Global Wine Medal Rating. Approach and maths

Getting started: The first look on the chessboard

As we have already mentioned in the introductory article about the Global Wine Medal Rating (GWMR), an attempt to create a wine rating based on the received awards implies solving a whole series of tasks, both organizational, analytical and mathematical.

From the perspective of collecting the information necessary to calculate the rating of any wine that has received at least one award at professional wine competitions, the Gustos.Life team has launched several parallel processes:

• Complete cataloging of information about all professional competitions in the world.
• Manual collection of information on medals received by wines from Moldova in order to quickly gather information necessary to calculate and set up the methodology with the framework of a specific country.
• Systematic and consistent “automated” data collection on all medals of each individual competition.
• Development of machine learning algorithms for structuring and systematizing the received information.

Having promptly received basic information about the Moldovan medalists, we began to formulate a mathematical model for calculating the GWMR.

As a result, the problem of calculating the aggregated medal rating was reduced to two large, but nevertheless solvable tasks:

• Calculation of the Rating of the significance of competitions.
• Calculation of the Rating of a particular vintage, depending on the number and value of its medals, taking into account the rating of the significance of competitions.

The basics of the rating of the Significance of the Competition: E2-E4

After a large-scale cataloging of wine competitions and consultations with experts who were repeatedly members of the jury of professional competitions, we came to the conclusion that the Significance of the competition depends on two main factors:

• Its scale (that is, the number of participating samples) — a parameter that shows how much winemakers and other participants trust this competition and strive to receive its awards.
• How much it is cited on the Internet — a parameter indicating the popularity of the competition and the resonance of receiving its awards.

Thus, we have derived 2 parameters, on the basis of which the index of significance of each individual competition would be calculated

• Average number of participating samples for the last 3 years.
• Google PageRank is a Google algorithm for ranking web pages based on the number of external links and mentions.

Competition significance index: Debut.

As a result, we calculate 3 parameters:

1.kNS (Number of Samples) is a trigonometric coefficient from 0.01 to 1, depending on the average number of samples.

As the graph shows, competitions with 5,000 samples and more will have values approaching the maximum, while competitions under 120 will receive the lowest possible scale factor.

2. kGPR (Google PageRank) — the citation index on the Internet will be linear in the range from 0.3 to 1 based on the Google PR indicator, which takes values from 0 to 10, although the maximum Google PR for wine competition sites is 7. The calculation formula is kGPR= (GPR + 3)/10, but not more than one.

3. kBtA (Bring to Adequate) — a coefficient from 0.1 to 2, which would equal 1 in 99% of competitions (that is, it would not affect the calculation) and would only sometimes be used for adjustment in exceptional cases when one of the above coefficients would not reflect the true popularity of the competition.

The kBtA parameter would change and thus affect the final rating only in exceptional cases. Examples:

• When the competition recently changed the Internet domain of the site and has not yet managed to acquire an adequate Google PR. Berliner, Asia and Portugal Wine Trophy are now in such a situation.
• When the competition deliberately limits the number of participating samples, for example, specialized single-variety competitions like Chardonnay du Monde or Pinot Masters.

The final calculation formula:

kNoT (Notoriety) = kNS * kGPR * kBtA

Below are some examples from the very top of the rating of competitions in which Moldovan wines participated, from its middle part and from the very end of the rating.

Vintage Rating: Strained Middlegame

Studying the possible Basic approach to calculating this rating, it would be based on 4 main concepts:

1. All medals are divided into 4 categories of value: Higher than gold, Gold, Silver, Everything that goes below.

2. The more medals in a certain category of value, the stronger the shift for all medals of this value from the minimum estimate of the interval towards the maximum.

A quick example. Let’s assume that the wine received only one medal — silver at the Concours Mondial de Bruxelles, where this value corresponds to an estimated range of 84–87 points. Since there is only one medal, its intermediate rating is at 1/5 of the path from the minimum rating of the interval to the maximum and is equal to 84.6.

Later, the wine received another silver medal at the Decanter World Wine Awards, where silver is awarded to samples that received between 90 and 95 points. Since there are already two silver medals, the rating for each of them is calculated as a 30% shift from the lower to the upper limit. Thus, the first medal increased in the rating from 84.6 to 84.9 points, and the second immediately received 91.5 points.

Then, a silver medal was added to the first two awards at the AWC Vienna, where it corresponds to the interval between 87 and 90 points. Since the number of silver medals has increased from 2 to 3, the rating for each of them is calculated as a 35% shift from the lower to the upper limit. Thus, the first medal increased in the rating from 84.9 to 85.5 points, the second from 91.5 to 91.75 and the third immediately received 88.05.

A similar logic applies to all four categories of the values of awards.

3. The final specific weight of each individual medal in the rating depends on the Rating of the Significance of the Competition.

Example: If a wine has 2 medals — one gold medal with an estimated rating of 90 points at an important competition with a significance coefficient = 1 and one bronze medal with an estimated rating of 80 points at a “parochial” competition with a significance coefficient = 0.2, then the specific weight of the gold medal will be 5 times higher and the bronze will only slightly lower the rating of the vintage below the “gold” level.

4. The primary rating is adjusted once again to encourage wines with a large number of medals received at prestigious competitions.

Vintage Rating Algorithm: A Confident Endgame

A few mathematical details. First of all, we calculate the rating of each separate medal:

Rm is the rating of each medal, taking into account the kNd (the correction coefficient of the evaluation bias within the evaluation intervals for each category of value).

This graph illustrates the growth of the shift that we described above in the example with one, two and three silver medals.

Then, we calculate the primary rating of the vintage, taking into account the rating (RM) and the coefficient of the significance of the competition (kNoT) of each of them:

pWMR (Primary medal rating) = Σ (Rm * kNoT) / Σ kNoT

As a finishing touch, we calculate the kFA (final correction factor) depending on the ΣkNoT (that is, the total significance of the competitions at which medals were received).

As the graph shows:

• The award of already 2 medals from the largest competitions (or 3–4 from average competitions) guarantees that the final correction will almost not affect the primary rating,
• A large number of medals will change the usual rating by less than 3.5%,
• While the presence of very insignificant medals will reduce the rating by no more than 5%.

GWMR (final vintage rating) = pWMR * kFA

Let us talk about it: All moves are recorded

In the near future, Gustos.Life will start publishing various sections of the received wine ratings. Ranking the winners in order and distributing the wines on pedestals is very likely to raise a series of questions. We are sure there’d be many who would have intuitively or objectively seen the list of winners of certain categories in a completely different order and/or composition.

In this connection, we strongly urge all those interested to carefully study our mathematical model and, should there be any observations, clarifications and comments, be sure to inform us about them.

We are always open to a constructive dialogue and improvement of our mathematical model in the future. We just remind you that you should pay attention to the level of principles, formulas and coefficients, and not directly to the ratings, which are only dry results of algorithm processing.

What is it all for?

This is what we shall write about in the next article, where we’ll show how the information we’d collected, as well as the resulting rating can be used and namely:

1. How a consumer can choose the best wine in the category he is interested in, especially in terms of price/quality ratio.

2. How winemakers choose the optimal strategy for participating in wine competitions based on the characteristics of each competition, the rating of competitions and the line of wines of a winemaker.

3. How wine sellers can use the information we’ve collected to increase their sales.

Authors:

• Andrei Radul (Linkedin profile)
• Veaceslav Cunev (Linkedin profile)

About the project Gustos.life:

The Gustos.life project specializes in providing information services for the wine industry within the created integrated ecosystem.

Main Services:
1. Service for holding professional wine competitions — https://winetaste.gustos.life/

2. Digital wine storage — https://dws.gustos.life/