Thursday, December 31, 2009

Did you win that Gold Medal by Chance?

When I visited Gloria Ferrer Winery in Sonoma, I was duly impressed by this display of gold and silver medals some of the wines had been awarded. This was until I read two articles in the Journal of Wine Economics by Robert T Hodgson. Nothing against Gloria Ferrer as they make some great wines but it turns out that winning at these competitions may strictly be a matter of luck and not so much a statement about the quality if the wines.

Statisticians are putting the world of wine on edge. In "An Analysis of the Concordance Among 13 U.S. Wine Competitions", Robert T Hodgson analyzed over 4000 wines entering 13 competitions across the U.S. What he found was that a wine winning a Gold Medal in one competition seldom won a second Gold Medal. You would think that a truly excellent wine could win repeated gold medals. He found that when wines did win repeated Gold Medals the frequency that this occurred at was no better than that of random chance according to a binomial distribution with a probability of p=0.09. For my statistics students, you may recall that binomial distribution is based on a process called a Bernoulli trial in which outcomes of repeated trials are wither win or lose. These repeated trials are independent and the probability remains constant from one trial to the next. A good example is a coin toss with heads or tails as the outcome. The inference from Mr Hodgson's work is the best way to win these competitions is simply enter as many of them as you can. Random chance will dictate that you will win some of them.

There may be an oversimplification here. Obviously some wines are better than other wines and probably do deserve recognition. Wineries probably choose to enter their best wines into the competition. Since all the wines are relatively good, then judging is the primary concern. This is the other area of research by Mr. Hodgson. In "An Examination of Judge Reliability at a Major U.S Wine Competition", Mr. Hodgson concluded that only about 10 to 20 percent of panel judges could replicate there scores on repeated tastings of the same wine. In total using Analysis of Variance techniques Hodgson showed that only about 50% of judges were influenced by the wine quality alone in their scoring. From a practical standpoint the other 50% of judges were not consistent and simply did not measure things the same.

The take away message from these articles is that judging of taste in food competitions or wine competitions may not be based on quality alone. Other factors are at play which may make the whole process random.


"An Analysis of the Concordance among 13 U.S. Wine Competitions", Journal of Wine Economics, Vol. 4, No. 1, Spring 2009, pages 1-9

"An Examination of Judge Reliability at a major U.S. Wine Competition", Journal of Wine Economics, Vol. 3, No.2, Fall 2008, pages 105-113

Monday, December 28, 2009

Is Expensive Wine really Better?

When you visit Napa and Sonoma you may observe a certain level of arrogance, prestige and quality at different wineries. Even within the same winery they may offer different levels of tasting experiences depending on if you are interested in the basic offerings or the reserve offerings. Wineries are trying to differentiate their brands and aim them towards consumers tastes and preferences. They want you to think that your personal image is associated with a particular brand and price level.

Answering the question do more expensive wines taste better, Robin Goldstein and Alex Herschkowitsch have published, THE WINE TRIALS 2010. The book is supported by an large academic research survey of over 6000 participants, details are published in "The Journal of Wine Economics" article entitled, Do More Expensive Wines Taste Better, Evidence from a large sample of Blind Tastings.

In the research study the participants were subjected to blind taste tests where an expensive and inexpensive wines were rated without prior knowledge to label, price or quality. The findings indicated that without prior knowledge to price, the participants rated less expensive wines higher than more expensive wines with statistically significant results. It was interesting that when a $150 Dom Peringon Champagne was compared against $12 Domaine Ste. Michelle Sparkling Wine from Washington State, participants favored Domaine Ste. Michelle by about 2 to 1.

What does this mean for the everyday consumer? In the book, Robin Goldstein encourages readers to conduct their own blind tastings. Decide for yourself what you like and don't like. The book presents reviews of 150 wines selling for less than $15 and decided on winners in this category.

If you subscribe to Wine Spectator or Wine Advocate, you may be influenced by the 100 point ratings they assign to wines. The authors suggest biases exist in the Wine Spectator or Wine Advocate rating systems favoring more expensive wines. These biases could be the result of an acquired taste or "perfect palate" of the expert wine drinkers. The majority of consumers are just not trained taste testers and probably cannot differentiate between black currants, blueberries, blackberries, cloves or other fragrances and tastes associated with wine.

Bottom line take away from this research study - Just because something costs more does not mean you will enjoy it more.

You may find more information at Robin Goldstein's website

Tuesday, December 15, 2009

Chance Encounters - What are the odds of that?

Our world is filled with systems, patterns and biases. The laws of physics and the universe demand a certain amount of order in nature and our surroundings. Yet despite this we often encounter random or sometimes unexpected events. At a distant airport while traveling on business you bump into a colleague you haven’t seen in 10 years. On vacation while waiting in line to enter the Magic Kingdom on the 4th of July in Orlando you see an old acquaintance. These are coincidences that seem to be unlikely random events. Just how unlikely are these encounters.

One view of these encounters is to look at the probability on any given day that any two people in a population will meet assuming they are independent events.

Let’s take Disney for example. According to the 2008 Attraction Attendance report by TEA/ERA about 17,063,000 people visited the Magic Kingdom in Orlando for 2008. This means on the average day, the park sees about 47468 visitors. The probability of any given individual being in Orlando might be 47468 park visitors divided by the US population 304059000 which equals 0.000157. This is one chance in 6368. If everything was considered independent we would say that the probability of Bill and his old acquaintance being in Orlando at the same time would be:

P=0.000157*0.000157 which is 2.64E-8 ( This is a very small probability number).

The odds might be expressed as one chance in 40.5 million. Wow this seems extremely low, however, this is not how the problem is solved.

Assume Bill is a moderately connected networker that can keep track of 500 contacts between family, friends, work associates and old classmates. Now we have to know a little about Bayesian probability and look at the probability that given a person is already at the Magic Kingdom, what are the odds an acquaintance is there. For this we have to look at the probability that Joe or Greg or Marge or Megan are there. There is literally a pool of possible candidates that may be at the park the same day of your visit. We add the probabilities together.

P=0.000157+0.000157+….. for all 500 contacts. This now gives a probability of 0.078 or about 1 chance in 13 that one of the acquaintances are at Disney on the very same day. Factor in the 4th of July and 20% additional attendance and the probability changes to 1 chance in 10. Since most people enter in the morning through the same set of park gates, there is good likelihood that you may be there at the same time as an acquaintance.

Airports, Malls, Amusement parks are large hubs where people gather and often at specific times based on schedules or seasons. It is not all that uncommon to hear stories of two people meeting that have not seen each other in years. These types of events although random may not be as unlikely or coincidental as we think.