Crowdsourcing blind tasting

More fun with stats

Not only can you hold a wine-centered Stanford student event without actually purchasing wine, but sometimes it's more fun that way. We had a variety of reasons for structuring our kickoff party the way we did, but probably the most lucrative was the fact that it took control away from the three of us dictators who have been determining the wine menu for the last month and allowed the crowd to bring the wines. Each person (or small group) brought a bottle of wine to share, with the only restriction being that the price fall somewhere between $5 and $20. Check out the night's results for a wide selection of affordable wines that a lot of us thought were quite lovely.

The wines were numbered and wrapped and guests were free to partake of whichever ones they wanted, and also urged to fill out simple evaluation forms that included room for describing the wine, ranking it on a scale of 1-5, and guessing what it was. The fact that rating was optional and the ordering of wines was random meant that even the number of ratings received by each wine ended up varying. So our results are far from rigorous, but that doesn't stop this correspondent from trying to identify trends anyway.

Some interesting results:

  • 17 wines were contributed: 9 reds and 8 whites. (Among the whites were 1 sparkling and 1 dessert wine, and both red and white included a good spread of varietals). Not only was it fortunate, but it was simply a remarkable distribution to have occurred by chance. This is epecially so, in light of my personal observation that friends who state a preference always prefer red wine, and, more substantively, the fact that 77% of our club's members who answered the question on their SWS membership forms listed a red wine as their favorite wine.
  • Most of the white wines appeared at the beginning of the night; it took the reds a while to catch up. I would very much like to believe that this is somehow results from our common training to start the evening with a white wine and finish with red.
  • A total of 44 reviews were submitted, distributed unevenly among the wines. The median number of reviews submitted per wine was 2.5, but two wines were not reviewed at all, and one received 7 reviews. The consumption of wine was also very uneven. Given that most guests had no clue what was in each bottle before it was poured (at least, this was true for the reds, which I poured), I think it's safe to assume that most of this variability was random. However...
  • The wine that was rated highest (4.6 on a scale of 1-5) also happened to be the wine that received the most reviews (7). The two wines that scored lowest (2.0) were among those that had received very few reviews (1 or 2). I find it very likely that people who liked a wine subsequently talked about it quite a bit (well, okay, I heard the talk and I engaged in some myself), which stimulated a greater number of people to drink the better-liked wines and to review them. (I should also disclose that, toward the end of the night, I gave in to the people asking for recommendations and pointed out to them the wines that had been most popular so far.) I find it equally plausible that the people who heard wine being talked about were convinced it was good and therefore rated it highly. Regardless of causality, the data are hinting at a linear relationship between number of reviews and the mean score given to a wine... ish. It has an adjusted R-squared of 0.1784 and a p-value of 0.06568.
    plot of mean rating v. number of reviews
  • If you paid any attention to our "Does Price Matter" tasting a couple weeks ago, where we found some indication that our members rank more expensive wines more highly, you may be curious to hear whether the same was true this week. I can say wholeheartedly that there was not, no matter how hard you squint. Nor am I surprised; I chalk it up to the social (decidedly not studious) atmosphere, to the irregularity in the review submissions, and to the confusing array of different wine types to compare this time around.
    plot of mean rating v. price

Well, take a look yourself. You can find the fully-tabulated results, complete with wine names and the crowd's descriptors, here. Prices were my own estimates based on