The Church of Kold Draft / The Cult of Hard Shake

[Kohai]

I assume many people have read these experiments by Eben Klemm, Alex Day and David Arnold? They were subject of one of the seminars at this year's Tales:

http://cookingissues.wordpress.com/2009/07...nce-of-shaking/

http://cookingissues.wordpress.com/2009/07...-of-shaking-ii/

These experiments, which appear to be fairly rigorous, seem to show that neither the type of ice nor the style of shaking make much difference in temperature or dilution of cocktails. This defies common sense, to my mind, but the studies appear to be well-executed and controlled.

A New York Times article recently summarized the studies (bottom of page 1):

http://www.nytimes.com/2009/08/26/dining/2...ardshake&st=cse

This challenges a lot of things I believed to be true. What do you make of this?

<edited to fix link>

Edited August 30, 2009 by Kohai (log)

[vice]

The last installment is also important here, as they delve into more subjective measures of shaking technique that do yield some qualitative differences.

http://cookingissues.wordpress.com/2009/08...ence-pegu-club/

[Kohai]

Brilliant, it seems I missed that one.

[slkinsey]

This is all really interesting stuff, and a great first step. But I'm not sure I'm willing to all any of it definitive, and I do wonder about some of these results and their real-world applicability.

For example, they compared similar weights of Kold-Draft and shell ice in the experiments. But who says that bartenders use similar weights of Kold-Draft and shell ice? They don't. In my experience, the bartender scoops the small end of the shaker full of ice and goes to town. This means that the shell ice shaker might have a larger amount of ice in it compared to the Kold Draft shaker, which might lead to faster dilution. I would have liked to see some experiments done comparing shakers-full of shell ice and Kold-Draft ice to see how they compared in this more real-world situation.

Unless it's a Ramos Fizz, people also don't generally shake their drinks more than 10 to 20 seconds. So, while the data from after 15 seconds is interesting, I'm not sure that we should be extrapolating those results back into the real world timeframe of shaking.

I also wonder about the sample size in some of these experiments. In particular, the dilution chart looks funny to me. When I see, for example, that the amount of ice melted in the shell ice drink at 9.5 seconds increased sharply compared to 8 seconds, but that the %ABV of the drink also had a sharp increase from 8 seconds to 9.5 seconds -- well, something just isn't right there. You can't melt more water into the booze and get a higher %ABV. Unless this was done with a very large sample size, I have to believe there was some kind of measurement error there.

Unfortunately, dilution in real-world conditions is really the only variable that's all that interesting to me. And, in particular, the relationship of dilution to shaking time. I don't have any trouble understanding that shaking a drink for 40 seconds tends to end up at right around the same temperature and %ABV for equal weights of ice. I've never been one who believed that Kold-Draft was somehow magically colder or resulted in lower dilution no matter how long the drink was shaken. What I do believe, however, is that in real-world conditions where the shaker is simply filled with ice (which may be important and may not be -- but those tests need to be done) and shooting for somewhere around 20-25% dilution, larger pieces of ice allow the bartender to shake the cocktail for a longer period of time before the target %ABV is reached. This longer shaking period (i) gives the bartender better control over the target %ABV, and also allows for more aeration of the drink due to the longer shaking time. I was spending a lot of time in all the first NYC bars that went over to using Kold-Draft, and rather than talking about how the drinks were colder and whatnot, most of the bartenders talked about how they could now shake the drinks a lot longer without watering them down.

The relationship of dilution to shaking time seems significant to me. And I think it's interesting to note a little quotation from the Pegu Club session: "From tests we have run earlier, we suspect that there are certain drops in ABV that register more than others. The difference in taste between a 21 and a 20% ABV drink isn’t as much as the difference between a 20 and a 19% ABV drink." It is important to keep this in mind looking at the earlier dilution chart, because those experiments were all done with 100 ml (about 3.4 ounces) of 40% ABV liquid. Very few shaken cocktails will start out with a %ABV anywhere near this high. Let's say, for example, that you're going to be making a Daiquiri with 2 ounces of 40% ABV rum, a half-ounce of simple and a half-ounce of lime juice. Before you start shaking the drink, it's already only at around 27% ABV. Assuming that 20% is a boundary condition below which the cocktail starts to taste watered-down, that doesn't leave very much time to chill and aerate the drink before you melt in an ounce of water and go below 20% ABV. This reality more accurately reflects what I was seeing around the time of the transition to Kold-Draft. At pre-KD Flatiron Lounge, for example, the bartenders had to give the drink perhaps 3-5 seconds of shaking and then go quickly to the glass, or the drinks would end up watered down. At post-KD Flatiron Lounge, they were shaking the drinks 10+ seconds, getting better aeration and less danger of watering the drinks. There has to be a reason for this.

None of the foregoing should be taken as a harsh criticism or invalidation of the excellent work done in these sessions. It's important to keep in mind, however, that these are preliminary experiments rather than definitive results. And, as a result, they have the same effect that first experiments always have: they raise more questions than they answer. This is the way these things always work. I don't think that Dave, Eben, et al. would suggest that they have definitively closed the door on these questions.

[Kent Wang]

The experiments are very interesting. Excellent post as well from Sam.

Another matter which should be obvious is making a highball and filling it with KD cubes. This will obviously dilute much slower than with smaller ice.

[Kohai]

I wonder if experiments on ice should be kept separate from experiments on shaking styles. Could combining the two cause some kind of interference or noise in the results? Would it be more insightful, at least initially, to simply add ice to alcohol and watch the natural, unassisted temperature and volume change? Hmm...

[vice]

I wonder if experiments on ice should be kept separate from experiments on shaking styles. Could combining the two cause some kind of interference or noise in the results?

Whenever one considers the effects of more than one factor, there is the potential for interactions between them. Off the top of my head, I can think of four factors that are addressed by the Cooking Issues experiments and Sam's discussion:

- ice type

- ice quantity

- shaking style

- shaking duration

If all these factors had only two 'levels' (which they don't; shaking duration, for example, is a continuous variable), there would be a total of 16 unique combinations to test. So, you can see how quickly such experiments can grow in size and complexity. When you consider that there are multiple response variables of interest as well (temperature, dilution, texture, etc), and that the treatment combination yielding the optimum for one may not do so for others, it can get still more complex. Indeed, it's enough to make one want a drink...

[slkinsey]

Ideally, I suppose you'd want to develop some kind of standardized shaking machine to separate out that potential variable. Perhaps a tweaked paint can shaker?

Then you'd want some way to quickly and precisely dump the drink and separate the ice from the liquid. I would think that some kind of hinged bottom on the shaker that would immediately dump the entire shaker contents into a large (chilled) sieve of the appropriate porosity (you would have to decide whether to let through ice chips or not) that would separate the liquid from the ice more or less instantly.

I think you'd also want to start with a mixture at no more than about 33% ABV, and you'd want to do some perceptual experiments beforehand to determine boundary conditions for "watered down" %ABV.

At this point, I would think you could fairly quickly run up very large sample sizes testing different weights and sizes of ice. I'd be interested to know what the difference in weight is between a full tin of shell ice and a full tin of Kold-Draft ice. I'd also be interested to scale the experiment all the way down to pellet ice and all the way up to fist-sized lumps of block ice, not to mention different kinds of shell ice.