Fun with data!

So, I’ve recently greatly expanded the types of information I track about my reviews and wanted to share with you all so you could maybe draw from my ideas to help find what you like about whisky, or rather, what trends there are in whiskies you find you really like. Or maybe you even have some suggestions for me to help me get more out of my scores. You can find screenshots of all of the mentioned tables of data here, though they are each included when they come up.

Initially I just tracked things like average and some other basic statistics and then expanded some to regional averages and a bit more, but didn’t get much more specific than that at first. But I’m a nerd and I love Excel so it obviously got crazier from there.

This table is that of the very broad statistics of averages, some other statistics like standard deviation, and then some regional/style averages, the respective counts of the scores included in those averages, and a High and Low. 

This section gives me a glimpse into the data as a whole, as well as some insight into how much I might like whisky from a certain region. And while today a whisky’s region is largely just a circumstantial bit of information about geographical location, there is still some meaning.

This table shows some distillery-specific data. 

This data tells me the averages by distillery for the bottles they released, the IB releases from that distillery, and the average of all things I’ve tried which that distillery produced (and the counts of each as well). I had some space, so I also made a formula to tell me the high and low of each distillery. The distillery averages are more useful to show how much I like a distillery relative to another and how distillery bottlings compare to IBs; the high/low is really just an interesting bit of information. This table shows information of various bottlers and how their casks/blends have performed for me.

This table just shows how many times I’ve given a score in a specific range to track how I might be leaning one way or another. 

Currently it’s pretty much a rightly skewed bell curve on the higher half of the scale, which is pretty much exactly what I’d expect given selection bias from both me choosing mostly whiskies I think I’ll like based on liking similar things and ones that seem to have performed well for others.

I recently realized I was scoring a bit high early on, and I eventually got curious about what my averages were over given ranges of scores over time, so this table has that data. 

This is a graph of a moving average of the first 10 scores, 2nd-11th, 3rd-12th, etc. (blue), and the total average up to that point (orange) with trendlines. 

The moving average trendline is interesting as I’d have expected it to be going down based on early scores being high, but kind of makes sense as I get more into non-entry level expressions warranting higher scores. 

Now for the more interesting stuff which gives some pretty valuable insight into what I like in a whisly. I got curious about whether I thought I liked peat a lot or whether I actually did, so I started a table to build averages based on how peated the whisky was. Rather than consider things like PPM which doesn’t offer much real meaning, I just used 4 levels of peat: None, Somewhat/Lightly, Normally, Excessive/Heavily (as well as an average of whether it was peated or not). Another part of this was me being curious about how much I liked things like Sherry influence and such so I added to that table various figures of maturation (all or bulk of time in that cask) and finishing (tail end of time in that cask): full/partial bourbon maturation, Sherry/wine/other maturation/finish, and a general note of some other type of cask information such as Laphroaig’s common use of quarter casks. In addition to tracking whether Sherry/wine casks were involved, I also tracked the type of Sherry where applicable. Not all figures have their own data (like cider casks or stout casks), but they could if/when those become more common. This table shows those averages, as well as the counts of the scores included in those averages and the percentage of scores which have that quality, the counts/percentages being there to determine how useful that data is at a given time. 

This was an interesting and worthwhile exercise as it taught me that I seem to favor wine maturation and should consider focusing on that when deciding what to purchase, as well as letting me know I don’t care for a Sherry finish nearly as much as a proper Sherry maturation. I’m looking forward to this data getting more meaningful as I build more data.

Finally, a recent review of Talisker 30 got me thinking. Anecdotally, and eventually backed up by my data, I like peat/smoke. This particular expression was far more fruity than smoky to me yet I loved it. I figured the extreme age of this is really the only thing different from the usual (or at least the most significant), so I realized I should try to track age as well just to see what it has to say. I decided on a few ranges of age statements and set up averages for them (as well as counts/percentages). This table shows these calculations and while the 30+ category is not yet useful with just 1 data point, it’s interesting that 5-10 years and 10-15 years did a fair bit better than things between 15-23 years old, offering some not-the-most-but-still-scientific evidence that older does not mean better and NAS whisky does not have to be trash.

I hope you found this interesting, or at least not a waste of your time. I’d love any suggestions to improve my data, and if anyone expresses interest, I’m plenty happy to talk about how these figures are calculated (mostly averageif(s) and countif(s)). Slainte!