About eight weeks ago I saw John Capobianco and Tim Bert were holding an online meeting about trying to pull down some NCAA football data using an API. I, myself, finally finished up the last certification exam I’ve had on my plate shortly thereafter. Since then, I’ve taken to doing a little bit of DataCamp and CloudAcademy each day.
DataCamp has a python track that teaches you the basics of python while adding in some packages a data scientist may use along the way, specifically, numpy and pandas. I’m square in the middle of DataCamp’s ‘intermediate python’ course.
Then, like a shower whose water never gets warm, I thought to myself, why not try and do some stuff with NBA data while going through the examples and practice during the course. So here we are 🙂
As the title says, this is going to be ‘basic’ as i’m just beginning and truth be told, i’ll prob do some non best practices going forward in this post. This is what learning in public looks like.
Getting Some Data
If I’m going to parse some data using what I’ve been presented thus far in my training, I need some data. In my current lesson we are doing basic data parsing with pandas series objects and pandas dataframe objects. I took to the googleverse and found an interesting github repo. This python package was created to make it easier to interact with stats.nba.com APIs. To install:
pip install nba_api pandas requests
From here it’s all about figuring out a bit from the documentation found on the github I linked above. Jimmy Butler is my current favorite NBA basketball player so the first thing I need to find out is his PLAYER_ID so I can use it to get further info on him.
>>> from nba_api.stats.static import players
>>> players.find_players_by_full_name('jimmy butler')
[{'id': 202710, 'full_name': 'Jimmy Butler', 'first_name': 'Jimmy', 'last_name': 'Butler', 'is_active': True}]
From the output, we can see that Jimmy Butler’s ‘id’ is 202710. I’ll use this when making my next call:
>>> from nba_api.stats.endpoints import playercareerstats
>>> Jimmy = playercareerstats.PlayerCareerStats(player_id=202710)
>>> print(type(Jimmy))
<class 'nba_api.stats.endpoints.playercareerstats.PlayerCareerStats'>
So, at this point we are almost there, just a little bit more mangling and we will get to the types of objects I need…I’m going to use the get_data_frames() function included in the nba_api and select the first table in this object with the ‘[0]’ and assign this to the variable ‘jimmy_panda’:
>>> jimmy_panda = Jimmy.get_data_frames()[0]
>>> print(type(jimmy_panda))
<class 'pandas.core.frame.DataFrame'>
>>> print(jimmy_panda)
PLAYER_ID SEASON_ID LEAGUE_ID TEAM_ID TEAM_ABBREVIATION PLAYER_AGE GP GS MIN ... OREB DREB REB AST STL BLK TOV PF PTS
0 202710 2011-12 00 1610612741 CHI 22.0 42 0 359.0 ... 23 33 56 14 11 5 14 20 109
1 202710 2012-13 00 1610612741 CHI 23.0 82 20 2134.0 ... 136 192 328 115 78 31 62 97 705
2 202710 2013-14 00 1610612741 CHI 24.0 67 67 2591.0 ... 87 243 330 175 127 36 102 106 878
3 202710 2014-15 00 1610612741 CHI 25.0 65 65 2513.0 ... 114 265 379 212 114 36 93 108 1301
4 202710 2015-16 00 1610612741 CHI 26.0 67 67 2474.0 ... 79 279 358 321 110 43 132 124 1399
5 202710 2016-17 00 1610612741 CHI 27.0 76 75 2809.0 ... 129 341 470 417 143 32 159 112 1816
6 202710 2017-18 00 1610612750 MIN 28.0 59 59 2164.0 ... 79 235 314 288 116 24 108 78 1307
7 202710 2018-19 00 1610612750 MIN 29.0 10 10 361.0 ... 16 36 52 43 24 10 14 18 213
8 202710 2018-19 00 1610612755 PHI 29.0 55 55 1824.0 ... 105 185 290 220 99 29 81 93 1002
9 202710 2018-19 00 0 TOT 29.0 65 65 2185.0 ... 121 221 342 263 123 39 95 111 1215
10 202710 2019-20 00 1610612748 MIA 30.0 58 58 1959.0 ... 106 280 386 350 103 32 127 81 1157
11 202710 2020-21 00 1610612748 MIA 31.0 52 52 1745.0 ... 94 265 359 369 108 18 109 71 1116
12 202710 2021-22 00 1610612748 MIA 32.0 57 57 1931.0 ... 102 234 336 312 94 27 121 88 1219
[13 rows x 27 columns]
Alright. We made it.
Finding Out Some Basic Info About our Data
Alright, above it says at the end of the output we have 13 rows x 27 columns. We definitely see the 13 rows but we don’t see anywhere close to 27 columns. Let’s see how we can see what the 27 column headers are, to do this i’m going to dtypes to see the info about each column in the table as shown:
>>> jimmy_panda.dtypes
PLAYER_ID int64
SEASON_ID object
LEAGUE_ID object
TEAM_ID int64
TEAM_ABBREVIATION object
PLAYER_AGE float64
GP int64
GS int64
MIN float64
FGM int64
FGA int64
FG_PCT float64
FG3M int64
FG3A int64
FG3_PCT float64
FTM int64
FTA int64
FT_PCT float64
OREB int64
DREB int64
REB int64
AST int64
STL int64
BLK int64
TOV int64
PF int64
PTS int64
dtype: object
So this table we have is kind of exciting, at least for a first go at this, am I right?! We can now select to print off only certain columns in the output or slice only specific years of Jimmy Butler’s career or both. One way to do this is to use pandas and make sure the object we are dealing with is a DataFrame, allowing us to use all the options assoicated with this object type. Let’s see how we can accomplish this:
>>> pd.DataFrame(data=Jimmy.get_data_frames()[0], columns=['SEASON_ID', 'PTS', 'AST'])
SEASON_ID PTS AST
0 2011-12 109 14
1 2012-13 705 115
2 2013-14 878 175
3 2014-15 1301 212
4 2015-16 1399 321
5 2016-17 1816 417
6 2017-18 1307 288
7 2018-19 213 43
8 2018-19 1002 220
9 2018-19 1215 263
10 2019-20 1157 350
11 2020-21 1116 369
12 2021-22 1219 312
###
Above we basically changed the Jimmy object into the Data
Frame object we want to play with. It's the same as doing
what's below because we've already assigned 'jimmy_panda' and
made sure it was a Pandas Data Frame
###
>>> pd.DataFrame(data=jimmy_panda, columns=['SEASON_ID', 'PTS', 'AST'])
SEASON_ID PTS AST
0 2011-12 109 14
1 2012-13 705 115
2 2013-14 878 175
3 2014-15 1301 212
4 2015-16 1399 321
5 2016-17 1816 417
6 2017-18 1307 288
7 2018-19 213 43
8 2018-19 1002 220
9 2018-19 1215 263
10 2019-20 1157 350
11 2020-21 1116 369
12 2021-22 1219 312
Or, an even simpler way to select columns, since we know ‘jimmy_panda’ is of the data type we need:
>>> print(jimmy_panda[['SEASON_ID', 'PTS', 'AST']])
SEASON_ID PTS AST
0 2011-12 109 14
1 2012-13 705 115
2 2013-14 878 175
3 2014-15 1301 212
4 2015-16 1399 321
5 2016-17 1816 417
6 2017-18 1307 288
7 2018-19 213 43
8 2018-19 1002 220
9 2018-19 1215 263
10 2019-20 1157 350
11 2020-21 1116 369
12 2021-22 1219 312
Alright, that was fun, let’s see how we can slice a specific year (row) out of this chart (how did i figure out how to do this you ask, I checked the pandas loc docs):
>>> jimmy_panda.loc[jimmy_panda.index[[12]], ['SEASON_ID', 'PTS', 'AST']]
SEASON_ID PTS AST
12 2021-22 1219 312
Let’s Compare
Now that we’ve seen we can pull out specific columns, rows and the like, we can now decide to do so based on certain thresholds. For example, it maybe useful to know only the years Jimmy Butler shot over 30% from 3-point range.
The first step we need to do is figure out which column we will be using to do our comparison, specifically, which column represents 3 point percentage. We can scroll up and look at the output of our dtypes command or do it again:
>>> jimmy_panda.dtypes
PLAYER_ID int64
SEASON_ID object
LEAGUE_ID object
TEAM_ID int64
TEAM_ABBREVIATION object
PLAYER_AGE float64
GP int64
GS int64
MIN float64
FGM int64
FGA int64
FG_PCT float64
FG3M int64
FG3A int64
FG3_PCT float64
FTM int64
FTA int64
FT_PCT float64
OREB int64
DREB int64
REB int64
AST int64
STL int64
BLK int64
TOV int64
PF int64
PTS int64
dtype: object
Alright, we can see we will be working with the ‘FG3_PCT’ which is the object type float. Next let’s check out the specific values we will be comparing, specifically, we are going to check which years Jimmy Butler shot the 3-ball better than 30%.
>>> jimmy_panda.loc[:,"FG3_PCT"]
0 0.182
1 0.381
2 0.283
3 0.378
4 0.312
5 0.367
6 0.350
7 0.378
8 0.338
9 0.347
10 0.244
11 0.245
12 0.233
Name: FG3_PCT, dtype: float64
Alright, now to see which years he shot beyond 30%; to do this we will simply add the operator at the end of the statement above:
>>> jimmy_panda.loc[:,"FG3_PCT"] > 0.300
0 False
1 True
2 False
3 True
4 True
5 True
6 True
7 True
8 True
9 True
10 False
11 False
12 False
Name: FG3_PCT, dtype: bool
Every ‘True’ is when Jimmy shot better than 30% and every false is when he failed to do so. Next, i’ll show a series of commands. First, i’ll assign the variable FG3_is_good to the comparison and then use that variable as an index. Lastly I’ll use the value_counts() function to see how many ‘True’ or ‘False’ in total, so you can see how many years he was better than 30% and how many years he was not.
>>> FG3_is_good = jimmy_panda.loc[:,"FG3_PCT"] > 0.300
>>> print(type(FG3_is_good))
<class 'pandas.core.series.Series'>
>>> jimmy_panda[FG3_is_good]
PLAYER_ID SEASON_ID LEAGUE_ID TEAM_ID TEAM_ABBREVIATION PLAYER_AGE GP GS MIN ... OREB DREB REB AST STL BLK TOV PF PTS
1 202710 2012-13 00 1610612741 CHI 23.0 82 20 2134.0 ... 136 192 328 115 78 31 62 97 705
3 202710 2014-15 00 1610612741 CHI 25.0 65 65 2513.0 ... 114 265 379 212 114 36 93 108 1301
4 202710 2015-16 00 1610612741 CHI 26.0 67 67 2474.0 ... 79 279 358 321 110 43 132 124 1399
5 202710 2016-17 00 1610612741 CHI 27.0 76 75 2809.0 ... 129 341 470 417 143 32 159 112 1816
6 202710 2017-18 00 1610612750 MIN 28.0 59 59 2164.0 ... 79 235 314 288 116 24 108 78 1307
7 202710 2018-19 00 1610612750 MIN 29.0 10 10 361.0 ... 16 36 52 43 24 10 14 18 213
8 202710 2018-19 00 1610612755 PHI 29.0 55 55 1824.0 ... 105 185 290 220 99 29 81 93 1002
9 202710 2018-19 00 0 TOT 29.0 65 65 2185.0 ... 121 221 342 263 123 39 95 111 1215
>>> FG3_is_good.value_counts()
True 8
False 5
Name: FG3_PCT, dtype: int64
One funny thing that jumped out at me is that Jimmy has been unable to shoot better than 30% from three after the age of 30. Interesting. Furthermore, we can easily conclude Jimmy has shot better than 30% from 3 61% of the seasons he’s played thus far.
Playoff Jimmy has been said to be a thing. One aspect of this is that he shoots better from three in the playoffs as well. We can use the same format as we did above to quickly figure out his percentage. Jimmy Butler’s playoff stats happen to be [2] index of the original ‘Jimmy’ object we pulled down using the nba_api:
>>> print(Jimmy.get_data_frames()[2])
PLAYER_ID SEASON_ID LEAGUE_ID TEAM_ID TEAM_ABBREVIATION PLAYER_AGE GP GS MIN FGM FGA FG_PCT ... FTM FTA FT_PCT OREB DREB REB AST STL BLK TOV PF PTS
0 202710 2011-12 00 1610612741 CHI 22.0 3 0 4.0 0 0 0.000 ... 0 0 0.000 0 0 0 0 0 0 0 1 0
1 202710 2012-13 00 1610612741 CHI 23.0 12 12 490.0 50 115 0.435 ... 45 55 0.818 9 53 62 32 15 6 16 26 160
2 202710 2013-14 00 1610612741 CHI 24.0 5 5 218.0 22 57 0.386 ... 18 23 0.783 6 20 26 11 6 0 3 13 68
3 202710 2014-15 00 1610612741 CHI 25.0 12 12 506.0 94 213 0.441 ... 59 72 0.819 18 49 67 38 29 9 21 27 275
4 202710 2016-17 00 1610612741 CHI 27.0 6 6 239.0 46 108 0.426 ... 38 47 0.809 9 35 44 26 10 5 15 10 136
5 202710 2017-18 00 1610612750 MIN 28.0 5 5 170.0 28 63 0.444 ... 15 18 0.833 3 27 30 20 4 1 5 9 79
6 202710 2018-19 00 1610612755 PHI 29.0 12 12 421.0 79 175 0.451 ... 63 72 0.875 22 50 72 62 18 7 22 20 233
7 202710 2019-20 00 1610612748 MIA 30.0 21 21 806.0 144 295 0.488 ... 164 191 0.859 46 90 136 127 41 14 59 37 467
8 202710 2020-21 00 1610612748 MIA 31.0 4 4 154.0 19 64 0.297 ... 16 22 0.727 6 24 30 28 5 1 9 6 58
9 202710 2021-22 00 1610612748 MIA 32.0 17 17 629.0 166 328 0.506 ... 111 132 0.841 41 84 125 78 35 11 25 25 466
Now I can quickly assign this to a variable so I can interact with it as a pandas object.
>>> Playoff_Jimmy = Jimmy.get_data_frames()[2]
>>> Playoff_Jimmy.loc[:,"FG3_PCT"] > 0.300
0 False
1 True
2 False
3 True
4 False
5 True
6 False
7 True
8 False
9 True
Name: FG3_PCT, dtype: bool
>>> Playoff_Jimmy.loc[:,"FG3_PCT"]
0 0.000
1 0.405
2 0.300
3 0.389
4 0.261
5 0.471
6 0.267
7 0.349
8 0.267
9 0.338
Name: FG3_PCT, dtype: float64
Kind of funny that Butler has a perfect True False pattern throughout his playoff career. Also, as with anything in IT there are at least 25 ways to do the same thing, the more you learn. Above I was able to pull the “FG3_PCT” for each row. Here is another way to do the same thing using a for loop:
>>> for column, row in Playoff_Jimmy.iterrows():
... print(str(row['SEASON_ID']) + ' 3ptfg%: ' + str(row['FG3_PCT']))
...
2011-12 3ptfg%: 0.0
2012-13 3ptfg%: 0.405
2013-14 3ptfg%: 0.3
2014-15 3ptfg%: 0.389
2016-17 3ptfg%: 0.261
2017-18 3ptfg%: 0.471
2018-19 3ptfg%: 0.267
2019-20 3ptfg%: 0.349
2020-21 3ptfg%: 0.267
2021-22 3ptfg%: 0.338
Conclusion
Well, if you read this far you realized we didn’t do anything mind bending. I suppose writing this out and doing the same type of practice activities I’m doing in my DataCamp course on some data I pulled down myself helped me learn quite a bit more than simply going only through the course.
I’ve always enjoyed doing data parsing in Linux. I’ve mentioned in previous blog posts how much I’ve enjoyed doing log parsing on the Linux shell using cut, grep, awk, uniq and sort to break down logs. This data science stuff seems to tickle a bit of that same excitement. I’m also excited to continue this project. Specifically, making visualizations should be fun, and i’m sure i’ll check back in here once I make it that far. I’ve seen some people plot shot charts as well, which would be a fun little side quest. I imagine a site like statmuse.com does a lot of what’s shown here behind the scenes when you ask a question.
See y’all around the bend.
The Art of Network Engineering 