Enologix?

[ChocoKitty]

Did anyone read the "Grapes of Math" article in this month's Wired magazine?  It's about a wine consulting company called Enologix -- they apparently guarantee that a winery using their method will obtain a certain score from Parker or Wine Spectator.

The article brought up some interesting viewpoints about how wine is rated, whether it's subjective or objective and whether critics' preferences and certain wine styles can be boiled down to a particular formulaic wine fingerprint.  At least I found it fascinating (but then, I'm a techno-geek as well as a food nut).

[Fat Guy]

The article isn't on the Wired site, but if I understand the information there correctly it will be posted on November 13. Perhaps someone will post the link then so we can discuss it.

[ChocoKitty]

Here's the link:

http://www.wired.com/wired/archive/9.11/wine.html

[Fat Guy]

Thanks!

[NewCat3]

Wine people interested in technology and wine ratings have tended to mention the story. The President of one of the exhibitors at MacWorld mentioned the story.  My sense is that technology will benefit the artisan winemakers much as CAD programs put power back into the hands of architects.  In other words, winemakers who can simulate quality of blends prior to blending wines must have a tremendous advantage. It enables them to increase volume or quality and estimate the outcome.  Wired really does cover the future.  

[Bux]

I met a wine maker in the Cotes du Rhone. Parker had just given his wine a pretty good rating and he was pleased that his bank was pleased. He noted that winemakers in the area speak of "parkerizing" their wines. I.e., making them to Parker's taste.

As for CAD empowering architects, it's hard to imagine Gehry's Guggenheim being built without a computer to put it on paper.

[NewCat3]

Bux mentioned Robert Parker who I respect in so far as being a consumer advocate.  However both the French and California wine industry need to figure out how to gain control of the ratings if they are really going to use Parker's 100-point scores to protect small wine farms in Europe and California wineries.  After all one day Parker will be gone, then what? Maybe the company in the Wired story can morph into a wine industry based ratings authority that protects small winemakers.  What does anyone think?

[project]

Okay, at

http://www.forbes.com/forbes/2004/0216/111_print.html

is another story on Enologix and its founder Leo McCloskey.

The Wired story is mostly better.

Did wonder why the Forbes story did not mention UC Davis more. The Wired story explains: UC Davis doesn't like McCloskey! Apparently UC Davis has fallen into a conceptual trap: They wanted measures with reliability and validity of 'fundamental' flavor sensations. Okay. So, this is an approach to 'reductionism' but, really, a weak one since taking that 'flavor wheel' and saying how to blend to get something that tastes good is missing.

Here is another and more direct, useful, reasonable, and valuable goal: Given a bottle of Chateau Effete Rotchild 1949, and given all these grapes, how do I harvest, make wine, and blend and age the wines to get something that tastes very much the same? Perfectly reasonable goal, and it's 'direct' in that we don't go through a lot of intermediate 'flavor wheel' analyses to say how.

For how, for what chemicals to measure, here's a guess: Looking at many wines, can begin to guess the top, say, 100 most important chemicals. Then, match Chateau Effete Rotchild on those 100, and might be pretty close. If extend to the top 200 chemicals and match on those, and otherwise match on grape variety, soil chemistry, sugar content and pH at harvest, details of vinification, use of egg whites, oak barrels, aging, etc., will likely also match reasonably well on the rest of the 50,000 or whatever chemicals are there and come really close. Then you're close to Chateau Effete Rotchild as far as human tasters are concerned. Then for any human buying a bottle of wine to drink, buy the imitation and why pay more?

The fundamental 'science' would not like this because this process did not identify 'fundamental' measures like velocity, acceleration, mass, energy, and momentum in physics. Instead, the process just matched on a lot of chemicals and process steps and hoped that the rest would just fall into place -- which it likely will.

To heck with the 'physics envy' of the 'fundamental science': Instead, let's just make huge quantities of great wine and get rich!

If want to match a Parker rating, proceed similarly. That is, get some of the wines Parker tastes, analyze those, and make some sense out of what combinations of the 100 top chemicals or so get what rating from Parker.

The Parker ratings and 'Parkerizing' need not be very relevant: If you like Montrachet, Meursault, Chambertin, Hermitage, classic Barolo or Chianti, or even a Greek Muscat, get a bottle, analyze it, match it, and drink the match.

This matching does not really permit making 'artificial' wine just from water, ethyl alcohol, and some chemicals: The matching on 100 chemicals or so promises to work only because the ingredients really are well made wines so that matching on the 100 chemicals promises to match on much more automatically.

Sounds to me like UC Davis got lost in some objectivity that was nearly irrelevant. There really isn't anything less objective about matching some given Chateau Effete Rotchild. Just match it, bottle it, sell it, and let people buy it and drink it. Of course, this process takes Chateau Effete Rotchild as a given and does not explain just why it is 'good'; we can sincerely promise to worry about this each day on our walk to the bank!

One curious point about the Forbes article is that they mention that Hahn Estates winery has 1000 acres and are shooting for 400,000 cases in 2004 and that Diamond Creek's wines sell for $400 a bottle. Hmm.

Gee, if Leigon could get even $100 a bottle at the winery for 400,000 cases in one year, then that would be

400,000*12*100 = $480,000,000

which is really serious money for a 1000 acre farm anywhere in this solar system. Heck, $10 a bottle would be serious money.

Sounds like wine prices will be coming down or we will be awash in wines. Yes, it's expensive to grow wine, but it's not a half million dollars an acre expensive!

[project]

Okay, let's see:

Suppose we are given a bottle of Chateau Effete Rotchild and a list of 300 other roughly similar wines and asked to blend the 300 wines to be like the Chateau Effete Rotchild. Suppose we measure the concentration of 100 relevant chemicals in each of the 301 wines. Suppose we are given prices per bottle for each of the 300 wines we are to use for blending.

Now, can we blend the 300 wines to match the Chateau Effete Rotchild on the 100 chemicals?

If so, then how do we do this at least total cost?

Hmm ....

It turns out, this problem has been investigated for about 55 years. The problem is now very thoroughly understood. The results are in the form of powerful theorems, elegant proofs, refined algorithms, very high quality software, etc.

Turns out, for this problem, the solution would be a piece of cake!

One result: If we can do the blending at all, then we can do it at least cost for no more than 100 blending wines -- we won't need the other 200.

How 'bout that!

[ctgm]

Blending is not as simple as all that:

Let's say you blend X and Y in March to get Z

and the same X and Y in September to get Z

the two Z wines will taste different.

The software may well tell you all the chemical components but is it going to tell you when to blend?

[balex]

Okay, let's see:

Suppose we are given a bottle of Chateau Effete Rotchild and a list of 300 other roughly similar wines and asked to blend the 300 wines to be like the Chateau Effete Rotchild. Suppose we measure the concentration of 100 relevant chemicals in each of the 301 wines. Suppose we are given prices per bottle for each of the 300 wines we are to use for blending.

Now, can we blend the 300 wines to match the Chateau Effete Rotchild on the 100 chemicals?

If so, then how do we do this at least total cost?

Hmm ....

It turns out, this problem has been investigated for about 55 years. The problem is now very thoroughly understood. The results are in the form of powerful theorems, elegant proofs, refined algorithms, very high quality software, etc.

Turns out, for this problem, the solution would be a piece of cake!

One result: If we can do the blending at all, then we can do it at least cost for no more than 100 blending wines -- we won't need the other 200.

How 'bout that!

Not at all.

You are only allowed a convex combination of the wines that you are mixing and there is no guarantee that your fancy wine is going to be in the convex hull of your cheap blending wines.

In any event, ctgm's point is valid: wines are not static linear combinations of chemicals but complex dynamical systems with interesting non-linearities in the ester reactions etc.

Still an open problem, IMHO.

[project]

balex:

"Not at all.

You are only allowed a convex combination of the wines that you are mixing and there is no guarantee that your fancy wine is going to be in the convex hull of your cheap blending wines.

In any event, ctgm's point is valid: wines are not static linear combinations of chemicals but complex dynamical systems with interesting non-linearities in the ester reactions etc.

Still an open problem, IMHO."

Wow! Fantastic! Someone on eG understands convex combinations!

Soon enough we will be into one of Caratheodory's results, Farkas lemma, and the Kuhn-Tucker conditions, constraint qualifications, etc. -- from each of which will need at least a bottle of wine to recover! Gee, why weren't Kuhn and Tucker in the movie on Nash?

I've been wrong before, but here I'm still essentially correct!

You claim that we cannot necessarily blend the esteemed coveted Rotchild from just any given 300 blending wines. Right. This is easy to see: Start with 300 bottles of water! Short of some religious experience, we can't do it!

I didn't claim that we necessarily could do it! Instead, I first asked:

"Now, can we blend the 300 wines to match the Chateau Effete Rotchild on the 100 chemicals?"

Good question. The answer involves what you said: Is Rotchild a convex combination of the 300 wines? This is an important question. Turns out, the answer is not trivial. But, the answer is available -- we do have the means! Of course, the answer is, we write out a problem in linear optimization with 101 equality constraints and 300 non-negative variables and ask if the problem is 'feasible'. If we attack this with the simplex algorithm, then we either discover that we can't do the blending or find a way to do the blending using at most 101 of the blending wines.

Then I continued:

"If so, then how do we do this at least total cost?"

So, here I did qualify the effort: Didn't try to find a least cost blending unless we already knew that some blending was possible.

For

"In any event, ctgm's point is valid: wines are not static linear combinations of chemicals but complex dynamical systems with interesting non-linearities in the ester reactions etc."

I was unsure of just how to respond. It's been a long time since I read Wagner's books on wine making!

It seemed that ctmg was assuming that we were considering blending as the wines were still changing rapidly. Ctmg didn't explicitly mention chemical reactions caused by the blending.

My first reaction to ctmg would be that we are trying to blend 'wines' which should mean after the pressing and more important fermentation steps are done. So, we are considering blending once we do have some liquids that are reasonably stable over a time interval of a few weeks.

My second reaction would be, if ctmg can tell us just how the chemical compositions vary over time, then in principal we could say when to blend at least to get a good result just after pouring and mixing -- assuming that the blending did not cause chemical reactions. Since wine aging has been studied, easy enough to guess that some changes could be predicted over time.

My third reaction would be that blending gets to be a tough subject if we are considering blending to cause chemical reactions the results of which will over some time interval, short or long, give us the Rotchild we crave. Instead, as you noticed, I was just assuming that the little molecules would just swim around after blending as before and not change due to the blending itself. Or, I was assuming that mixing and aging are commutative operators!

Of course wine making is awash in chemical reactions, and aging involves chemical reactions. But blending is not supposed to be as active as pouring vinegar over baking soda!

Blending in wine making has been around for a long time. If the process of blending was wildly nonlinear (blending and aging not commutative), easy to believe that the intuitive techniques would just have become lost. Instead, likely a good first-cut approximation is that for reasonably stable 'wine', blending is 'linear' or close enough for the approximations being attempted.

But, maybe I'm wrong: Maybe the extra tannins in one wine will affect the esters in another.

I do agree that there are likely many open questions in blending wines to approximate the flavor of another wine.

Edited February 14, 2004 by project (log)

[balex]

Of course wine making is awash in chemical reactions, and aging involves chemical reactions. But blending is not supposed to be as active as pouring vinegar over baking soda!

Blending in wine making has been around for a long time. If the process of blending was wildly nonlinear (blending and aging not commutative), easy to believe that the intuitive techniques would just have become lost. Instead, likely a good first-cut approximation is that for reasonably stable 'wine', blending is 'linear' or close enough for the approximations being attempted.

But, maybe I'm wrong: Maybe the extra tannins in one wine will affect the esters in another.

I do agree that there are likely many open questions in blending wines to approximate the flavor of another wine.

Basically I agree with you -- but in my experience with wine the more I know, the more I know I don't know.

In particular I think there are a number of assumptions that are problematic.

First there is the assumption that it is only a few (100) chemicals that are important. Secondly what we are intersted in is not chemical similarity but sensory similarity -- and it is not clear that unscaled Euclidean distance to the target wine is the right metric. Thirdly there is the issue of chemical stability that you touch on.

I think people have really no clue about exactly what goes on chemically in a bottle of wine over fifty years. The holy grail is to type "Chateau Cheval Blanc 1947 as it was in 1985" into your InstaWine box and have it dispense a glass. But it is a long way to that point.

[project]

balex:

"First there is the assumption that it is only a few (100) chemicals that are important. Secondly what we are intersted in is not chemical similarity but sensory similarity -- and it is not clear that unscaled Euclidean distance to the target wine is the right metric. Thirdly there is the issue of chemical stability that you touch on."

Of course we have to guess that more than 100 chemicals are important. So, if we just started with 100 bottles of pure chemicals, then blending would likely be hopeless. But if we are seeking Montrachet 2001, have 300 Chardonnay wines all made to imitate Montrachet, and match on 100 chemicals, then we are likely matching fairly well on more than just the 100 chemicals we are considering.

First-cut for what you mentioned with convex combinations and what I outlined in linear optimization, we are not really using a Euclidean metric; instead we are matching on the 100 chemicals essentially exactly.

But a Euclidean metric could have a role: If the linear optimization problem with the 300 blending wines and the 100 chemicals was infeasible, then we could look for an approximate solution. The usual way to proceed would be to solve the convex quadratic problem of minimizing the sum of squared errors. Here, then, as you suggested we would be 'scaling' so that we would essentially be willing to accept being 1 microgram off on some tannin as equivalent to being 1 microgram off on some ester. Instead of minimizing the sum of squared errors, just minimizing the absolute value of the worst error is also attractive and in some ways easier.

But we are implicitly exploiting some more things:

First, we hope that as the number of chemicals we would match on grows toward infinity and past the actual number of chemicals in any bottle of wine, we really would get a match, essentially an exact match. So what we are doing with 100 chemicals is one correct step on an asymptotic process that in the limit will match exactly. Or, if 100 is not large enough for good accuracy, then try 200; eventually we will have to win (modulo the issues of chemical reactions, etc.).

Second, we hope that whatever 'topology' is important for 'continuity' in the sense of the taste of one wine being close to that of another is the same as the 'usual' topology from the Euclidean metric. Then, any asymptotic process that converges in some one metric will converge in the topology and also in all metrics that generate that topology.

We can hope that, given some complex system where we take measurements on several related variables, approximate exactly on all these variables, and continue as the number of variables grows to infinity, in the limit we will represent the system exactly. So, to test this, let's roll back to something simple: For numbers x between 0 and 3, consider the function f(x) = 1 for 1 <= x <= 2 and f(x) = 0 otherwise. Not a complicated function! For some positive integer n, pick distinct points u(i), i = 1, 2, ..., n, between 0 and 3 and set v(i) = f(u(i)). Now we have pairs of numbers (u(i), v(i)). We can write down a polynomial of degree n - 1 that will pass through all n points exactly. Now, as we increase n and pick the points u(i) to be dense between 0 and 3, do we get a sequence of polynomials that converges to f(x) exactly for all x? No! Instead we get polynomials that oscillate wildly near 1 and 2. Ah, hopes dashed again!

So, we are hoping that nothing analogous holds in wine blending. One hope is that in practical terms there are only finitely many relevant chemicals, but this alone is less comfort than we could want.

But in many areas we hope nothing holds analogous to the oscillatory polynomial disaster: We can write out what we want in an ideal girlfriend -- height, weight, age, hair color, eye color, education, talents, interests, accomplishments -- and not fear that getting what we do specify increases the chances of getting something really bad where we did not specify -- voice of Boris Karloff, smile of Bela Lugosi, mating behavior of a Black Widow spider, etc. Hmm .... Interesting: Where does specifying what we do want increase the chances of getting something especially bad where we did not specify?

Or, we can squeeze the toothpaste tube as hard as we want in as many places as we want and we will still not succeed in making the tube smaller; instead we will only cause the tube to bulge out more in some place we didn't squeeze. If we ask our ideal girlfriend to play the Bach 'Chaconne', the Paganini 'Caprices', and the the Bruch 'Scottish Fantasy', then she may have had less time for cooking, sewing, hair, nails, and hostess skills! Or, as we know, at times there are trade-offs!

So, if we want to blend Chardonnay wines from California, Chili, Australia, Italy, and even around Macon to match Montrachet, does matching one long list of chemicals increase the chances of something really bad on some chemical we ignored? Does the terroir and microclimate of Montrachet make that wine something really different in the way a step function is not a polynomial so that any attempt to match on some chemicals will just cause a really bad mismatch on some other chemicals? My guess is no, but it would be good to have some solid support for such a guess.

Ah, we get all the esters and tannins just right, and at the triumphant victory tasting all anyone can taste is hydrogen sulfide! Naw: Since the only linear combinations we are doing are convex combinations, the hydrogen sulfide concentration in the blend can never be higher than in any one of the blending wines!

The idea of measuring 100 variables from a complex system and asking if these are the 'right' 100 variables, how should we 'scale' these 100 variables, etc. are quite broadly important. These questions are not simple to pose well, but some progress is needed.

Here's one: We take 299 Chardonnay wines and one additional wine, measure the concentration of each of our 100 chemicals in each of the 300 wines, and ask for a non-trivial test with rate of false positives of 1% "is the additional wine a Chardonnay?". We take 299 rocks from a valley and one additional rock, get data on each of several variables for each rock and ask if the additional rock was from the valley? Similarly for many cases of looking for anomalies.

[balex]

What I meant about sensory similarity was more about things like 'masking' where the presence of one chemical reduces the sensory importance of another. TCA is an example. Now if we are not measuring the presence of both of these chemicals we are completely screwed. Even if both of these are in our low dimensional subspace, this still induces non-linearities. In particular since I think exact matching is not feasible, we want to find the point in the convex hull of the mixers that is closest to the target. IF we have some wierd function that measures sesnory distance, then the sets of points equidistant from the target wine will not be hyperspheres -- indeed might not even form convex stes -- in which case the whole problem ceases to be convex and you have multiple local optima and probably need to use some stochastic optimisation to find the best solution , and that's assuming we have good model of the distance function.

And the number of chemicals in a mature wine is probabky really large. TCA can be detected by humans at concentrations of a few parts per trillion. So in principle we could be talking about a billion to a trillion different chemicals.

[project]

balex:

"What I meant about sensory similarity was more about things like 'masking' where the presence of one chemical reduces the sensory importance of another. TCA is an example. Now if we are not measuring the presence of both of these chemicals we are completely screwed. Even if both of these are in our low dimensional subspace, this still induces non-linearities. In particular since I think exact matching is not feasible, we want to find the point in the convex hull of the mixers that is closest to the target. IF we have some wierd function that measures sesnory distance, then the sets of points equidistant from the target wine will not be hyperspheres -- indeed might not even form convex stes -- in which case the whole problem ceases to be convex and you have multiple local optima and probably need to use some stochastic optimisation to find the best solution , and that's assuming we have good model of the distance function.

And the number of chemicals in a mature wine is probabky really large. TCA can be detected by humans at concentrations of a few parts per trillion. So in principle we could be talking about a billion to a trillion different chemicals."

If we are to try to blend some 300 Chardonnay wines to duplicate Montrachet, then we should start with good Chardonnay wines. Wines with strong off flavors will not be very useful. So, even if we match on only 100 chemicals or so, we would have to analyze for more, e.g., for TCA from bad corks or wherever it can come from. For TCA, we would likely not be 'matching' on it but just 'rejecting' blending wines based on it.

If we have 300 blending wines and want to match Montrachet on some 100 chemicals, then mathematically either the Montrachet is in the convex hull of the 300 wines or it is not. If it is in the convex hull, then at least mathematically we can match exactly -- as usual, modulo no chemical reactions. Moreover, as noted, we will need at most 101 wines to do the matching. In practice, such matching would be as close as we can expect from finite precision computer arithmetic and quantitative chemistry, and these are quite precise.

In any normed linear space, the norm generates a metric, and in that metric the collection of all points within a given distance of a given point is convex. So, if flavor corresponds to some such metric, then 'spheres' will be convex.

Yes, flavor sensations are tricky to understand. I cut back on the amount of heavy cream and suddenly discovered that the prior amount of salt was too much!

But we do not really need some version of 'convexity' to approximate flavors. Something like continuity in the usual topology is enough. In this case, at least in principle, and considering only the chemicals we are matching on, however close we want to be in flavor, there will be some Euclidean distance on the space of vectors of chemical concentrations such that any mixture of the chemicals as close as the distance in chemical concentrations to our coveted Montrachet will also be as close as we have specified to that Montrachet on flavor. E.g., take 100 gallons of our desired Montrachet and put in one tiny drop of distilled water or really good Chardonnay and try to tell the difference!

Here we do not have to assume that the set of all chemical vectors as close as we desire on flavor will be convex. In principle this lack of convexity could complicate finding a least cost blend within a given closeness on flavor. I doubt that this situation would be very important in practice because there we would likely just be going for a really close match on the chemical concentrations and not trying to exploit flavor 'masking'. Perhaps we would, then, be leaving ourselves open to someone with a really good nose stirring up some swill from dried cattle blood, ground wood bark, vanilla, and plums and bettering our efforts at matching Chambertin -- well, maybe someone's imitation of Chambertin!

Without convexity, for finding least cost approximate matching, we would be into nonlinear optimization. First cut, we would look for the Kuhn Tucker necessary conditions. Ah, in convex combinations, all the constraints are linear so that there is no problem with constraint qualifications!

The meaning of 'stochastic optimization' is ambiguous:

One meaning is solving a problem that evolves with some unpredictability over time. So, an example would be saying how to call plays in football. A key here is the possibility for 'recourse' where we don't have to call the play for second down until we see the results from the play on first down; moreover, the play we call on first down should be in light of this. This whole subject is fairly well understood mathematically and there is called 'stochastic optimal control'. The now famous but relatively simple example is the Black Scholes option pricing model.

Another meaning is that we have an optimization problem to solve, a 'deterministic' problem, that is, one without uncertainties, and we make some use of a random number generator to guide our search. In such things, usually we scatter a lot of points, like sowing seeds, and then try to make improvements locally from each of those. We have essentially been assuming that our blending problem was deterministic. Alas, to use this technique for wine blending (for the case of least cost approximate matching on chemical concentrations to come within some specified closeness on flavor), we would have to have some quite impressive details on human taste sensations or 'put humans in the loop' to do a lot of tasting -- wonder if we could recruit the panel members on eG, after we warned them on just how much tasting there would be?

Yes, if in our blending wines we have to pay attention to parts per trillion and if those wines could contain just anything, then in principle we would have to analyze all chemicals that could conceivably be present in parts per trillion, would likely have to analyze billions, maybe many more than trillions, of chemicals, and would need a comparable number of blending wines for an exact match. This situation strikes me as too pessimistic.

Gee, we can take a grape, Pinot Noir, Cabernet Sauvignon, Merlot, Chardonnay, etc., load up some oak barrels, head for Italy, California, Chili, Australia, etc., select ground, plant, grow, prune, grow, harvest, crush, ferment, rack, filter, age and do these things well and come up with excellent wine, even wine comparable with the originals in Bordeaux or the Cote d'Or. Here we have 'matched' on only a few things -- the grape, the yeast, the barrels, etc. -- maybe dozens, not millions.

For the whole issue of blending to match a given excellent wine: Easy? No. Possible? I suspect so.

[Bob Marley]

Yikes! You math geeks are making my brain hurt. Enologix is interesting, and controversial, but I don't think that even the founders are suggesting that you can reproduce an existing wine. They are merely suggesting that they can compare your wine with others, using known characteristics, and then make an educated guess as to how it will score, again compared with other known wines/scores. Then, they may make a suggestion as to which compounds were different in your wine vs. the know high-scorers, and what might bring your wine more in-line with others that performed better.

THE BOTTOM LINE IS THIS: IF WE PUT LESS IMPORTANCE ON THE PALATES OF A FEW, VERY SUBJECTIVE CRITICS WITH QUESTIONABLE TASTE, AND JUST BOUGHT AND DRANK WHAT WE LIKED, THEN COMPANIES LIKE ENOLOGIX AND LONG HEADACHE CAUSING POSTS WOULD BE TOTALLY UNNECESSARY! WE COULD SPEND THE TIME THAT WE WASTE READING BIASED SCORES AND OPINIONATED POSTS DRINKING GREAT WINE.

Well, its just a thought......

[project]

Bob Marley:

My posts in on this thread were mostly to make a point close to your objection about "THE PALATES OF A FEW, VERY SUBJECTIVE CRITICS WITH QUESTIONABLE TASTE".

So, my original point was:

"Here is another and more direct, useful, reasonable, and valuable goal: Given a bottle of Chateau Effete Rotchild 1949, and given all these grapes, how do I harvest, make wine, and blend and age the wines to get something that tastes very much the same?"

And we don't have to use "the palates of a few", either. Or, one possible goal of the work would be just to mix up some wines like ones we already know we like!

For "I don't think that even the founders are suggesting that you can reproduce an existing wine". Right. To "reproduce an existing wine" was my guess at one possible goal for the future.

[Florida Jim]

For those who have access to the N.Y. Times Sunday Magazine this week, on page 36 begins an article about a company called Enlogix, which arose in 1993, and its owner and principal architect, Leo McCloskey. Evidently, Mr. McCloskey has discovered the secret formula which makes any wine capable of scoring 90+ points from the noted critics in the industry and hence, more valuable in the marketplace. To quote the article’s author, David Darlington, Mr. McCloskey, “. . . insists that high-scoring wines can, through chemical analysis, be scientifically proved to be the best wines on the market. In other words, there is accounting for taste.”

In brief, Enologix divides wines into four categories. For reds, style 1 is pale in color and low in tannin (pinot noir, mostly); style 2 is pale in color but higher in tannin (eg. Barolo); style 3 is dark in color and high in tannin (many cabernet sauvignons and Bordeaux); and, style 4 which is dark in color and lower in tannin.

The last category is where one finds the majority of successful, flagship mainstream wines, the most elegant and popular wines in the world, according to McCloskey. And clearly, this is the category that is most favored by wine critics with their scores and the one that winemaker’s should aspire to.

One of the most interesting and compelling quotes attributed to McCloskey is that, “By 1990 everybody was discrediting the Score, but I saw that the critics were going to win because Americans wanted to reduce their risk of purchase and winemakers weren’t filling the information void.”

Mr. McCloskey’s ideas are prevalent in the industry and his client list includes such wineries as Beaulieu, Benziger, Diamond Creek, Merry Edwards, Niebaum-Coppola, Ridge, St. Francis, Sebastiani and Chappellet. Apparently, there are many more clients that wish to remain anonymous.

Some of those skeptical of his ideas include Randall Graham, Joel Peterson and Roger Boulton, a professor of oenology at U.C. Davis.

The article, though certainly brief, is a good read and tracks the history of Enologix and its proprietor in the industry. It is, IMO, more factual than critical but tries to be somewhat even-handed.

Worth a read, I’d say.

Best, Jim

[Brad Ballinger]

Enologix was on the screen for a few minutes in the Mondovino documentary. Robert Mondavi Winery was a client until 2001, presumably when they opted to use Michel Rolland's services instead.

As long as there is a consumer market for wines that achieve certain scores, there will be a client list for Enologix, Rolland, and others like them.

[Rebel Rose]

So, if people drink Enologix wines, does that make them Enologeex?

(Sorry, couldn't resist!)

[Daniel Rogov]

As I posted elsewhere: Sorry, but it all seems rather tautological to me. After all, what are they saying: (a) A wine that has certain characteristics will be popular with people who like those characteristics; (b) With appropriate manipulation wineries can "make" wines with those characteristics; © Some of the people who like those characteristics are critics and when they find those characteristics they will write well about and award high scores to those wines.

I'm also afraid that I agree with Professor Boulton in that it all sounds just a bit too much like alchemy and not quite enough like either winemaking or verifiable chemistry. I am also growing increasingly weary of people who buy wines primarily on the basis of scores. On a more positive note, I honestly believe that an increasing number of people who read the critics are learning to perceive scores as nothing more than a "shorthand quality summary statement" and then turning to read the notes to see if they really might enjoy the wine.

Those who subscribe to the NYTimes on line can find the article in question at

http://www.nytimes.com/2005/08....html?

[JohnL]

Daniel,

I agree with you. Many people do "move beyond the scores."

And tautology aside--there is a lot of "voodoo" if you will involved in wine making. Science has not been totally successful in quantifying what makes wine what it is and the element of "magic" is part of what makes wine so attractive.

As for the scores, all the hoo ha has somewhat mystified me.

Wine has been "rated" for a long long time. This is not some new phenomena. The 1855 classification of Bordeaux created a rating system--one based upon--consumer taste --sort of a focus group, if you will. That is wines classified, in part, based upon the prices people were willing to pay.

A network of "middlemen" importers etc offered "advice" to their customers as to what to buy etc. The trade was firmly in control.

Later wine writers emerged who made pronouncements and offered their opinions to customers/the public. The trade influenced many of these writers and indeed, many of these writers were part of the trade. Both offering advice and participating in the promotion and sale of wine to the public. a wine's quality, tasting notes alone are, in essence, the act of rating a wine--indirectly maybe, but still reflecting a wine's perceived quality. The attempt to quanitfy what one tastes and believes and to communicate that to others has led to an attempt to be "scientific" to quantify.

Whether it is verbage, stars, the Davis scale or the 100 point scale wine is and has been rated in one way or another forever.

I believe that much of the "debate" over critics, writers, rating systems is created and carried on by the wine trade itself.

It comes in large part from, from people who fear losing influence, control, if you will, and other petty feuds and silliness among winemakers, exporters, importers, wholesalers, retailers, writers and critics etc. There is a lot of hypocrisy, jealousy and petty sniping. The world of wine (our world in general) has seen a seachange over the last forty years or so and the entire wine business has been wrestling over how to deal with this change. Wine is truely global and it is no wonder that winemakers who produced wine and sold it locally for local consumption are "sweating" over the fact that there are more and more wines from more and more places around the world on local shelves competing for the consumers they had a monopoly on for centuries--it is no wonder there is sniping and character assasination at play.

As for the enologix thing.

Science has been a big part of this sea change. Peynaud leads to Rolland "gurus" emerge. remember Guy Accad in Burgundy? How about Turley? UC Davis, the "anti UC Davis gurus etc etc.

I would ask is "Biodynamics" any more or less voodoo than what enologix deals in?

Is it any more or less valid?

Look at the whole debate over terroir.

In the end though time marches on and progress is what it is. I think this is all fun to watch and this is a great time for all consumers of wine. If some guru's advice results in a fine distinctive wine that's great, if bodynamics results in a delicious Loire Sauvignon Blanc or a Leroy Burgundy--that's great. If we have fewer wines tainted by hail or other weather problems as a result of some biologist or a shaman that's

great. If there are fewer weedy cabs and merlots because of UC Davis research indicating where certain varietals should be planted--great! And if enologix' mix of science and who knows what results in better finer wines then that's a good thing.

All I can say--from my vantage point is, today is a lot better for a consumer of wine than yesterday!

[Florida Jim]

John,

Your points are well taken.

Regarding scoring; I do have difficulty with points or any attempt to quantify quality. That may work very well if the only person involved in the tasting and scoring is you, but I do not believe it translates beyond that.

Whereas, using the language to describe a wine is a more descriptive, more subjective and does not appear to be an attempt at precision. If you will, it qualifies quality rather than quantifies it.

Perhaps, that is hair-splitting, but I am much more comfortable with verbal opinion than I am with numbers, stars or puffs.

Regarding Enologix; I guess it boils down to the use of the term 'better' in the concept 'better wines.' Better to whom? IMO, once a wine is determined to be sound (unflawed) then there is never a time when one wine is better than another in the universal sense. For you it may be, for me it may be, but for everyone, that's nonsense. When it comes to quality, the one universal truth I can hang my hat on is that the best wine in the world is the wine you like best.

Of course, much of this is likely taking this all a bit too serious, but here we are, doing just that.

Best, Jim

Edited August 8, 2005 by Florida Jim (log)

[Craig Camp]

Tom Wark's impassioned comments on the Enologix article that appeared in last Sunday's New York Times Magazine (NY Times link) on his "Fermentations" Blog are well worth a read as is the interesting article by David Darlington in the Times discussed above.

Here is the link to Tom Wark's Fermentations

Edited August 10, 2005 by Craig Camp (log)