Home team: SF first_downs 12 total_yds 209 passing_yds 124 rushing_yds 85 penalty_cnt 9 penalty_yds 102 turnovers 0 punt_cnt 5 punt_yds 298 punt_avg 54 pos_time 31:07 ——-


:::python
game11 = nflgame.one(2011, 1, "SF", "SEA")
    print game11.nice_score()
    print '*****'
    print "2011 Away team: %s" % (game11.away)
    # convert the named tuple to dict and iterate with items
    for i, v in game11.stats_away._asdict().items():
        print i, v
    print '-------'
    print "2011 Home team: %s" %game11.home
    # convert the named tuple to dict and iterate with items
    for i, v in game11.stats_home._asdict().items():
        print i, v

    print '-------'
    week_11 = nflgame.games(2011, week=1)
    players = nflgame.combine(week_11)
    for p in players.rushing().sort("rushing_yds").limit(25):
        if p.playerid == "00-0025399" or p.playerid == "00-0023500":
            print "--------"
            print p.player, p.formatted_stats()

    print "-------"

SF (33) vs. SEA (17) ***** 2011 Away team: SEA first_downs 18 total_yds 219 passing_yds 155 rushing_yds 64 penalty_cnt 11 penalty_yds 72 turnovers 3 punt_cnt 7 punt_yds 342 punt_avg 33 pos_time 28:53 ——- 2011 Home team: SF first_downs 12 total_yds 209 passing_yds 124 rushing_yds 85 penalty_cnt 9 penalty_yds 102 turnovers 0 punt_cnt 5 punt_yds 298 punt_avg 54 pos_time 31:07 ——- ——– Frank Gore (RB, SF) rushing_lngtd: 0, rushing_tds: 0, rushing_twopta: 0, rushing_lng: 16, rushing_yds: 59, rushing_att: 22, rushing_twoptm: 0, receiving_tds: 0, receiving_lng: 12, receiving_rec: 3, receiving_twopta: 0, receiving_yds: 19, receiving_lngtd: 0, receiving_twoptm: 0 ——-


 :::python
 game10 = nflgame.one(2010, 1, "SEA", "SF")
    print game10.nice_score()
    print '*****'
    print "2010 Away team: %s" % (game10.away)
    # convert the named tuple to dict and iterate with items
    for i, v in game10.stats_away._asdict().items():
        print i, v
    print '-------'
    print "2010 Home team: %s" %game10.home
    # convert the named tuple to dict and iterate with items
    for i, v in game10.stats_home._asdict().items():
        print i, v

    print '-------'
    week_10 = nflgame.games(2010, week=1)
    players = nflgame.combine(week_10)
    for p in players.rushing().sort("rushing_yds").limit(25):
        if p.playerid == "00-0025399" or p.playerid == "00-0023500":
            print "--------"
            print p.player, p.formatted_stats()

    print "-------"

SEA (31) vs. SF (6) ***** 2010 Away team: SF first_downs 14 total_yds 263 passing_yds 214 rushing_yds 49 penalty_cnt 8 penalty_yds 60 turnovers 2 punt_cnt 6 punt_yds 249 punt_avg 35 pos_time 32:45 ——- 2010 Home team: SEA first_downs 14 total_yds 242 passing_yds 165 rushing_yds 77 penalty_cnt 5 penalty_yds 35 turnovers 1 punt_cnt 5 punt_yds 207 punt_avg 36 pos_time 27:15 ——- ——- *******

Conclusion

So we took a look at the team stats for the last three, first game, match-ups to see if we could uncover any actionable information. We see that Frank Gore played in 2011 but not in 2010. Marshawn Lynch did not play in 2011 nor in 2010. Not much to go on there…

Rushing yards were low in the first two games, which may have suggested that the third game would have been similar. However, the 2012 match-up had both teams with over 100 yards rushing which may or may not be an outlier. Now to be thorough, I should compare the running backs I had on the bench in a similar manner. Then, ultimately, make a reasonable guess.

Not all is lost in this simple analysis, I learned a bit about tuples in the process. I can also say that I am learning while playing which is ‘win-win.’ Perhaps the trophy on my mantle is a bit pre-mature at this point, as I really need to learn statistics (stay tuned for more on that front).

Here is a neat little tidbit about Named tuples that I learned during this. The team stats are created in game.py and are delivered as a namedtuple(). The ._asdict() method on Named Tuples will return an ordered dict as described here. Once you have a dict, simply use (chain on) the .items() method to iterate over the key and value. If you want the keys or values by themselves, use the .keys() or .values() methods respectively.

I plan on using this extensively next season. In preparation, I have been learning statistics and the Pandas data analysis python library.

Note: Everything I work on in this space is using iPython Notebook and will be published for you to run as well…