The time has come. The Bordeaux 2022 En Primeur has come to an end and, as for every vintage, we run a comparison between the Saturnalia scores and judgements by the critics.

In case you have never heard of the Saturnalia score, it is calculated automatically using an artificial intelligence model after the harvest. It leverages the data collected from satellites monitoring single vineyards and combined with weather data. The model is trained with scores from previous vintages from wine critics.

First, let’s compare them with the average of scores published by the critics. The chart (Figure 1) shows wines ordered by deviation between the scores. Saturnalia shows a habit of underestimation of the scores.

Figure 1. Saturnalia scores and average of critics scores.

The histogram helps in understanding even better. Each bin counts the number of wines within a certain deviation. On a total of 98 comparable wines, 95 (96.9%) have a deviation from -2.5 to 1. Moreover, 72 out 95 (75.8%) have a deviation lower than 0.

Figure 2. Histogram of differences between Saturnalia score and average of critics scores

Let’s now include all the critics scores. The Saturnalia score stayed within the window of scores by the panel of critics 81 times (82.6%).

Figure 3. Comparison with all the scores available.

The chart below helps to understand the variability of the critics. Quite interestingly, in most cases, they disagree among each other on a range from 3 to 5 points. 

Figure 4. Range of scores assigned by critics

In addition, we have calculated a correlation matrix, reported below. For example, we can highlight that the Saturnalia score correlates better with Jeff Leve, Jean-Marc Quarin, Lisa Perotti-Brown and Peter Moser. Lowest correlation with Lin Liu, 70%. 

Surprisingly, JA has correlations below 80% with all the other critics and the lowest correlation (56%) is between Antonio Galloni and Lin Liu.

Figure 5. Correlation matrix

Here some scatter plots to compare single distributions

Figure 6. Saturnalia vs Neal Martin
Figure 7. Saturnalia vs Peter Moser
Figure 8. Saturnalia vs Jean-Marc Quarin

In conclusion, the analysis showed once again the good confidence of the Saturnalia scores – which I remind once again, are calculated and distributed in November. As for most of previous vintages, the model confirmed its trend to underestimate the scores. This can be explained by the unique features of the 2022 vintage, so hot and dry that no one – not even our model – expected scores so high. Anyway, the model will benefit once again from these extreme conditions as it will learn and improve.

If you want to benefit from using the Saturnalia scores, contact us!


© 2021 Ticinum Aerospace. All rights reserved

En Primeur 2022 – Comparison of Saturnalia scores with critics