Let's talk about beer!

Beer. Pivo. Cerveza. From ancient Persia to old-man pubs to the craft bars of 2020, it's been with us for quite some time. The Sumerians even had a goddess of beer! It is also the third most popular drink in the world. The first two are water and tea.

Beer taps in Bar Es, Keswick, England. (source: https://www.esbarltd.co.uk)

I found this amazing dataset on Kaggle (https://www.kaggle.com/rdoume/beerreviews), which contains about 1.5 million reviews from BeerAdvocate.

So let's have a look at what are the beers we brew, drink and like.

This dataset consists of 5840 breweries, 66055 beers and 104 beer styles.
Some of the styles could indeed have been merged into one, such as Russian and American Imperial Stouts, different sours could have been all grouped together and so forth. For a purpose of this blog, I left the beer styles intact as they were abstracted from BeerAdvocate.

First let's have a look at what are the most common beer styles in our dataset. Most beers were brewed in the style of American IPA (3611), American Pale Ale (3349) and American Amber / Red Ale (2511). Surprising to see Fruit Beers, Imperial IPAs and even Hefeweizen above Lagers and Stouts.

I'm sure you spotted English Bitter in the 8th position. It is author's personal opinion that these beers are not fit for human consumption. But the mean average rating for English Bitter is relatively high, 3.78 (out of 5). It seems that at least some BeerAdvocate users do indeed like warm beer with no head.

It would be interesting to monitor this over time and see for example the rising popularity of sour beers that has taken place over the past few years.

Now we know which are the most common beer styles to brew. But what are the most often reviewed styles?

Here we've got American IPA in the lead (117 586 reviews), then a big drop to second American Imperial IPA (85 977) and American Pale Ale in the third place (63 469). I'm pleased not to see English Bitter at the top 10 anymore.

Now let's look at our final top 10. The highest rated beer styles.

Here we've got American Wild Ale in the first position with the average ratings of 4.093, Gueze coming second with 4.086 and Quadrupel taking bronze with 4.07.

Interesting to see sour beers occupying 1st and 2nd (also 4th) and American and Russian Imperial Stouts right beside each other also.

I'm also chuffed to see Rye Beers making the cut at 10th position.

Now some food for thought. If we were to monitor these three lists (most commonly brewed styles, most commonly reviewed styles and highest rated styles) for let's say 20 years, what would it look like? Would the highest rated list look the same regardless the most often drunk and brewed styles or could we see people's palate changing (or pretending to do so) over the years?

If we break it down to single beers, the most often rated beer is 90 Minute IPA from Dogfish Head Brewery (American Imperial IPA, 3289 reviews), Old Rasputin Russian Imperial Stout (North Coast Brewing Co., Russian Imperial Stout, 3110 reviews) and Sierra Nevada Celebration Ale (Sierra Nevada Brewing Co., American IPA, 2999 reviews).

Altogether the top ten consists of 6 IPAs (Imperial or otherwise), 2 Imperial Stouts and 2 Ales.

Only one of the 10 most often rated beers drops ever so slightly under an average rating of 4.0 (Sierra Nevada Pale Ale, avg, rating 3.9951).

Now we can stop beating around the bush: The beers with the highest alcohol content.

No. 1 & 2 are Schorschbräu Schorschbock 57% and Schorschbräu Schorschbock 43%, both Eisbocks from a brewery called Schorschbräu. Between these two and a 40% Eisbock from the same brewery we've got Sink The Bismarck! at the third position, an Imperial IPA from the famous Scottish punks BrewDog at 41%.

What about the other side of the spectrum, what are the best non-alcoholic beers? They are in general rated very poorly, but the best beers with the abv of 0.5 or less are Clausthaler Golden Amber (Binding-Brauerei AG, mean average rating of 3.39), Clausthaler Premium (Binding-Brauerei AG, 3.05) and Erdinger Weissbier Alkoholfrei (Erdinger Weissbrau, 2.67). It appears that German speaking countries with their culture of workmen having lunch-break beers nailed the low alcohol ones. For reasons explained later I only included beers with at least 30 reviews to this list.

Now we know what are the most alcoholic beers in our dataset. What is the overval distribution of alcohol content though? For better clarity, the graph below only shows values between 0 and 15 (% abv) as there are very few beers going above those values.

And it is clear that the most beers have abour 5% abv, which makes sense as most of the "casually drunk" pub ales and lagers would fall somewhere around there.

When we try to find the "best" beer, things start to get a little tricky. There are 751 beers with the average rating of 5.0.

This will be down to the fact that many of those beers will have only very few reviews, but all of them giving the highest score.

The way to get around it was to count how many times was each beer reviewed and then filter out beers with only a few reviews.

I decided the threshold to be 50, so these are the top 5 rated beers that have been rated more than 50 times:

1. Armand'4 Oude Geuze Lente (Spring) - Gueuze from Brouwerij Drie Fonteinen, average rating 4.73

2. Hoppy Birthday, American Pale Ale, Alpine Beer Company, 4.68

3. Geuze Cuvée J&J (Joost En Jessie) Blauw (Blue), Gueuze, Brouwerij Drie Fonteinen, 4.63

4. Citra DIPA, American Double / Imperial IPA, Kern River Brewing Company, 4.63

5. Cantillon Blåbær Lambik, Fruit Lambic, Brasserie Cantillon, 4.63

This table shows what seems to be the general consensus when it comes to beer: Belgium is where it's at.

For the same reason as above, it'd be misleading to simply search for the "best" brewery, so I looked for the highest rated breweries with 3 or more beers under their belt.

The top one is Cheriton Brewery with average ratings of 4.67, followed by Brauerei Zehendner GmbH (4.61) and Brass Monkey Brewing Co. (4.60).

The Alchemist in the 4th position (4.58) and finally no. 5 is Brouwerij Westvleteren (4.54).

The dataset I'm looking at here consists of single reviews by BeerAdvocate users. Let's have a look at them too!

There are 33 387 unique users in this dataset, but only 1870 of them are very active with 200 or more submitted reviews.

If we look at reviews submitted by single users, we can gain quite a bit of knowledge about them. Do they use the whole 1-5 scale of possible ratings? Are they lovers/haters with 1 or 5 being the only options? What styles of beers do they drink?

Let's look at a user named hoppymeal, who submitted whopping 517 reviews. Their mean average overall rating is 4.08. Beside the main overall rating I've been using so far, BeerAdvocate also lets its users rate different beer aspects separately. Hoppymeal seems to be more strict with aroma and appearance (mean average ratings 3.97 for both) than they are with palate and taste (4.12 and 4.04).

The two graphs below show the distribution of hoppymeal's overall ratings and also of the alcohol content in beers reviewed by them.

Before we turn the music off and cover the taps, I got three more plots for you.

First one shows how many times are beers usually reviewed (only shows up to 400 reviews) and the second one shows how are the beers usually rated.

We can now join these two graphs together and see that the better the beer, the more reviews it tends to have.

Which makes sense as when you see a high-rated beer, you are more likely to give it a go and then rate it yourself. You will see similar shaped graphs if you look at let's say IMDb datasets with movie ratings.

No matter what beers you're into, whether you're sipping Imperial Stouts, appreciating Lambics, quenching your thirst with Lagers after your weekly football sesh with the lads or god forbid enjoying the silage aroma of English Bitters, don't forget that the beer world is big and there's always more to explore, so don't be afraid to try something new every now and again. And most of all, don't be a snob.

Cheers!