Frolian's blog standing on the shoulders of PyGiants

• About
• Archive

Brewdog Beers Statistics

Thu, 22 Aug 2019

Food Dataviz

I've recently started having an interest in beermaking.

Watching a very interesting presentation from EMF Camp 2016 called Beer for everyone, I'm now trying to reproduce the setup used by the author to make beer (spoiler: it involves using a classic percolator-type coffee maker).

from IPython.display import IFrame

IFrame("https://media.ccc.de/v/emf2016-269-beer-for-everyone/oembed", width=1024, height=576)

In this process, I've come across the Brewdog catalogue of beer recipes for the home-brewer, called DIY DOG.

Looking at the recipes, I've realized that a few standard parameters that can be compared between recipes are the mashing temperature and time as well as the fermentation temperature. Since the setup presented in Beer for everyone does not indicate strong rules on these parameteres, I'm trying to answer the following question: How sensitive is the brewery process to these parameters?.

One way to answer this is to chart all this information for every recipe in the book. However, the PDF file is not easily machine-readable. Luckily for me, someone already did the hard work of making it machine readable and built a website using it, called Brewdog recipes.

So the plan of action is two-fold: get all the data from this website and then chart the three parameters:

• mashing temperature
• mashing time
• fermentation temperature

Scraping the data

We start with the index pages and get the links to the individual recipes. The index pages all have URLs like https://brewdogrecipes.com/pXX where XX is a number.

import requests
from bs4 import BeautifulSoup
import pandas as pd

dfs = []
for index in range(14):
    index_url = "https://brewdogrecipes.com/p{}".format(index)
    r = requests.get(index_url)
    soup = BeautifulSoup(r.text, 'html.parser')
    recipes = soup.find('div', class_='recipe-list').find_all('article')
    recipe_urls = [recipe.find('a').attrs['href'] for recipe in recipes]
    recipe_names = [recipe.find('h2').text for recipe in recipes]
    dfs.append(pd.DataFrame({'names': recipe_names, 'urls': recipe_urls }))

df = pd.concat(dfs)

Let's now work with the dataframe we have obtained. Let's make sure there are 249 recipes and then let's extract the data for each recipe.

recipe_df = df.drop_duplicates()

recipe_df.describe()

	names	urls
count	249	249
unique	249	249
top	#244 AB:21	https://brewdogrecipes.com/recipes/b-sides-whi...
freq	1	1

import tqdm

series = []
for recipe_url, name in zip(tqdm.tqdm(recipe_df.urls), recipe_df.names):
    r = requests.get(recipe_url)
    soup = BeautifulSoup(r.text, 'html.parser')
    tag = [item for item in soup.find_all('div', class_='panel panel-brewdog') if item.find(lambda tag: tag.find('h3') and tag.find('h3').text == 'METHOD/TIMINGS') is not None][0]
    items = [item.text for item in tag.find_all('td')]
    try:
        values = {'beer_name': name, 'mash_temp': items[0], 'mash_time': items[2], 'fermentation_temp': items[3]}
    except IndexError:
        print('could not retrieve data for url: {}'.format(recipe_url))
    series.append(pd.Series(values))
    
df_parameters = pd.DataFrame(series)

 21%|█████████████████▏                                                               | 53/249 [00:14<00:53,  3.69it/s]

could not retrieve data for url: https://brewdogrecipes.com/recipes/bramling-x

 35%|████████████████████████████▋                                                    | 88/249 [00:23<00:41,  3.88it/s]

could not retrieve data for url: https://brewdogrecipes.com/recipes/ab-19

 82%|█████████████████████████████████████████████████████████████████▏              | 203/249 [00:55<00:12,  3.66it/s]

could not retrieve data for url: https://brewdogrecipes.com/recipes/no-label

100%|████████████████████████████████████████████████████████████████████████████████| 249/249 [01:07<00:00,  3.70it/s]

df_parameters.head()

	beer_name	mash_temp	mash_time	fermentation_temp
0	#1 PUNK IPA 2007 - 2010	65°C	75mins	19°C
1	#2 PUNK IPA 2010 - CURRENT	66°C	75mins	19°C
2	#3 THE PHYSICS	65°C	75mins	19°C
3	#4 RIPTIDE	65°C	75mins	20°C
4	#5 HOP ROCKER	65°C	75mins	10°C

We have all the data we need, let's now get to work with the plotting part.

Plotting the data

First, we have to convert the text data to numerical data.

df = df_parameters.copy()

df['mash_temp'] = df['mash_temp'].str.extract('(.*:?)°C').astype(float)
df['mash_time'] = df['mash_time'].str.extract('(.*:?)mins').astype(float)
df['fermentation_temp'] = df['fermentation_temp'].str.extract('(.*:?)°C').astype(float)

df.head()

	beer_name	mash_temp	mash_time	fermentation_temp
0	#1 PUNK IPA 2007 - 2010	65.0	75.0	19.0
1	#2 PUNK IPA 2010 - CURRENT	66.0	75.0	19.0
2	#3 THE PHYSICS	65.0	75.0	19.0
3	#4 RIPTIDE	65.0	75.0	20.0
4	#5 HOP ROCKER	65.0	75.0	10.0

Let's use this data exploration as an excuse to try out Altair, a relatively recent visualization API. First, let's examine the distribution of each parameter using simple scatter plots.

import altair as alt

alt.Chart(df).mark_point().encode(
    x='mash_temp',
) & alt.Chart(df).mark_point().encode(
    x='mash_time',
) & alt.Chart(df).mark_point().encode(
    x='fermentation_temp',
)

As we can see, the distribution of mash temperature tends to be clustered around 65°C. The mash time varies a bit more, though, while the fermentation temperature is roughly around normal outside temperature.

Let's see this as histogram charts:

alt.Chart(df).mark_bar().encode(
    alt.X("mash_temp:Q", bin=True),
    y='count()',
) & alt.Chart(df).mark_bar().encode(
    alt.X("mash_time:Q", bin=True),
    y='count()',
) & alt.Chart(df).mark_bar().encode(
    alt.X("fermentation_temp:Q", bin=True),
    y='count()',
)

Finally, we can also combine this and do a two-dimensional chart with tooltip:

alt.Chart(df).mark_point().encode(
    x='mash_temp',
    y='mash_time',
    tooltip=['beer_name', 'mash_temp', 'mash_time', 'fermentation_temp']
)

df.describe().loc['mean'].to_frame()

	mean
mash_temp	64.405622
mash_time	72.610442
fermentation_temp	23.151020

Apart from a few outliers then, an average beer can be obtained with the following parameters:

• mash temperature of 65°C
• mash time of 72 minutes
• fermentation temperature of 23°C

This data-based conclusion is roughly along the lines of what the tutorial for beer making from EMF2016 recommends.

With this satisfying thought in mind, I can now go back to the main task of brewing beer instead of doing data exploration.

This post was entirely written using the Jupyter Notebook. Its content is BSD-licensed. You can see a static view or download this notebook with the help of nbviewer at 20190822_BrewdogBeers.ipynb.

---

This article has been tagged with the following terms:

Food Dataviz

---

• 
• 
• 
• 
• 

---

Comments