Monday, 16 April 2018

Why comparing wine-quality scores might make no sense

There is no mathematical meaning to comparing wine-quality scores between different critics.

If you do want to compare scores, then it can only validly be done between those scores produced by any one critic (eg. the same critic tasting different wines, or even the same wine on different occasions). There is no mathematical justification for comparing scores between critics (eg. different critics tasting the same wine, even at the same time). That is, quality scores provide a ranking only for the wines tasted by any given critic, nothing more.


Background

Wine-quality scores are an important concept in the wine industry, for several reasons. First, wine critics produce them, and there would be precious little wine writing without them. Second, wine drinkers and buyers use them to help make their wine-purchasing and wine-drinking decisions. Third, marketers use them as an advertizing tool, usually along with a lot of flowery words about the wines.

So, these scores are not going away any time soon, no matter how many pundits proclaim their demise. Instead, what we need to do is come to terms with their characteristics, so that we can use them effectively.

To this end, there is actually a small body of professional literature about the vagaries of the range of wine-quality scores that are currently in use; and I have produced several blog posts myself, trying to make sense of what is going on.

Reasoning

I have finally concluded that there are two fundamentally different sorts of wine-quality scores in use: (1) what we might call an objective score, based on explicitly assigning points to a series of pre-defined wine characteristics, and then summing them to get the wine score; and (2) subjective (but expert) scores, where the overall score comes from whatever characteristics the scorer wants to express. There are many variants of these two score types, especially the subjective scores, but for our purposes in this post this variation is not relevant.

What is important, instead, is that these two types of scores should not be confused, although most people still seem to do this — people often refer to "wine scores" as though they are all the same. However, the two types have fundamentally different mathematical behaviors. Their mathematical behavior is of the utmost importance because this is what numbers are all about — if numbers have any meaning then it must be a mathematical meaning, otherwise words would be enough.


So, we need to distinguish between the scoring scheme, which contains the information about wine quality, and the scale, which is the way that the quality is expressed (stars, points, words, etc). Formally, for the objective scores there is a single scoring scheme and a single scale being used to express that scheme (eg. the numbers 1-20 used by the UCDavis quality score). However, for the subjective scores there are many different scoring schemes, even though a single scale is being used to express those schemes (eg. the numbers 50-100 used by the majority of wine critics, as well as by community sites such as Cellar Tracker or Vivino).

This distinction can be illustrated using this 2x2 table:

Objective scores:
Subjective scores: 
Scale
 one (eg. 20 points)
 one (eg. 20 points)
Scoring scheme
 one (pre-defined)
 many (chosen by critic)

This means that: for the objective points scores, since there is only one scoring scheme, then differences in points always reflect differences in the perceived qualities of the wines; but for the subjective points scores, there is a wide choice of scoring schemes — the scoring scheme can mean anything the person wants it to mean. In both cases, there can be personal choices about the wine quality, but in the subjective case there are also choices about how to interpret the scoring scheme (ie. what it actually means).

Of most importance, then, is that the objective scores can be directly compare between critics, because any difference in score will almost always represent a difference of opinion about the quality of the wine. On the other hand, the subjective scores cannot be compared, because any difference or similarity of the scores could be interpreted as either (i) a difference of opinion about the wine quality or (ii) the use of different scoring schemes — for example, there can be different schemes for reds versus whites, or sweet versus dry wines, or even different grape varieties.

For subjective wine-quality scores, we thus cannot tell what numerical similarity or difference of scores actually means. The same scores could mean different qualities (because the scoring schemes are different), and different scores could mean the same quality (because the scoring schemes are different). How on earth are we to know? We can't!

I have listed some of my previous posts at the bottom of this page, which provide illustrative examples of just how many different scoring schemes there are among critics, even when they are ostensibly using the same scale.


Finally, note that that it is the combination of a value judgment with a variable scoring system that is the issue. Variable scoring systems on their own are not problematic, provided they are measuring an objective phenomenon. For example if we are measuring the length of something, then it does not matter whether we use yards, meters or cubits, because the length itself will be the same in all three cases, and we are just describing this using different units. But wine quality is not an objective phenomenon, in this same sense — it is to a large extent a value judgment; and this creates the problem. Different scores may mean different judgments or they may mean different scoring schemes.

Conclusions

Harvey Steiman (Editor at Large, Wine Spectator) once wrote (Are ratings pointless? June 15, 2007):
The main reason I like to use the 100-point scale is that it lets me communicate more to my readers. They can tell that I liked a 90-point wine just a little better than an 89-point wine, but a 94-point wine a lot more than one rated 86.
And that ranking is all the score does — we cannot compare Mr Steiman's numbers to anyone else's numbers. This is a pity.

However, the average quality score produced by community sites like Cellar Tracker might possibly have some meaning, but only if it is an average of enough scores. I have no idea what "enough" would be in this case, but it has to be a large enough set of scores to "average out" the fact that the many people producing the scores may all mean different things. If variation among the scores varies randomly about some average value (as is likely), then calculating an average score will, indeed, address the issue. But an average derived from a small number of scores is itself subject to random variation, although this decreases as the sample size increases. Trying to work out the required sample size might be the topic of another blog post.

Moreover, as I have emphasized, if a consistent scoring scheme is used (ie. an objective score), then the scores can naturally be compared among tasters. I have done this comparison, for example, when I have used data from the Vintners Club, which employs the standard UC Davis 20-point scoring system for its tastings (see the list of posts below). Here, averaging the scores does, indeed, make perfect mathematical sense, because all of the scores are based on the same scoring scheme — differences in scores can only mean differences of opinion about wine quality.

Finally, there are a number of contributions to the professional literature that cover the implications of this topic for wine competitions; but I will cover that in another post.

Previous blog posts illustrating the differences between scoring schemes
Previous blog posts using an objective scoring scheme

Monday, 9 April 2018

The rise, rise, fall and rise of Australian wine

Global Wine Markets, 1860 to 2016: a Statistical Compendium (by Kym Anderson, Signe Nelgen and Vicente Pinilla. 2017. University of Adelaide Press) lists a number of countries that are net exporters of wine, in the sense that wine production per capita exceeds wine consumption per capita, including (in decreasing order) Spain, Chile, Italy, New Zealand and Australia.

A few weeks ago I discussed the inexorable rise of New Zealand wine over the past couple of decades (The rise and rise of New Zealand wine). I presented some graphs comparing New Zealand with Australia, and noted the somewhat longer and more rocky road the latter wine has traveled. Here, I will look at that road in a bit more detail.


Technically, the story of Australian wine starts on January 1st 1901, which is the Australian equivalent of July 4 1776 for Americans (except that the British decided to avoid having the Australians shooting at them, to make them go away). Until that time, it was Empire wine, not Australian wine.

Nevertheless, there was a pre-history for Australian wine. For example, Kim Brebach has noted:
Australian wine had a heyday in the latter part of the 19th century, when the Gold Rush brought all kinds of adventurers to the country. Boat people, migrants, refugees. They came from the old world and the new (America). They were seekers of fortune, followed by entertainers, suppliers of mining needs, cooks and market gardeners.
The depression of the nineties saw a return to a simpler life for most Australians, where survival was the order of the day and food and wine took a back seat. Little changed through the first decade of the new millennium, and the second, which saw the Great War, as they called it. By the end of the third decade, things began to look up, but then the sky fell in on the stock market, the Great Depression followed, and yet another Great War.
For the first half of the 20th century, winemakers survived by turning much of their fruit into Brandy, Sherry and Port ... Table wines were a rarity because drinking wine with food had become a vague memory from a bygone era to which most people had lost the link ... Of course there were people who drank table wine during the war years and the depression — doctors, lawyers, academics and other professionals — but their number was small.
So, for our purposes, the real story of Australian wine actually starts mid-century, with what has been called the post-war Baby Boomer generation. This, incidentally, is my generation.

Australia versus Europe and the USA

The first thing to get clear is that there is actually no such thing as "Australian wine". Australia is a continent, the size of the continental USA and larger than continental western+central Europe, as shown in the above figure. So, there are as many radically different wine-growing regions within Australia as there are anywhere else on this planet; and the diversity of wines and styles reflects this. Don't let the fact that it is a single country fool you into thinking that there is a single wine style.

Nevertheless, Wine Australia, the nationally funded statutory service body for the Australian grape and wine community, likes to talk about "Brand Australia", and who am I to argue? So, I will look solely at the national level in this blog post. Most of the data for the following graphs come from the book by Kym Anderson (with the assistance of Nanda R. Aryal) Growth and Cycles in Australia’s Wine Industry: a Statistical Compendium, 1843 to 2013 (University of Adelaide Press, 2015) .

Wine exports

Let's start by looking at the graph that inspired the title of this blog post. Each point represents one year, from 1901 to 2017, showing the value of the wine exports in A$. Note that the value scale is logarithmic, so that a straight line on the graph represents exponential growth.

Australian wine export value 1901-2017

The graph shows that wine export value increased at a roughly constant exponential rate until 1980, followed by a increase in the exponential rate until the early 2000s, then a sharp decrease, followed by a recent recovery. So, to the rest of the world, Australian wine has shown a rise, rise, fall and rise.

Let's look at what might lie behind these patterns. The most obvious place to look is wine production, of course, as shown in this next graph. Note that production has increased at a constant exponential rate, all the way from 1901 to the present. So, this explains the initial rise in wine exports from 1901 to 1980 — the increase in exports was simply tracking the increase in wine production.

Australian wine production 1901-2013

This is what economists like to see, but it is not really sustainable, because suitable land has a finite area, and so there is a limit to production.

Moving on, what happened after 1980? Why was there a sudden increase in exports? To look at this, we need to consider Australia's population size, and the behavior of the people. This next graph shows the population size through time. Note that the graph scale is not logarithmic.

Australian population 1901-2015

The graph shows three distinct periods of different rates of population growth. Australia's population growth is, and has been since 1788, dominated by migration — currently, more than 50% of the populace are either first or second generation migrants. (There is no other country on the planet that is like that.) So, what the graph shows is three periods of migration rate, with a dramatic increase in migration immediately after World War II, and another increase at the turn of the current century.

The importance of these patterns for Australian wine exports is that the population growth has been linear, whereas wine production has grown exponentially. Wine production has out-stripped the population.

This has been counter-acted to some extent by changes in consumption of wine, as shown in the next graph. Wine consumption (per adult, of course) rose dramatically during the 1960s, peaking during the 1980s. It was the recent Italian, Greek and other European migrants who re-introduced the locals to the delights of good food and wine after World War II. Like California at the same time, Australia was still producing bulk wines with faux-French names well into the 1960s; but this changed during the 1970s, to a celebration of Australian wine styles for their own sake.

Australian wine consumption 1922-2013

So, up until the 1980s, increased consumption (not population growth) matched the exponential increase in wine production. After that, the domestic system fell apart — wine consumption reached a plateau but wine production kept increasing. This is why there was a sudden increase in exports during the 1980s, as shown in the next graph.

Australian wine exports 1901-2013

The Australian wine industry started a concerted effort to attract the world's wine consumers to Australian wine; and it was at this stage that the world wine media first started paying serious attention to Australian wine. As an aside, the odd peak of exports in 2007 (as seen in the graph) was due to a dramatic decrease in wine production that year.

However, by 2010, exports had also reached a plateau, as a percentage of production. Indeed, by 2005 export volume had exceeded 50% of production, and this was actually the peak in the value of wine exports (as shown in the original first graph). It has remained at 50-60% since then, but the dollar value of exports has decreased.

It was at this stage that the Australian wine industry was losing its way. With the exception of the boom in Yellow Tail wines in the USA (see Yellow Tail and Casella Wines), the first 15 years of the new millennium were poor ones for Australian wine exports.

Basically, the exports focused on the cheap end of the wine market; and this is no way to make money. The Australian wine industry became known as a source of inexpensive wine, mostly exported in bulk. I have discussed this issue in previous posts (Global wine exports; United States wine imports and exports), noting that, for example, Australia has recently made less money per liter out of its wine exports to the USA than anyone else, in spite of being the second-largest supplier, as a result of the industry's approach to doing business that has confused volume for profitability.

The current situation

Domestically, Australia has gone the same route as the United Kingdom, so that almost all wine is sold through the two biggest local supermarket chains. In the UK, the chains started buying cleanskin wines (usually from wine co-ops or factory-scale commercial wineries), and marketing them with own-brand labels — actually, they used to hire Australian winemakers to "clean them up" for the tastes of the UK market!

So, the Australian supermarkets currently own the retail (>70% market share), the distribution and the winemaking, as far as domestic consumption is concerned (see the list of wine brands at Who makes my wine?). Apart from the own-brand wines, the supermarket-owned liquor chains mostly favor the big wine companies (Treasury Wine Estates, Accolade, Pernod Ricard), which own about 70% of Australian wine production (their brands are also listed at Who makes my wine?). The other 30% of production is made by c. 2,500 small- to medium-size wineries.


That is why there has been so much recent focus by Wine Australia on getting the export market back on track. Indeed, it has been reported that in 2017 Australian wine exports hit a record high in both volume and value. Furthermore, the Australian Government recently enacted the Wine Australia Regulations 2018 (replacing the Australian Grape and Wine Authority Regulations 1981) to help regulate and protect wine exports as far as product, shipment and licensing are concerned (ie. protect Australian wine brands and the reputation of Australian wine exports).

By region, Europe currently receives c. 38 million cases of Australian wine, North America 26 million cases, and Asia 22 million. Within Europe, the United Kingdom is the biggest market, followed by Germany and then the Netherlands. More importantly, Australia's Jacob's Creek is the top wine brand in New Zealand!

Australia has recently signed a number of free trade agreements, notably the Trans-Pacific Partnership (Australia plus: Brunei, Canada, Chile, Japan, Malaysia, Mexico, New Zealand, Peru, Singapore, Vietnam. The current US president withdrew his nation from their original intention to be part of this deal, which may disadvantage Australia, given the size of the US market; but on the other hand, he is also embroiled in a tariff dispute with China, which will probably benefit Australia. The expansion of free-trade agreements is notwithstanding the current problems between Canada and Australia.

So, the recent focus has been on China. The current free trade agreement between Australia and China took effect in December 2015. France is still the dominant wine seller to China, holding about 40% of the imported wine sales market. However, Australia has been in second place for a decade; and, while French sales growth has been steady, Australian exports have skyrocketed. China is now Australia's largest wine export destination (A$848m), with the USA now second (A$449m) and the UK falling to third (A$348m). Indeed, it has been reported that, due to a smaller 2018 harvest plus the growing Chinese market, demand is now outstripping supply for bulk exports of Australian red and white wines.


The recent focus has also been on premiumization. Treasury Wine Estates, in particular, have finally realized the dollar value of selling premium wines rather than getting involved in wine discounting — they are now selling less wine but doing so much more profitably (leading to a rapid rise in its share price). The key export brands are currently the premium brands: Penfolds and Wolf Blass. TWE's main focus is now firmly in China — it is not for nothing that the super-premium Penfolds G3 wine was launched at the exclusive Liang Yi Museum in Hong Kong, not in Australia.

The Australian wine industry at large is slowly learning this same lesson. For example, over the past decade, the value of Australia’s wine exports to China has expanded roughly twice as much as volume. China is the country where Australian exports have the highest average dollar per liter (AU$5.55), while the USA is only now seeing growth in Australian export wines that sell above $15 retail.

Conclusion

So, here is the first graph again, with a couple of lines added, showing that: (i) the initial rise in wine export value was simply a balance between population and wine consumption on the one hand and wine production on the other hand; (ii) the second rise in exports was a concerted effort to deal with saturation of the domestic wine market after 1980; (iii) the fall at the turn of the century was due to saturation of the export wine market and a focus on unremunerative bulk exports; and (iv) the final rise is due to a recent focus on premium exports, particularly to China.

Australian wine export value 1901-2017

Addendum

I have been asked about the affect of inflation on the dollar value of the exports (Australian inflation has been an average of 4% annually since 1901). So, I have used the Inflation Calculator available from the Reserve Bank of Australia, to convert all of the export values into 2017-equivalent AUD. This is shown as the reddish line in this next graph, for comparison with the unadjusted data.

Exports expressed in 2017 dollars

Monday, 2 April 2018

Artificial intelligence in the wine industry? Not yet, please!

The world is changing rapidly, and the wine industry needs to keep up. However, we should not trip over our own feet in a mad rush to do this. We need to think carefully about just which bits of the modern world will be beneficial, and in what capacities. To this end, I have already written about Big Data, and about the use of social media, and about the vagaries of community wine-quality scores, along with some cautionary tales.

At the risk of becoming a perpetual nay-sayer, I must now say something about Artificial Intelligence (AI). Computing is something I know about, and the potential problems with AI are just a bit too obvious for me to let them pass by unnoticed. Once again, I feel that the enthusiasts are being a bit too enthusiastic, and not quite critical enough for clear thinking. The wine industry deserves better than this.


AI is just what it says — artificial. Whether it is also intelligent I will let you decide for yourselves, below. Human intelligence is sometimes called into question, usually for a good reason, but we should always call artificial intelligence into question.

What is artificial intelligence?

Humans learn by example. Given suitable examples, we can learn to do some pretty impressive things. This is what we mean when we say that human beings are intelligent — we interact with the examples, using trial and error to work out how to do whatever it is that we are trying to do. Sadly, if we are presented with bad examples, we can also learn some pretty bad habits — that is the trade-off, which we have happily accepted.

On the other hand, when we have previously devised machines to aid us in our endeavors, we have designed them to function in very specific ways. The machine does not interact with the world to learn new functions, but instead we have to devise these new functions ourselves and then re-design the machine. Pens dispense ink but cannot learn to compose text; knives cut food but cannot learn to cook that food; and cars cannot learn to fly, even if we add wings to help them do so.

This situation is now changing with the advent of Artificial Intelligence. Computer programs based on AI are not told by humans what to do — they learn by example, not by instruction. That is, they are presented with a collection of examples, plus a programming system that allows them to devise their own behavior from whatever patterns they detect in those examples. This is an example of what is called Machine Learning. It is a probabilistic system — the AI system may not make the same decision each time it meets a new situation, but instead it will have a probability associated with each of several possible behaviors. This is unlike our previous machines, where each machine should repeatedly do the same thing under the same circumstances.


We have very little control over what it is that the AI systems learn — we can only control the actual examples, not what patterns the AI system finds in those examples. If a system learns bad habits, for example, all we can do is keep giving it more and more good examples, and hope that it eventually re-learns. Just like people, right? Indeed, just like any Complex System, the outcome can be unpredictable, as well as uncontrollable.

Let's first look at a few successful examples of AI usage; and then we will look at what sorts of things can go wrong.

Some examples of AI

Perhaps the best-known early application of Artificial Intelligence has been in the matter of designing computer programs to play competitive games, such as chess or poker. Here, the process is relatively straightforward, because the program input is a series of game situations plus their outcome under particular future plays, from which the AI program can deduce the probabilities of success when following any given strategy. The most recent, and most successful, chess example is the AlphaZero program. At the moment, the AI successes are restricted to 2-person games.

Other commonly used examples of AI include the digital "personal assistant" apps, such as Apple's Siri and Amazon's Alexa, along with the predictive film-choosing technology from Netflix and the music-choosing technology from Pandora. In a more modern but less-common vein, predictive self-driving features of Tesla cars are all based on AI. A bit of the history of AI and some other examples are included in The WIRED guide to artificial intelligence.

A not-so-good example (from the wine world)

A classical use of AI is in the Google Translate system, which allows us to translate online text between a wide range of languages. Here, I present a simple example taken from my own experience, in which some text from a Swedish wine site, describing three wines, is being allegedly translated into English.

Original text:
Translated text:

The titles alone tell you that something is wrong, because the translated title makes no sense — it should say "Less than SEK 70". Note that the word "kronor" has successfully been translated in the title — this is the Swedish currency, which would translate literally as "crowns", but SEK is the accepted financial abbreviation.

However, look at the way the other three occurrences of "kronor" have been translated! The text actually has four different translations of this one word, even though the format of the text is unvarying, and all four occurrences should be translated the same way — we have: "SEK", "billion", "$" and "crowns". The first and last are correct translations, but the other two are complete nonsense. Note, especially, the direct translation of Swedish currency to dollars without using an exchange rate — this is not unusual for Google Translate, which is also known to translate "meter" to "foot" without a conversion, for example.

The issue that I am highlighting here is that we cannot ask why the AI system has done this. There is nothing in the programming that tells the system to use any given translation. The system is simply given a large body of text (original text plus a translation), and the algorithm tries to find repeated patterns connecting them. From this deduced information, it makes its probabilistic decisions with each new piece of untranslated text. In this case, Google Translate has learned four different possible translations, and decides which one to use on each occasion.



The only way to correct this problem is to keep providing more and more text (original plus translation), until the system starts to get its decisions right (by finding the correct patterns). We cannot tell it what to do — it is "intelligent", and therefore must work it out for itself.

This solution will eventually work. For example, a couple of years ago the Google translation of Swedish text always ignored the Swedish word "inte". This was a problem because the word translates as "not", which creates the negative of the sentence (see Wikipedia). You can image how silly the translations were, when they said impossible things could happen! Fortunately, Google Translate has now corrected itself (through 2 years' worth of more examples), and "inte" is currently translated correctly.

Along the same lines, if you really would like to see some bizarre translations, try getting Google Translate to convert some Latin text into English (or any other non-Romance language of your choice).

The take-home message

The issue with Artificial Intelligence is this. The old-style approach to computing and machines involved specialization — each machine did one thing only, and did it well. The AI approach to computing and machines involves them being generalists — each of them can do a lot of things, but this risks that they do none of them well. So, in my example, traditional translation systems involve only one pair of languages at a time, and these are translated properly. Google Translate is a system that tries to do all pairwise languages, and at the moment it doesn't do any of them particularly well.

We need to make a choice — we can't have it both ways.


The wine world of AI

So, what are we getting ourselves into, if we bring AI into the wine world? What are people suggesting that we use it for?

Perhaps the most widely touted use of AI in the wine industry is the sort of predictive technology mentioned above for Netflix and Pandora — given certain basic pieces of information about the customer, a computerized assistant should be able to make sensible suggestions regarding wine purchases or food/wine pairing.

This idea is based on having a database of wine information, which is connected by expert knowledge to some sort of consumer "profile". In short, both the wines and the consumers are "profiled" is some way, and the two datasets are connected by an AI system.

This general sort of idea is being (or has been) pushed by a number of companies, producing mobile apps or online sites, such as Next Glass, WineFriend, Hello Vino, Wine Ring, and WineStein. These AI systems usually ask the user a set of questions, and then suggest new wines based on the answers, and possibly also on previous wine consumption.* Wine Ring, for example, has even made it into reports on CNBC and Go-Wine.

This AI approach has also been pushed by some of the social networks, which started out as ways to record what you drink and whether you like it, but have recently morphed into general-purpose wine sites. So, sites such as Vivino now use AI to provide new wine recommendations according to the wines already rated or bought by the consumer. Even Wine-Searcher, which mainly connects consumers with wine prices from an array of retail shops, is testing a recommendation chatbot, called Casey.


This idea may be the least problematic use of AI in the wine industry. It can work well, depending on the quality and quantity of the database containing the wine-related information, and how well it is connected to the customer information. Novices, in particular, can benefit greatly from this use of AI, if it is implemented effectively — but don't be surprised by unpredictable or unexpected wine suggestions, since the AI system itself is dealing with probabilities only. Moreover, speaking as a biologist, the oft-used biological metaphor of the AI database functioning like a "genome" is utter nonsense (see the most popular blog post I have ever written: The Music Genome Project is no such thing).

However, the computational scientists are keen to push these ideas much further. The Google internet search engine is a pretty straightforward implementation of a database search strategy (with a lot of bells and whistles). However, Wolfram Alpha touts itself as a "computational knowledge engine", based on AI — instead of finding a web resource that might contain the answer to a given question (as Google does), it tries to compute the answer from the knowledge in its own databases. It can certainly do some pretty fancy things (see 32 tricks you can do with Wolfram Alpha, the most useful site in the history of the internet). However, if we compare a query for "climate zones" (see last week's post) in each technology — Google returns links to a series of web pages about climate and climate zones (prominently including Köppen's climate classification), whereas Wolfram Alpha returns nothing more than some data about the climate in the town of Zone, Italy. Artificial Intelligence is alright in its place, but we need to understand what that place is, if we are to use it effectively. Horses for courses, as the saying goes.


At the other extreme from simple predictive technology, it has been pointed out that one likely consequence of AI technology is the automation of many tasks currently employing millions of people (Google Chief Economist Hal Varian argues automation is essential). The only real question is whether this will occur sooner or later, not whether it will occur. The point is that, in the past, only repetitive jobs could be automated by machines, but with AI a much winder range of jobs can now be learned by newly designed machines. Self-driving cars are an obvious example, following on from the long-standing use of autopilots in aeroplanes. The issue here is that flying a plane is actually easier to automate than is driving a car!

In the wine industry, as far as autonomous vehicles are concerned, we already have the WineRobot, which wanders the vineyards gathering information about the state of the vines (such as vegetative development, water status, production, and grape composition), just like vineyard managers used to do. We also have the Wall-Ye V.I.N. robot, which carries out the labor-intensive vineyard tasks of pruning and de-suckering; and TED, a robot that neatly weeds between the vineyard rows (for those people who don't use sheep to keep their weeds under control). Other ideas about what is now called Robotic Farming are covered in a short video from the Australian Centre for Field Robotics, at the University of Sydney— a farm is a much safer place for autonomous vehicles than is a public road. [Aside: I learned both my biology and my computing at this university.]


In between these two extremes, the most obvious use of AI systems is likely to involve computerized forecasts, such as early-season vintage forecasts in a vineyard, or sales and price forecasts in a shop. [Note: a forecast is different from a prediction, as I will discuss sometime in a future post.] In these cases, the forecasts are expected to improve through time, as more and more data are gathered, and the AI system continually adjusts itself based on newly found patterns in the data. These forecasts are, thus, adaptive.

It is here that I am most skeptical about the benefits of Artificial Intelligence. My example above of the issues with Google Translate seems to be all too pertinent here. Forecasts are problematic no matter how they are implemented; and AI will not necessarily help. The issues with forecasts lie much deeper than mere "intelligence", with the fact that the future is often so disconnected from the present and the past. The old finance "40% Rule" seems all too apt — one can look like a good forecaster simply by following any proposition with a 40% probability (see How do pundits never get it wrong? Call a 40% chance).

For a selection of other, rather enthusiastic, discussions of AI and wine, see:

* Have you ever noticed that the only two groups who refer to their customers as "users" are the computer industry and the illicit drug industry? I think that this is very revealing.

Monday, 26 March 2018

Wine is sunlight held together by water

The title quotation is usually attributed to Galileo Galilei (along with a lot of other things). Whoever said it certainly had the grapevines' physiology right, although as far as the vines' environment is concerned it should mention both warmth and water.

When we are told about where grapevines will grow and produce nice wine, we are often shown a map something like this first one. This indicates that the world's wine-producing regions are generally between latitudes 30° and 50°, both north and south.

Latitudinal range of vineyard areas

However, this view is rather simplistic, because grapevines don't really know or care about latitude, as such — they avoid the tropics (which is too warm and wet) and the poles (too cold). Besides, the vineyards of places like Sweden are further north than 50°, because it is warm enough for them to grow there.

So, a more refined map looks like this next one, which emphasizes temperature as a prime determinant of vineyard location — the world's wine-producing regions are generally in areas where the median temperature is between 10 °C and 20 °C (or 50–68 °F).

Temperature range of vineyard areas

Clearly, this is not good enough, either. As shown in the map, there are plenty of grapevines outside this temperature range, including those of Sweden.

We could refine the temperature idea by focusing on the temperature solely during the growing season, as in this next map. This particular map is for the world as it was back in 1950 — by the turn of the century the temperature boundaries had detectably contracted north- and south-ward, as the world warms up.

Growing-season temperature range of vineyard areas

This idea was taken to its logical conclusion with the work of Maynard Amerine and Albert Winkler in the 1940s, who developed the idea of Growing Degree-Days — the sum of the daily temperatures above 10 °C (50 °F) during the grapes' growing season. This allows vineyard regions to be classified according to the Winkler Index, indicating which grape varieties are likely to do well: Region I (eg. Chardonnay, Pinot noir, Riesling, Sauvignon blanc), Region II (Cabernet sauvignon, Chardonnay, Merlot, Semillion, Syrah/Shiraz), Region III (Barbera, Grenache, Nebbiolo, Syrah/Shiraz, Tempranillo), Region IV (Carignan, Cinsault, Mourvèdre, Sangiovese, Tempranillo) and Region V (Fiano, Nero d'Avola, Palomino, Primitivo).

However, we still haven't taken water into account in this discussion. Indeed, geographers have long used the idea of combining temperature with water availability (particularly seasonality) to define geographical regions of the earth. Probably the best-known example of this is shown in the next map, where each color represents an area (zone) with a particular climate regime. Clearly, there is more to the location of these zones than simply latitude or temperature isotherms.


The person credited with initiating the drawing of this climate-classification map is Wladimir Köppen, a geographer who worked mostly in Austria and Germany between 1880 and 1940. The final version of his map was a joint work with the German climatologist Rudolf Geiger, and so the map is usually credited as the Köppen-Geiger Climate Classification. The current (updated) version of this map is available from the Climate Change & Infectious Diseases Department at the University of Veterinary Medicine, Vienna, with another version also available on Wikipedia.

For our purposes, we are interested in the following climate zones:
Indeed, almost all of the world's vineyards are in zones Csa, Csb, Cfa or Cfb. The definitions of these zones are:
     Csa — temperate, summer dry, hot summer
     Csb — temperate, summer dry, warm summer
     Cfa — temperate, fully humid, hot summer
     Cfb — temperate, fully humid, warm summer
The "C" classification (named Temperate) refers to the mean temperature of the coldest month being > -3 °C. The "f" and "s" classifications refer to whether there is enough rain throughout the year (f) or little to no rain during summer (s). The "a" and "b" classifications refer to areas with the warmest month ≥ 22 °C (a) or at least four months with ≥ 10 °C (b).

This classification works for our purpose here because of the climate requirements of viticulture:
     mild winters with low freeze / frost risk;
     stable, cloud-free conditions during bloom;
     warm and dry summers;
     dry maturation period with moderate to high daily temperature range.

Köppen-Geiger climate zones around the Mediterranean

The validity of these classification zones can be seen in the above map of the area north and south of the Mediterranean. All of the current vineyard areas fit into zones Csa, Csb, Cfa or Cfb, including those of Sweden. The same is true, for example, of Australasia (South Australia: Csb; Western Australia: Csa, Csb; Victoria, Tasmania, New Zealand: Cfb; New South Wales: Cfa, Cfb), South America (Argentina: Cfa, Cfb; Chile: Csb, Cfb), and South Africa (Csa, Csb).

North America also fits into this scheme, with one group of intriguing exceptions. As shown in the final map, for the USA we have: California (Csa, Csb), Oregon and Washington (Csb, Cfb), and much the rest of USA (mostly Cfa, with bits of Cfb, such as the vineyards of New Mexico, the nation’s oldest wine-growing region). For Canada, the vineyards of British Columbia also fit into zones Csb and Cfb.

Köppen-Geiger climate zones of North America

We would also conclude from this map that certain parts of North America are likely to be very poor for viticulture, including southern Florida (too warm and humid), and most of the Rocky Mountains (too dry at times when the vines need water). To be consistent, this conclusion should also apply to all of the north-central and north-eastern US states, plus all of central and eastern Canada — all of this area is in climate zone Dfb (in purple on the map).

The "D" classification refers to the mean temperature of the coldest month being < -3 °C. This is the so-called Snow climate zone rather the Temperate zone. This zone violates the viticulture requirement of mild winters with low freeze / frost risk.

And yet, there are vineyards throughout this Dfb zone, including the US states of Colorado, Minnesota, Wisconsin, Illinois, Michigan, New York, Vermont and Maine, plus the Canadian provinces of Nova Scotia, Ontario and Quebec. Mind you, none of these are extensive viticultural areas, although the wineries of the Finger Lakes, at least, are hardly unknown.

Grape-growing is not an easy proposition in zone Dfb, which is why few people try it. In the past, some people in these areas have chosen to use native vine species (including Vitis labrusca and Vitis riparia) instead of the Mediterranean one (Vitis vinifera), on the grounds that they will grow happily in their local environment; or people have produced hardy hybrids of these native species with V. vinifera. Nevertheless, most vignerons have persisted with the introduced vine species, and dealt with the climate zone as best they can. Perhaps one of the important climatic factors is the presence near many of the vineyards of large bodies of water: The Great Lakes — this will significantly moderate the effects of freezes and frosts.

Old Man Plain (from Wikipedia)

Almost no-one else tries to grow wine grapes as a paying proposition in climate zones outside the C (Temperate) classification; at least, not without considerable help. For example, Australia's largest wine-producing area is largely in a BSk zone (B = arid, S = steppe, k = cold arid). This area is capable of Western-style agriculture only because it has a massive flood-irrigation scheme (called the Riverina) — the area is otherwise semi-arid (see the photo). When you see a bottle of wine that says it comes from "South-eastern Australia", rather than from a named viticultural region, this is where it originates (eg. Yellow Tail).

Finally, Guido Grimm has recently pointed out to me the existence of a small amount of Tanzanian wine, which is produced in a BSh zone (h = hot arid) of Sub-Saharan Africa.



This post was inspired by a post on the blog of my friend and colleague Guido Grimm (Wladimir Köppen — a climate-vegetation genius, and what this has to do with wine), to whom all credit is due.

Monday, 19 March 2018

Laube versus Suckling — their scores differ, but what does that mean for us?

There seem to be two general attitudes toward professional wine-quality scores. First, they can be seen as the sum of assessments of various sensory "components" of the wine. The classic example of this is the UCDavis 20-point score, which was originally designed to train students in detecting wine faults. This approach has been perhaps taken to its logical extreme in the fascinating book by Clive S. Michelson, Tasting and Grading Wine (2005. JAC International).

The alternative view is that the scores are expert, but subjective, opinions about the quality of the wine. For example, on March 15 1994, in response to a reader query, the Editor of the Wine Spectator magazine noted:
In brief, our editors do not assign specific values to certain properties of a wine when we score it. We grade it for overall quality as a professor grades an essay test. We look, smell and taste for many different attributes and flaws, then we assign a score based on how much we like the wine overall.
This seems to be the approach adopted by most of the professional media, especially when they use the 100-point scale. Some of them claim to be considering wine components individually (eg. complexity, concentration, balance, texture, length, overall elegance), but there is little evidence of this in their final scores.

James LaubeJames Suckling

I have shown in several blog posts that professional wine commentators do not necessarily provide comparable wine-quality scores when tasting the same wine. This can happen for many reasons, including different expertise, different personal preferences, different wine bottles, and different tasting conditions. This is why we seem to both love and hate wine critics. Let's look at this issue in more detail.

An interesting exercise

To look at variation in wine-quality scores, it is of interest to eliminate the last two factors listed above (bottles and tasting conditions), by having the scores be produced from the same bottle at the same time. This, of course, is what happens at most group wine tastings; but rarely do we see published the scores from several people at a single tasting, to make the direct comparison.

However, one pair of commentators where we can do this is James Laube and James Suckling who, at various times, have both provided wine-quality scores to Wine Spectator magazine regarding Cabernet wines, with Laube as the California expert and Suckling as the Bordeaux expert. Suckling has subsequently parted company with the magazine, but Laube remains as their California correspondent.

The dataset I will use here is from the "Cabernet Challenge" of 1996 (see Wine Spectator for September 15, 1996, pp. 32–48), in which the two James tasted 10 California Cabernet blends and 10 Bordeaux red wines from both the 1985 and 1990 vintages. This gives us 40 bottles of wine with which to compare their scores.

The data are shown in the first graph, with Laube's scores vertically and Suckling's horizontally. Each point represents one of the 40 bottles.

Suckling vs. Laube for 1985 and 1990 cabernets

I don't know about you, but this does not look too good, to me, in spite of the fact that Marvin Shanken, as the Editor of the article, claimed: "For the most part, our two critics found themselves in much agreement". To me, there is a wide spread of points in the graph — the scores differ by up to 9 points, with 5 of the bottles differing by more than 6 points. Furthermore, the mathematical correlation indicates only 29% agreement between the two sets of scores.

However, it is worth noting that the average scores from the two critics are almost identical (90.5), with very similar maximum (100 vs. 98) and minimum (both 82) scores. On average, Laube gave slightly higher scores to the California wines than to the Bordeaux wines; and Suckling gave lightly higher scores to the Bordeaux wines than to the California wines.

Now, let's look at what we might expect from critics who do agree. This next graph shows what perfect agreement would look like (the solid line) — for bottles whose points are on this line, the two James perfectly agreed with each other. Clearly, this is only 5 out of the 40 bottles. The Laube score is > the Suckling score 18 times, and 17 times it is the other way around.

Suckling vs. Laube for 1985 and 1990 cabernets

The two dashed lines in the graph show us ±2 points from perfect agreement — for bottles between the two lines, the two James' point scores were within 2 points of each other. This allows for the approximate nature of expert opinions — technically, we are allowing for the fact that the scores are presented with 1-point precision (eg. 88 vs. 89 points) but the experts cannot actually be 1-point accurate in their assessment.

There are only 23 of the 40 bottles (58%) between the dashed lines. So, even when we allow for the approximate nature of expert opinions, there is not much more agreement here than there is disagreement.

Another way of dealing with the approximate nature of expert scores is to greatly reduce the number of score categories, so that all the experts need to do to agree is pick the same category. This is the reasoning behind using star scores instead of points (eg. 3 or 5 stars), or word descriptions instead of numbers. The Wine Spectator does it this way:
95 – 100
90 – 94
85 – 89
80 – 84
75 – 79
50 – 74
 Classic: a great wine
 Outstanding: a wine of superior character and style
 Very good: a wine with special qualities
 Good: a solid, well-made wine
 Mediocre: a drinkable wine that may have minor flaws
 Not recommended

So, I have shown this scheme in the third graph. For bottles within the boxes, the two James' point scores agree as to the word categories of wine quality. Once again, this is only 25 of the 40 wines (63%). So, even this broad-brush approach to wine quality assessment provides only two-thirds agreement between the two critics.

Suckling vs. Laube for 1985 and 1990 cabernets

As an aside, it is worth noting the overall low scores given to the wines. Only 17 of the wines scored >90 points, even though they are all quite expensive. The only one of the 40 wines that I have tasted is the 1985 Château Mouton-Rothschild, and I was no more impressed by it than was either of the two James (85 vs. 89 points).

What does this mean for us?

The magazine is presenting their scores as representing some sort of Wine Spectator standard of quality, but clearly this is not an objective standard of quality. The scores are personal (but expert) judgments by their individual critics, who may have very little in common. At issue here is whether quality is an intrinsic property of wine, or whether it is mainly context dependent (see Jamie Goode).

The formal explanation for the degree of disagreement is this: the tasters are not using the same scoring scheme to make their assessments, even though they are expressing those assessments using the same scale. This is not just a minor semantic distinction, but is instead a fundamental and important property of anything expressed mathematically. As an example, it means that when two tasters produce a score of 85 it does not necessarily imply that they have a similar opinion about the wine; and if one produces 85 points and the other 90 then they do not necessarily differ in their opinion.

This situation is potentially a serious problem for all wine-quality assessments, when the scores represent expert, but subjective, opinions. Scores will look the same because they are written using the same scale, and people will inevitably try to compare them. But, if the scale does not have the same meaning for any given pair of people, then the numbers cannot be validly compared, because they have different meanings.** Not only would we be comparing apples and oranges, we would be comparing different (but unknown) numbers of apples and oranges. What is the point of that?

I will look at the mathematical consequences of this topic in a future post, illustrating the issue with a well-known data set.

Finally, one practical consequence of this mathematical characteristic is clearly being exploited by wine marketers. When looking at these scores on the web, it quickly became obvious that the wine stores are simply choosing to report the higher of the two critics' scores, when advertising any of the 40 wines, almost never producing both scores. This is an interesting example of "cherry picking".

Reproduced from Robert Dwyer at Palate Press

Thanks to Bob Henry for all of his help with this post — he has long championed the use of standardized wine-quality scoring schemes, often in vain.

** As a specific example, here are quotes from each of the two critics. James Suckling: "I was more concerned with the texture and aftertaste of the wines than with their aromatic qualities or flavor characteristics." James Laube: "I like my wines young, rich, highly concentrated and loaded with fruit flavors."

Monday, 12 March 2018

The rise and rise of New Zealand wine

Wine from New Zealand has very much become flavor of the month in many parts of the world. According to the recent book by Warren Moran (New Zealand Wine: The Land, The Vines, The People. 2017), the New Zealand wine industry only really got going 50 years ago, and as recently as 1990 Müller-Thurgau was it’s principal grape variety. However, the NZ vineyard area has been growing rapidly for the past 20 years (see the first graph), since it started to focus on Sauvignon blanc, instead. And yet, even the Wine Advocate did not have its first comprehensive report on New Zealand until 2008.


Wine exports have continued to increase along with production, particularly at the premium end of the market. I thought that a good way to explore this might be by comparing recent exports trends with those of their nearest geographical and socio-political rivals, Australia.

The wine export data come from Wine by Numbers (Il Corriere Vinicolo n. 23, July 2017), which covers the years from 2005 to 2016, inclusive, plus the recent Wine Export Report for 2017. This next graph illustrates the annual trends for both countries, using Australian dollars for the wine export value.

Annual wine exports from Australia and New Zealand

As I noted in a previous post (Getting the question right), Australia has had its ups and downs in the export world this century. Australian wine was itself the flavor of the month back in the 1990s, when good wine was available from Australia more cheaply than from many other countries. Times have changed since then; but a comeback of sorts appears to be in progress, starting from 2014.

By comparison, the rise and rise of New Zealand wine exports has been inexorable. They haven't caught Australia's value yet, nor do they look like they will do so any time soon; but this may depend on exactly what the exports markets consist of. Apparently, the New Zealand wine industry has a target of NZ$2 billion of exports by 2020 (New Zealand Winegrowers Annual Report 2017).

Annual wine exports from Australia

So, let's take a look at Australia, first, for comparison, before we turn to New Zealand in more detail. The third graph shows Australian wine exports (to 2016) by general world region.

As you can see, the North America (mainly the USA) and Europe (mainly the UK) patterns run in parallel. The peak export period coincides with the initial rise of the Yellow Tail brand. The decline, on the other hand, at least partly reflects the decline of what was then Australia's largest wine company, Southcorp. This decline occurred because they were involved mainly in the budget end of the wine spectrum, where volume sales are essential, margins are tight, and sustainable returns are not likely. This conglomerate was bought (in 2005) by a different conglomerate, the Fosters Group (Australia's biggest beer company), which later (2011) split it's wine-making arm off as what is now Treasury Wine Estates (currently Australia's largest wine producer by total revenue).

The sharp rise in the pattern for exports to Asia after 2014 reflects the fact that the Australian wine industry is now specifically targeting the premium wine market in China (the world’s fastest-growing major wine market). Notably, in 2012 China became the biggest destination for Australian bottled exports sold at AU$7.50 per liter and above; and currently China accounts for more than half of Australia’s exports of wine at AU$10 or more per liter. China has become a premium market.

Annual wine exports from New Zealand

Turning now to New Zealand, the same three geographical regions all show a different pattern from that of the Australian exports. Exports to the Asia/Oceania region principally reflect sales to Australia (not China, yet). These exports have been dominated by Sauvignon blanc, of course, which apparently now accounts for 75% of all NZ wine exports; and since 2008 this has actually been the biggest selling white wine in Australia. This market seems to have been saturated since 2012, with no further increase in sales.

Exports to the European region are principally to the UK, as they are for Australia. The graph shows that this market may also be becoming saturated for NZ wine, with a recent plateau in sales. This means that the North America market is now the biggest one for New Zealand exports, mainly to the USA; and it is this market that has shown an inexorable rise since 2011.


This suggests that we should look at the data in a bit more detail, by country rather than region. The data below also come from Wine by Numbers, this time the annual edition for 2016, which covers the year 2015 only.

By volume, almost exactly two-thirds of New Zealand's 210 million liters of wine exports for 2015 were bottled wine, as opposed to the remaining exported in bulk. The three biggest destinations for bottled wine were the USA (28%), the UK (24%) and Australia (24%), while those for bulk wine were the UK (36%), the USA (30%) and Australia (24%). So, most of the NZ wine (81%) went to just three countries (from where it may have been re-exported, of course), with the USA now their global biggest market.

However, more revealing is the premiumization of new Zealand wine. We can see this by looking at the dollar value per liter of the bottled wines being exported to various of the top countries, as shown in the table.

Country
Japan
China
Singapore
Hong Kong
United Arab Emirates
Canada
Denmark
USA
UK
Sweden
Australia
Germany
Ireland
Netherlands
NZ$ / liter
17.26
15.01
13.45
12.57
10.38
10.37
9.07
8.52
8.16
8.04
7.97
7.44
7.10
6.81

This indicates that the Asian market takes principally premium bottled wines, whereas the North American and European markets also take some of the cheaper stuff. Indeed, none of the bulk wine actually goes directly to Asia, while France takes slightly more expensive bulk wine than does anywhere else. Premiumization is an important topic in the world of wine marketing, and the New Zealanders seem to have learned the lesson — sometimes, less is more, financially.


Unfortunately, the combined bulk and bottled export figures do seem to show that the New Zealanders are dumping their cheap wine principally in Australia, a country that already has enough cheap wine of its own. The only export countries where the price per liter of NZ bottled wines was cheaper than Australia were: the Netherlands, Ireland and Germany; and combined these countries imported only 28% of the volume sent to Australia. Similarly, the only bulk-export country where the price per liter was cheaper than Australia was Canada; and it imported only 3% of the volume sent to Australia.

The Australia — New Zealand rivalry clearly continues!

Monday, 5 March 2018

When did California become a "red wine" state?

The short answer is: in 1998.

California is often perceived as a state that specializes in the making of red wines. However, that image seems to be set mainly by the wines of the Napa Valley, where Cabernet sauvignon wines are king. Indeed, the "cult" wines of the USA principally come from this region, enhancing the state's reputation, which is often the focus of wine writers.

However, the most widely planted grape variety is actually Chardonnay, not Cabernet, with 91,333 bearing acres in 2016 versus 84,584 acres, respectively (California Grape Acreage Report 2016 Crop). Furthermore, Chardonnay comprised 14.5% of the 2017 grape-harvest crush, versus 14.2% for Cabernet sauvignon (California Grape Crush Report Preliminary 2017).

Indeed, the California Department of Food and Agriculture's data show that it was not all that long ago that the white-wine grape crush exceed that of red-wine grapes. The data in the first graph are from the Crush Report for 2017, and show the crush size for each of the years from 1989 to 2017, inclusive, categorized by grape type. Eating grapes include both Table (fresh) grapes and Raisin (dried) grapes.

Crush tonnage of grapes in California through time

Obviously, before 1996 more white grapes were crushed than red, while the two crushes were almost identical in 1996 and 1997. Since then, the red crush has progressively outstripped the white. Note, also, that in 1992 the red-wine grape crush was even less than that for the edible grapes!

So, red-wine grapes have been the most important in California only for the past 20 years. This is not necessarily surprising, as the Bordeaux wine region of south-western France, globally recognized for its red wines, only moved from mostly white wines to mostly reds in the 1970s.

The most likely reason for this change in emphasis is also revealed in the same Crush Report. The following graph, taken directly from that report, shows the price paid for the various grape types during the past 10 years.

California grape price through time

The graph makes the interesting point that the prices of Table grapes, Raisin grapes, and White-wine grapes have not changed much over the past decade. However, the price or Red-wine grapes has shown an increase of c. 50%. The has been distinctly so in the Napa Valley, of course.

The report also shows that in 2017 the Cabernet sauvignon production was up 17% above the 5-year average, while the Chardonnay production was down 13% below the 5-year average. Indeed, the Chardonnay crop was the smallest since 2011, although it remained just ahead of Cabernet as the biggest percentage of the crush (as noted above).