Behind WineStein

This world has countless numbers of people loving good food. And most of them are no wine-experts, nor do they have the time or desire to become one. They just wanna drink good and well-matching stuff.

For the good stuff, they rely on the many information sources available. But which wine matches best with what you set your mind on to have for dinner tonight? This enters the field of expertise of the sommelier. Good restaurants have one. At home, the odds are that you don’t. And that’s where WineStein steps in. Basically, the idea behind WineStein was simple: you enter what you’re having for dinner (or lunch), and WineStein tells you the good matches.

In order to make this work, you can try relying on rules. But… there’s a lot of rules to be taken into account here. Ask a good sommelier what (s)he’s thinking when considering the best match for a specific dish. We did that, and it didn’t make us happy. Frankly, there are too many rules and exceptions to ever make a rule based system really work, unless you want to settle for very basic pairing advice.

So, what then? Luckily, modern science has also taken up the challenge of getting to terms with a lot of highly complicated decision making. It’s the world of data mining, neural networks and other types of advanced math that – when done properly – will succeed in ‘capturing’ the way an expert person reasons. And that’s what we opted for in developing WineStein.
Not that things got suddenly very easy from there. It took us five years of research and development to get where we are now. But now that we have found the trick, we have time at hand: since software of this type is self-learning, our ‘kid’ will only get better.

To give you an idea of what goes on behind your App: there’s a dazzling amount of number crunching in WineStein. It calculates a big range of taste and aromatic aspects of both wine and dish. Basically, we convert everything that matters in a match into numbers. How exactly is the chef’s secret of course, and like good cooks, we’ll never tell. However, we can tell you that once WineStein has done the ‘basic’ math on what you tell it you want to eat, it will start a search among the over 50,000 examples of wine&food pairings (and mispairings!) that is was trained with. And after having made its way through over 200 million (!!) possible permutations of stuff that potentially matters in wine&food pairings, it comes up with results.
So, in case you’re wondering why our App takes a few seconds now and then, this is why. We’re doing our best to make it as fast as possible, but 200 million is a decent lot, even for the serious computer system that WineStein runs on.

Impressed by what we pulled off? Thanks. But what matters most to us is that you like what we built. We at WineStein / winewinewine.com all love good food and good wine. And we want as many people possible to enjoy how much better both will get when properly paired.

Oh, and in case you’d run into anything that doesn’t make sense: do feedback us (it’s in your App). We use it to make WineStein even better. Thanks for that in advance!

You can find a YouTube demo here.

Our iPhone App is available here