Part of the appeal of good wine for me is the social aspect of discovery, and I don't think I'd want to trade that in for a technical one. I don't want a perfect wine so much as I want to drink a sincere recommendation from a friend or family member, and hopefully have that same wine again with them, and have something to talk about. In fact when I'm with certain people, even bad wine is fondly remembered. "What was that chalk we were drinking last month... I crave it when I think of you."*
The funny thing about all this is that his big data quest for more good wine was launched because of the same reason I enjoy wine: It was his brother-in-law's recommendation that found him his favorite, not crunching the #cassis #vanilla #persistentfinish numbers. And he's doing this to find bottles similar to the one from a beloved family member, most of which are far more expensive (eyeballing the graph), but if I were him, I don't know if I could ever find a better bottle than "my brother-in-law's wine." Isn't sentimentality worth more than the data, especially when working with the most fickle data-set of all?
Also, come on, don't make a post like this and not give us the list!
* (We were drinking Troublemaker, a Paso Robles blend, which I swear is 50% syrah and 50% limestone.)
You have to start somewhere, but I'd love to be able to get wine recommendations based on my personal favourites + some heuristics that come up with something less obvious or that I wouldn't pick otherwise.
I don't care so much about his list, but being able to use the same engine he did with my own starting point would be great.
On a side note, I wonder how the recommendations are influenced by the audience. i.e. wine.com is presumably used by US customers, so perhaps it could explain why French wines are described in certain ways compared to California. But you might get the opposite if you look at a European website.
That's the best definition of better in this case. I have a friend who is a wine maker(enologist) and he always says the same, the best wine is the one you enjoy not the one some magazine/person/web says. You can but the most expensive wine in the world and do not like it then it doesn't matter if it's good or not.
The article display some cool use for NLP, but as a winemaker and wine judge I find methodology funny and flawed.
You have to understand that wine descriptions are nothing more then marketing message that have very little resemblance to what is inside of the bottle. It would be interesting to see how those description change during the time as this message is changing all the time to catch the current trend.
Wine market is actually quite simple: you have few tiers of wine/wineries: from your basic mass market wines to mid tier to premium and ultra premium. Each category have some price point and higher the price better fruit (should) come to it.
There are differences in between the regions, but lower level wines are usually same no matter the marketing message and premium wines are usually very different and special in each own way anyway.
Agree that using marketing descriptions provided by the winery isn't the best approach. It would be much more useful if the data used stuck to just the reviews (though _which_ reviewers are used is its own issue):
> A Quid network shows 3,380 nodes, where each node represents a description and review of a single bottle of Bordeaux-style red wine from Wine.com.
FWIW, this is a topic we've explored in a few ways at VinePair:
Using user review are not very helpful as well. Each taster use different vocabulary and there are even different way to analyse wine: some people describe the flavours (this wine is strawberries, apple, lime) some, including me, are more of a structural tester (comparing "length" of a wine, width and character of the taste and after taste).
Second problem is (and there is even statistical proof) that after wine pass some threshold of kinda good wine, the resulting score is random. This is caused by few things: your mood, what you drink before it, anticipation, the meal and even music.
Scoring (and recommending) wine and selecting what you like is very complicated subject and probably best way is still ask some good wine buyer, describe what you like and he might recommend wines he likes that loosely fits your description.
Agree user reviews do prevent their own issues. One of those linked articles is about the confusion over the meaning of 'dry' in the contexts of sweetness and tannins.
> ...probably best way is still ask some good wine buyer, describe what you like and he might recommend wines he likes that loosely fits your description.
Best advice I've gotten/give is find an importer/distributor (eg Kermit Lynch) and rely on their selections.
Moreover, the "kinda good" level is not at a very high price point. I've never had an undrinkable bottle that cost over $15 (unless it was "corked"). A majority of wines I've tried over $10/bottle are very drinkable. I think it has to do with the amount of money it takes to literally put the wine in the bottle.
I'm very curious. When you say "undrinkable", what do you mean? I've heard a lot of wine fans say this and I've never really gotten a straight answer. Do you instantly pour out your glass and throw out the bottle? Do you finish your glass but throw away the bottle? Or maybe finish your glass but save the rest of the bottle for cooking?
I'm also curious about what wines you're getting over $10/bottle that are not drinkable. Here in the Netherlands we have fairly cheap wine, but not outrageously cheap by any measure, and the majority of what I drink is 6-10 euros/bottle. I don't think I've ever gotten an "undrinkable" bottle over 6 euros. The only "undrinkable" bottle I can remember buying in recent memory was part of a sale - some "fancier" wine at 5 euros a bottle if I remember - and though I don't think it was corked the taste was off enough that I decided to toss the bottle.
For me undrinkable = faulty wine, but those are very rare these days. I just throw those wines away.
Bigger problem with wines in region of 5 - 15 euros is that they are boring. They taste like generic wines and nothing more. I put aside those wines and use them for cooking.
> Wine market is actually quite simple: you have few tiers of wine/wineries: from your basic mass market wines to mid tier to premium and ultra premium. Each category have some price point and higher the price better fruit (should) come to it.
Also important to note that more money == better wine is not linear. For me (and most people) it does not take much money to get to a 'good' wine. Having drank wines in all price ranges ($5 big bottles to $1000 Chateau Margaux/Laffites - never had a DRC though not sure that's even possible now with all the fakes out there, but I digress), I have hard time spending more than $20 on a bottle nowadays and would prefer $10-$15.
For me the wine is all about culture. I do like to read the profiles on wineries: history, how they work in vineyards, how they treat the wine in cellar, why they choose of what they do. What is the history of the region and how it was shaped during centuries (yeah, I am not really into new world wines).
Then I have at least some background on what I am drinking and if I find some winery I might like, I try to taste multiple wines and if I like them I just keep buying it. During the years I found some good selection for any kind of occasion. (Yes and most of the wines I buy are usually 10 - 15 USD + few more expensive for some special occasions).
Ah yes, wine and history go together so well. I've found that many (most?) people who like one like both. I took a wine tour once in Hvar, Croatia and it was really a history tour of the island with wine mixed in.
I used to know a couple people where wine was a status symbol, and that's how I got to taste some insane things. Personally, I look at wine as you do and want to know about the history, culture, tradition, love (for lack of a better word) that went into babying a finicky vine like the grape and turning it into something that can be so complex and delicious.
I had a heuristic that the flashier and more colorful the label the worse the wine was. However, bad wine makers have seemingly adapted to this and now have understated labels.
I actually like this analysis. I think there's something there. But man-oh-man, this is about as "small data" as it gets. Has "big data" become just a short name for an assorted bag of statistical tools?
>>>Has "big data" become just a short name for an assorted bag of statistical tools?
Yes. Or as @mhoye on twitter put it: "FACT: most "big data" is actually two small datas standing on each other shoulders wearing a trenchcoat and sunglasses."
In 2013, I got taste and aroma for around 1100 wines. I used the data to create flavor profiles and predict the price based on the chemical composition. In general, more expensive wines are more complex and the predicted price matches the actual price quite well. This breaks down on the extremes. Boxed wines such as Franzia and Vella are under priced. Not by much, but their juice is closer to 4-5 dollar bottled wines. At the time of the calculation a bottle of Franzia was around 2 dollars. So, it's a pretty good deal. On the other side, most new world wines over 35 dollars are over priced. There's nothing chemically different to the the wines just below 35 dollars. There were a few wines in the middle that were way off. I remember a Clos Du Bois was under priced by 7 dollars, but most of them were in the right range. The biggest offenders were wines that built their reputation in a well known appellation and moved a level higher. It's very common for wines to start in Sonoma Coast or Carneros and end up with a California appellation.
Did you try the underpriced ones, and were they good? A particular wine can be complex but have some specific quality that makes it unappealing. I'd be surprised if complexity is the main variable between different vintages of the same wine that have very different prices.
I didn't try any. I drink very little wine. I did show the list around and no one was surprised. It wasn't Kendall-Jackson that was under priced. It was smaller wineries that had been getting solid reviews. There must have been some surprises, but I knew very little of wine back then. I also knew very little about machine learning, R or anything else. I knew enough to use Random Forests, R for graphing and the opinion of others to judge the results.
I tried finding the file with the list of all the wines and the predicted prices. I didn't even know enough to name the file correctly.
I think the large companies are very consistent with their product. They have many sources for their juice and because of that their product will be more average. The small companies can have years where they do everything right and create a great product, but it's not like they can just change their price to reflect the better quality.
27 comments
[ 3.0 ms ] story [ 65.3 ms ] threadThe funny thing about all this is that his big data quest for more good wine was launched because of the same reason I enjoy wine: It was his brother-in-law's recommendation that found him his favorite, not crunching the #cassis #vanilla #persistentfinish numbers. And he's doing this to find bottles similar to the one from a beloved family member, most of which are far more expensive (eyeballing the graph), but if I were him, I don't know if I could ever find a better bottle than "my brother-in-law's wine." Isn't sentimentality worth more than the data, especially when working with the most fickle data-set of all?
Also, come on, don't make a post like this and not give us the list!
* (We were drinking Troublemaker, a Paso Robles blend, which I swear is 50% syrah and 50% limestone.)
I don't care so much about his list, but being able to use the same engine he did with my own starting point would be great.
On a side note, I wonder how the recommendations are influenced by the audience. i.e. wine.com is presumably used by US customers, so perhaps it could explain why French wines are described in certain ways compared to California. But you might get the opposite if you look at a European website.
https://www.theguardian.com/lifeandstyle/2013/jun/23/wine-ta...
http://www.realclearscience.com/blog/2014/08/the_most_infamo...
http://www.economist.com/blogs/graphicdetail/2017/05/daily-c...
It defies common sense to think all wine is equal quality. The fact that there's much nonsense in the wine world doesn't mean it's all nonsense.
We'd need to see much stronger evidence than what has been presented.
You have to understand that wine descriptions are nothing more then marketing message that have very little resemblance to what is inside of the bottle. It would be interesting to see how those description change during the time as this message is changing all the time to catch the current trend.
Wine market is actually quite simple: you have few tiers of wine/wineries: from your basic mass market wines to mid tier to premium and ultra premium. Each category have some price point and higher the price better fruit (should) come to it.
There are differences in between the regions, but lower level wines are usually same no matter the marketing message and premium wines are usually very different and special in each own way anyway.
> A Quid network shows 3,380 nodes, where each node represents a description and review of a single bottle of Bordeaux-style red wine from Wine.com.
FWIW, this is a topic we've explored in a few ways at VinePair:
* https://vinepair.com/wine-blog/how-much-do-the-labels-on-win...
* https://vinepair.com/articles/wine-industry-touch-consumers/
* And a more humorous look at these descriptions @ https://vinepair.com/robert-robot/
Second problem is (and there is even statistical proof) that after wine pass some threshold of kinda good wine, the resulting score is random. This is caused by few things: your mood, what you drink before it, anticipation, the meal and even music.
Scoring (and recommending) wine and selecting what you like is very complicated subject and probably best way is still ask some good wine buyer, describe what you like and he might recommend wines he likes that loosely fits your description.
> ...probably best way is still ask some good wine buyer, describe what you like and he might recommend wines he likes that loosely fits your description.
Best advice I've gotten/give is find an importer/distributor (eg Kermit Lynch) and rely on their selections.
I'm also curious about what wines you're getting over $10/bottle that are not drinkable. Here in the Netherlands we have fairly cheap wine, but not outrageously cheap by any measure, and the majority of what I drink is 6-10 euros/bottle. I don't think I've ever gotten an "undrinkable" bottle over 6 euros. The only "undrinkable" bottle I can remember buying in recent memory was part of a sale - some "fancier" wine at 5 euros a bottle if I remember - and though I don't think it was corked the taste was off enough that I decided to toss the bottle.
Bigger problem with wines in region of 5 - 15 euros is that they are boring. They taste like generic wines and nothing more. I put aside those wines and use them for cooking.
Also important to note that more money == better wine is not linear. For me (and most people) it does not take much money to get to a 'good' wine. Having drank wines in all price ranges ($5 big bottles to $1000 Chateau Margaux/Laffites - never had a DRC though not sure that's even possible now with all the fakes out there, but I digress), I have hard time spending more than $20 on a bottle nowadays and would prefer $10-$15.
Then I have at least some background on what I am drinking and if I find some winery I might like, I try to taste multiple wines and if I like them I just keep buying it. During the years I found some good selection for any kind of occasion. (Yes and most of the wines I buy are usually 10 - 15 USD + few more expensive for some special occasions).
https://en.wikipedia.org/wiki/History_of_California_wine
https://en.wikipedia.org/wiki/Zinfandel
I used to know a couple people where wine was a status symbol, and that's how I got to taste some insane things. Personally, I look at wine as you do and want to know about the history, culture, tradition, love (for lack of a better word) that went into babying a finicky vine like the grape and turning it into something that can be so complex and delicious.
Yes. Or as @mhoye on twitter put it: "FACT: most "big data" is actually two small datas standing on each other shoulders wearing a trenchcoat and sunglasses."
I tried finding the file with the list of all the wines and the predicted prices. I didn't even know enough to name the file correctly.
I think the large companies are very consistent with their product. They have many sources for their juice and because of that their product will be more average. The small companies can have years where they do everything right and create a great product, but it's not like they can just change their price to reflect the better quality.