How Much Do You Need to Pay for Good Wine?

Posted on Apr 29, 2020

One of the first challenges in finding a good bottle wine is the enormous range in price.  Most supermarkets have bottles for under $10.  At the other end of the scale, in October 2018 somebody paid $558,000 for one bottle of the 1945 Romanée Conti. Using the data collected around the world for sold wine, we can determine the average price of a good wine. 

The prospect of learning how to find good value wine can seem daunting.  Indeed, to truly master all the nuances of regional variations, differences in vintages, and the styles of the individual winemakers requires a full-time and lifelong commitment.  Very few of us will choose to devote ourselves to that task.  However, good wine can be delicious, so how can the average consumer identify good quality wines for a reasonable price? I’m an enthusiastic wine consumer myself, so I decided to take a look through and analyze the data there to see what I could tease out.

Professional Ratings

We’re lucky that some people really did decide to spend their whole lives studying wine.  On about half of the 5,748 wines they sell have reviews from one or more professional raters, and these wines range in price from $8.99 (for the Dominio de Eguren Protocolo Tinto 2017) to $4,999.97 (for the Chateau Petrus 2000).  Within that wide range, the average wine is surprisingly expensive.  The median price for a 750 ml bottle is $89.99 and the mean is just under $176.  That’s a lot for a bottle of wine!

Ratings on come from 14 different raters, although some are more prolific than others.  Not surprisingly, the largest number of ratings come from America’s most influential (and perhaps, most controversial) critic, Robert Parker.

The box plot below shows the top 10 raters (by number of ratings) and some statistics on how their ratings are distributed.

The first thing that jumps out from this is that there are NO wines on with a rating below 87!  In fact, when I looked more closely at the data I discovered there were very few below 90.  I suspect just doesn’t think those wines are saleable so they don’t list them.  I’m sure the various raters have given a much wider range of ratings than you see on this curated list from

So essentially, we have information about wine quality that’s on a 10-point scale from 90 to 100, and we have 10 raters who’ve contributed a meaningful enough number of ratings to do some statistical analysis.  Let’s have a look at what we can find out.

Analyzing Data from Professional Raters

I was interested in answering two questions about the raters.  The first was:  how correlated are their ratings to price?  After all, if somebody is just rating more expensive wines higher, that’s not going to be any help to me in identifying value.

The second thing I wanted to look at was: how correlated are they to each other?  Interpreting this one is a little more nuanced.  Perhaps if somebody has a low correlation to others it means they are an original thinker and may uncover some bargains for us. On the other hand … wouldn’t we hope that there really are some wines that are just, objectively, well made and should be recognized as such by seasoned professionals?  So perhaps low correlation might indicate a low level of expertise, or very unconventional tastes. I’m not sure about this one, so let me show you the data and you can decide.

Early on in the analysis I decided to hone the dataset down to only wines I’d actually be interested in buying.  I threw out everything with a review below 90 and everything with a price over $500 for a 750 ml bottle.

First, let’s look at correlation to price.

This is encouraging! Many of the raters are not particularly correlated to price, so we have a chance of finding some bargains.  It’s notable however how big the range is.  Firstly,’s own in-house reviewer (Wilfred Wong) showed the highest correlation to price.  Is this suspicious?  Is just trying to hype its most expensive wines?  Next highest was Wine & Spirits Magazine.  We’ll be coming back to them later.  On the other hand, Wine Spectator, Jeb Dunnuck, and Vinous’ Antonio Galloni all had correlations in the mid 40s.

Next, let’s look at the correlations between the raters.

The first thing you might notice about this grid is there are some empty cells.  In order to calculate a correlation between two raters, I started out by finding all the wines that they had both rated.  For some pairs of raters there just weren’t enough datapoints to estimate a meaningful correlation, so rather than show you spurious numbers I’ve left these blank.

The grid shows you the pairwise correlations, and the rows below give some summary statistics.  The first is the average correlation for that rater, weighted by the number of ratings.  So for example, if Robert Parker had a high correlation with Jeb Dunnuck, but he had fewer common ratings with him than the other raters, that correlation would be down-weighted.

I then produced an ‘Influence Score’ that multiplied the weighted average correlation by the number of reviews.  It’s very arguable whether this metric is truly tracking ‘influence’ because somebody who just follows the herd will score high on it – if they produce enough reviews.  That said, it’s a simple stat to understand so I thought I’d show it.  Robert Parker clearly dominates in terms of number of reviews.

There is an outlier though, and it’s Wine & Spirits magazine – to a quite striking degree.  I don’t have a good sense for why that is, so I hope to receive comments on this post from people with views.  However, based on this and the price correlations, I decided to drop both Wine & Spirits and from the analysis from here on.  I kept Burghound despite his low number of reviews because there’s a good reason for this: he only reviews Burgundy.  However, he’s a well-regarded expert in that region so I wanted to keep his ratings in the dataset.

Using Ratings to Find Good Value Wine

Armed with a sense of whose ratings seemed useful, I set out to build a tool to find the best value wines on  The first step of this was to look at some very simple questions.

  • For a fixed budget (say $30 a bottle) what are the highest-rated wines I can find?
  • For a fixed rating (say 95) what are the cheapest wines I can find? And,
  • For a given region (say, France/Burgundy) what are the best value wines available?  This one was a bit more tricky.

I built a wine chooser app to do this.  The static picture below shows some sample output, and the link here will take you to the dynamic version of it.  (It can take a couple of minutes to load – sorry!)

There are a couple of clear lessons to be learned from looking at the regional data.  Again, there’s a static picture below and a more interactive chart that you can hover over to explore here.

The orange circles are all France and the largest at the top is Bordeaux. (The size of the circles is proportional to the number of ratings on This took me by surprise:  Bordeaux, on average, is pretty good value.  Much more so than Burgundy (the next largest orange circle, distinctly below the regression line), or Champagne (to its left).  The rightmost French wine region is Rhone, which showed close to average value.

The dark green circles are America, with the largest being California.  That’s pricey too!  The better value US wines are the smaller two circles on the left: in the middle is Washington, and to the left is Oregon.  The light green circles are Italy,  where Tuscany shows up as better value (above the line) than Piedmont below it.

Overall it’s worth noticing that the smaller circles are nearly all above the regression line.  That means you’ll be able to find good value if you venture off the beaten track and try regions like Chile, Argentina, New Zealand and South Africa.

Defining a ‘Value for Money’ Score

Can we encapsulate ‘value’ into a single number and use this to search for wines? Clearly the answer is ‘yes we can’ – but the important part of the question is what definition of this number will give us the most useful predictions.  I experimented with a few different definitions before hitting on one that seemed to give me the most useful results.

‘Value’ is certainly a subjective measure.  One person may truly get a thrill from tasting 100-point wines while another might just view them as unnecessarily expensive.  It’s not obvious that the difference between a 90 and 93 rated wine should be worth the same as the difference between a 97 and 100 rated wine.  Perhaps a value score shouldn’t be a straight line but instead something that grows more rapidly at the top end.

Ultimately I decided my own preferences for wine to drink are close enough to a linear scale to stick with a simple model.  For each wine rating in the sample, I generated an expected price by using its rating and a linear regression on all the wines in the sample. (That’s the purple line you see in the chart above). 

Then I compared the actual price of the bottle to this expected price to see what discount (or premium) it had. I could have taken this price difference in dollars as my value function, but I realized this would be more likely to pick the more expensive wines. Thus, the score function I ended up using was the percentage discount the wine offered to its expected price.

Here's a what this function picked out of the global sample:

And here’s what it picked for Bordeaux:

Putting It Into Practice

The nice thing about is the vast majority of what’s on their site is available for immediate sale.  I decided to put my rating tool to the test and purchase a case of its top recommendations globally (with a bit of filtering for my personal taste.)  I’ll let you know what the verdict is over the next few weeks.  Happy drinking!

The skills the authors demonstrated here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.

About Author

Simon Yates

After running global equity derivatives for Credit Suisse and Citibank, Simon Yates switched to the quant hedge fund world, joining Two Sigma in 2014. He is a fan of all things quantitative, and an avid runner and skier.
View all posts by Simon Yates >

Related Articles

Leave a Comment

No comments found.

View Posts by Categories

Our Recent Popular Posts

View Posts by Tags

#python #trainwithnycdsa 2019 2020 Revenue 3-points agriculture air quality airbnb airline alcohol Alex Baransky algorithm alumni Alumni Interview Alumni Reviews Alumni Spotlight alumni story Alumnus ames dataset ames housing dataset apartment rent API Application artist aws bank loans beautiful soup Best Bootcamp Best Data Science 2019 Best Data Science Bootcamp Best Data Science Bootcamp 2020 Best Ranked Big Data Book Launch Book-Signing bootcamp Bootcamp Alumni Bootcamp Prep boston safety Bundles cake recipe California Cancer Research capstone car price Career Career Day citibike classic cars classpass clustering Coding Course Demo Course Report covid 19 credit credit card crime frequency crops D3.js data data analysis Data Analyst data analytics data for tripadvisor reviews data science Data Science Academy Data Science Bootcamp Data science jobs Data Science Reviews Data Scientist Data Scientist Jobs data visualization database Deep Learning Demo Day Discount disney dplyr drug data e-commerce economy employee employee burnout employer networking environment feature engineering Finance Financial Data Science fitness studio Flask flight delay gbm Get Hired ggplot2 googleVis H20 Hadoop hallmark holiday movie happiness healthcare frauds higgs boson Hiring hiring partner events Hiring Partners hotels housing housing data housing predictions housing price hy-vee Income Industry Experts Injuries Instructor Blog Instructor Interview insurance italki Job Job Placement Jobs Jon Krohn JP Morgan Chase Kaggle Kickstarter las vegas airport lasso regression Lead Data Scienctist Lead Data Scientist leaflet league linear regression Logistic Regression machine learning Maps market matplotlib Medical Research Meet the team meetup methal health miami beach movie music Napoli NBA netflix Networking neural network Neural networks New Courses NHL nlp NYC NYC Data Science nyc data science academy NYC Open Data nyc property NYCDSA NYCDSA Alumni Online Online Bootcamp Online Training Open Data painter pandas Part-time performance phoenix pollutants Portfolio Development precision measurement prediction Prework Programming public safety PwC python Python Data Analysis python machine learning python scrapy python web scraping python webscraping Python Workshop R R Data Analysis R language R Programming R Shiny r studio R Visualization R Workshop R-bloggers random forest Ranking recommendation recommendation system regression Remote remote data science bootcamp Scrapy scrapy visualization seaborn seafood type Selenium sentiment analysis sentiment classification Shiny Shiny Dashboard Spark Special Special Summer Sports statistics streaming Student Interview Student Showcase SVM Switchup Tableau teachers team team performance TensorFlow Testimonial tf-idf Top Data Science Bootcamp Top manufacturing companies Transfers tweets twitter videos visualization wallstreet wallstreetbets web scraping Weekend Course What to expect whiskey whiskeyadvocate wildfire word cloud word2vec XGBoost yelp youtube trending ZORI