How does one select wines with value?

Lots has been made recently of young vintages being released at more expensive prices than a similarly scoring – but more mature – vintage.

You hear a variation on the below often..

“Chateau Plonk just released their 2023s at £250 / bottle. They only got a 92 from my favourite critic – Jacques Hyperbolé. I could buy a bottle of the 2014 for £160, and Hyperbolé gave that a 96!

So how do you work out the best value wines? For every vintage of every label of every producer you compare the relative price points, maturity, critic scores, vintage quality and a number of other factors.

You then pick the wines that seem to be undervalued based on your comparison to other vintages, other labels, other wines etc.

By the time you’ve done this for the last 10 years in Burgundy – they probably will have released the newest set of wines.

By the time you’ve identified the undervalued wines, the prices have probably changed.

This is the point at which the discerning wine investor must gracefully secede to a computer. One thing that computers are much better than humans at is computing lots of values very quickly.

Logic at Scale

Our model does this better than a person can for three reasons.

Firstly, as I said above – it can work out which ones are undervalued by which metric much faster than you can.

When I say much faster I am comparing hours (the computer), to years (the humble wine connoisseur).

Secondly, because we have historic price data – we can actually backtest which of the comparisons actually affects the future value of the wine the most. Not only is the model faster, but it is more accurate.

Thirdly, we can build in variables accounting for the investment characteristics of the wine. We can not only identify undervalued wines, but we can identify which of the undervalued wines shows the most potential for future appreciation.

Have you ever thought “I wonder whether the price of second wines from highly regarded producers acts differently to the first wines…”? or “Do wines from off vintages appreciate at different stages in their lifecycle to wines from good vintages…”?

We have. And we’ve tested it. And it’s factored it into the model.

The Human Touch

As with most deployments of AI that I have come across – this process works best when you combine human expertise.

AI can provide the logic at scale, but ultimately, an experienced eye must interpret the data, apply market context, and make the final decision.

The question is no longer “How do I find the best value wine?”—it’s “How do I best combine technology and experience to maximise returns?”