Beer Alcohol per Calorie (APC)¶

A friend found a nice table of beer alcohol by volume and calories so I thought I'd try something obvious. Maybe alcohol isn't what drinking beer is all about, but it's definitely also about calories. They don't call it a beer belly for nothing. So maybe APC should be a thing? If you've ever wondered about how alcohol and calories play together in your favorite brew, you may find the interactive plotly charts at the bottom of this notebook interesting.

In [1]:

pip install plotly

Requirement already satisfied: plotly in c:\program files\python\python 3.7.8\lib\site-packages (5.8.2)
Requirement already satisfied: tenacity>=6.2.0 in c:\program files\python\python 3.7.8\lib\site-packages (from plotly) (8.0.1)
Note: you may need to restart the kernel to use updated packages.

In [2]:

import pandas as pd
import lxml
from matplotlib import pyplot
import plotly.express as px

Thank you homebrewacademy.com for the simple to import table!

In [3]:

beer = pd.read_html('https://homebrewacademy.com/beer-alcohol-content-list/')

In [4]:

print(f'Total tables: {len(beer)}')

Total tables: 2

In [5]:

# Need the first table, and need to strip off the '%' symbol and make it a float
df = beer[0]
df['ABV'] = df['ABV'].str[:-1]
df['ABV'] = df['ABV'].astype('float')
df.head()

Out[5]:

	Brand	Calories	ABV
0	Abita Amber	128	4.5
1	Abita Golden	125	4.2
2	Abita Jockamo IPA	190	6.5
3	Abita Light	118	4.0
4	Abita Purple Haze	128	4.2

In [6]:

# Start with a simple plotly express plot
fig = px.scatter(df, x="Calories", y="ABV",  hover_data=['Brand'],
                width=400, height=400)
fig.show()

In [7]:

# But the residual isn't really what's important. How about a ratio?
df['apc'] = df['ABV']/df['Calories']
df = df.sort_values(['apc'],ascending=False)

In [8]:

# So which beers give you the most alcohol per calorie?
df.head(10)

Out[8]:

	Brand	Calories	ABV	apc
136	Michelob Ultra Pure Gold	85	4.3	0.050588
153	Natural Ice	130	5.9	0.045385
149	Molson Canadian 67	67	3.0	0.044776
33	Bud Ice	123	5.5	0.044715
193	Rolling Rock Green Light	83	3.7	0.044578
52	Corona Premier	90	4.0	0.044444
154	Natural Light	95	4.2	0.044211
133	Michelob Ultra	95	4.2	0.044211
37	Bud Light Platinum	137	6.0	0.043796
142	Miller Lite	96	4.2	0.043750

So it's mostly "light" and "ice" beers winning the "bang for the buck" test but the highest ABV beer is also in the top 30 (truncated for ease of display). Way to go Dogfish Head.

Maybe try something a little fancier...

I like quintiles for graphics because the middle value contains the median and then you get high, low, and extremes.. Quartiles are for boxplots.

In [10]:

pd.qcut(df["apc"], q=5, labels=False)
df['Quintile'] = pd.qcut(df["apc"], q=5, labels=False)
df['Quintile'] = df['Quintile'] + 1  # add one because zero index not helpful here
df.head(10)

Out[10]:

	Brand	Calories	ABV	apc	Quintile
136	Michelob Ultra Pure Gold	85	4.3	0.050588	5
153	Natural Ice	130	5.9	0.045385	5
149	Molson Canadian 67	67	3.0	0.044776	5
33	Bud Ice	123	5.5	0.044715	5
193	Rolling Rock Green Light	83	3.7	0.044578	5
52	Corona Premier	90	4.0	0.044444	5
154	Natural Light	95	4.2	0.044211	5
133	Michelob Ultra	95	4.2	0.044211	5
37	Bud Light Platinum	137	6.0	0.043796	5
142	Miller Lite	96	4.2	0.043750	5

Plotly's default color ramp from blue to yellow not the best because the largest quintile gets assigned yellow which is difficult to see. Try discrete color rainbow after taking making quintiles categorical.

In [11]:

df['Quintile'] = df['Quintile'].astype("string")
df = df.sort_values(['apc'],ascending=True)
# Now color data points by quintiles of ABV/Calories
fig = px.scatter(df, x="Calories", y="ABV",  hover_data=['Brand'], color="Quintile"
                 ,color_discrete_map={
                     "1" : "blue"
                     ,"2" : "green"
                     ,"3" : "yellow"
                     ,"4" : "orange"
                     ,"5" : "red"
                 }
                 ,width=700, height=700)
fig.update_layout(title_text="Alcohol per calorie by ABV and Calorie", title_x=0.5)
fig.show()

No surprise that the highest quintile APC are above where the regression line would be but that "Dogfish Head 120 Minute IPA" red dot at 18% is interesting. With that exception, the mass of the highest quintile ratio distribution is at the lower calorie range. So maybe there is something to light beers after all - just not taste!