Comparison of Michelin and Gault Millau ratings

There are two well known rating systems for high end restaurants in Europe, Michelin and Gault Millau. If you regularly look up both ratings for a specific restaurant you can notice the obvious correlation between the two. Which made me wonder, what the average Gault Millau score of the three star ratings of Michelin would be.

Additionally I've read that Gault Millau purely rates the quality of the food while Michelin is focused on the whole experience. So there will probably some restaurants with more focus on one or the other that show up as outliers. 

To answer this question I wrote a script (details below) to scrape both websites and return the score of all restaurants they rated. I limited this to Belgian restaurants to keep this project quick and easy, since i'm just doing this for fun and out of personal interest (I live in Belgium).

The results can be seen in the plots below. On the box plot we can see the average scores of each star rating. The 1* restaurants have an average score of 15,44 (note the huge variance in this group, 12-18). 2* restaurants have an average of 17,52 and for what it's worth the two 3* restaurants we have, have an average score of 18,75. Additionally there is a group of restaurants that did receive a Gault Millau rating but don't have any stars, the 0 group, they score 13,16 on average. 

To get a different view, I plotted all the points individually. I added a little jitter to the scores so they don't overlap. The distinct banded pattern is simply because Gault Millau only scores in half points so this is not really a continuous variable.

There's clearly some outliers here. So, I selected the lowest and highest scoring restaurants of each group to see which restaurants they are. Of course the 0 group has a lot of low scorers since it's range restricted because restaurants below a certain score don't even get published. The 3* group is so small that I excluded it as well (only two restaurants: Zilte and Boury).

 Remarkably high Gault Millau scores. 

 Remarkably high Michelin scores.                 

Presumable the high (Gault Millau) scorers of each (Michelin) group would be restaurants that outperform their peers in the kitchen. On the other hand the low scorers of each group would offer amazing experiences relative to their peers. On the other hand, the reviewers could just be in disagreement. 

In the end, there's only one way to find out what they're really like.

Bon appetit :)

For those interested, the script is publicly available here: https://github.com/Martijn-DG/resto-rating

I wrote it in R since that's just the language I'm most familiar with. Please ignore the lack of comments and shitty variable naming. This really was just a quick and dirty job.

As always with web-scraping, don't run it too much, to be respectful of the two websites in question. I included an RData file with the scraping results you can load in, so you can just play with the resulting data yourself from there.