(Updated 2/19/17: fixed broken links, converted charts to a Tableau Public story, added a new section about a Harvard University science & cooking lecture from 2012 that included similar plots)
One of my early theories about why the cookies were so good was that they simply had a lot more chocolate than the average cookie. That line of thinking reminded me of Meg’s brilliant post about her “mean” chocolate chip cookie recipe, where “mean” had two meanings: slangy (excellent) and statistical (the average value of a collection of numbers).
To make her mean cookie, Meg took twelve distinct cookie recipes (but surprisingly not the classic Toll House recipe), entered them into a spreadsheet, calculated the average quantity for each ingredient, and then used her baking judgment to determine an average set of mixing procedures. The result:
These cookies were pretty damn good! I’d expected the worst. I’d expected they’d be inedible, or burnt, or floury and gooey at the same time. I had a hint they might not be too bad when I tasted the dough. But when I pulled them from the oven, I was amazed. The first bite revealed a cookie crispy around the rim, warm and chewy on the inside. A few hours later, they were firmer, but still tasty. The best chocolate chip cookies ever? I’m not sure, but I baked A Mean Chocolate Chip Cookie. And that’s enough for me.
I took the spreadsheet (which Meg made available on her website as an Excel file) and did what comes naturally to me: make some charts. The first thing I did, however, was correct a small oversight that Meg made when she did not normalize the recipes to a fixed point, like the quantity of sugar or flour. Not normalizing means that the importance of an individual recipe in the final result depends on the size of the recipe: in other words, the weight of a recipe calling for 4 cups of flour and 3 eggs would have more influence on the final result than one with 1 1/3 cups of flour and 1 egg. Since these were all pure flour recipes (no oatmeal), I scaled the ingredients in each recipe so that they would have two cups of flour.Clearly, this is not an ideal normalization because not all cookies rely on the same amount of flour for their structure, but it seems like a reasonable approximation. Other normalizations, like the quantity of butter or amount of chocolate, could be used too, but as we’ll see below, I’m not sure it really matters.
In some provinces of Geekdonia, people have been know to chart first, ask questions later. As I have been thinking about this post, I’ve realized that I’m one of those people. There were many questions that I should have asked and answered before making the charts, like “How much do I like each cookie?”, “What style is each cookie (thin and crispy, thick and chewy, in-between)?” And so, alas, the charts that I carefully prepared are more like pieces of art than explanatory figures. The charts are presented as a Tableau Story; to move between them, click the rectangular descriptors at the top or the arrows. If you move your cursor over a symbol, a description of that data point will appear (including the recipe source).
The first chart shows the volume of chocolate chips in each recipe after the flour normalization. The names of the recipes on the vertical axis are taken directly from Meg’s Excel file (except for Toll House, which I added myself); I provide links to each one (when available) at the bottom of this post. We see quite a significant variation, from 1 cup all the way to almost 2.5 cups (1 cup = 8 fluid ounces = 236 mL). Part of the variation, of course, is that different cookies have different styles — crispy, cakey, etc.
The second chart shows how butter and chocolate quantities relate for the collection of recipes. The x-axis is butter (in sticks — 4 ounces, 1/2 cup, 113 grams), the y-axis is the volume of chocolate chips. If there was some universal ratio of butter to chocolate, the points would fall on a line, but they clearly do not. The ‘butter outlier’ to the left is “The Cookie Book” recipe; the two high chocolate recipes are Torres/Leite and David Lebovitz.
The third chart shows the relationship between chocolate chips and sugar (the sum of brown and white types). These points are even more scattered than the butter points, because 1) sugar probably plays less of a structural role than butter, and 2) recipe developers might be influenced by the ‘sticking’ of butter, i.e., picking one, 1 1/2 or two sticks for a recipe instead of 3.3 ounces or another uneven number. In this chart, the Torres/Leite recipe has the lowest amount of sugar. Combine that with a high quantity of dark chocolate and you get a seriously good cookie.
The fourth chart compares butter and sugar quantities for the recipes. The Steingarten recipe is farthest to the upper right with the most butter and most sugar.
(Here is a link to my chocolate chip cookie chart in my Tableau Public space if you want to view it there)
Although I’m sure that the food industry has people that make lots of charts exploring the make-up of chocolate chip cookies — attempting to understand how to improve flavor, reduce costs, and so on — for a hobbyist like me the charts turned out to be a somewhat futile exercise, resulting on only a reinforcing of the idea that cookie recipes can vary significantly and confirming my initial impression that the Torres/Leite cookies had more chocolate than average.
When a three-day rainstorm hits the Bay Area on a weekend this winter, perhaps I’ll try taking this exercise to the next level: baking each cookie and comparing my preference with their ingredients — a chart comparing my ranking with the butter to sugar ratio, for example. Like this exercise, it might not tell me anything new, but the experimental samples will be a worthwhile result.
The list of recipes used for the charts (for the recipes in the comments at Megnut, you’ll need to scroll down and/or do a search because the comments have lost their links).
- Alton Brown
- Robyn (in comments at Megnut)
- Cook’s Illustrated
- The Cookie Book (in comments at Megnut)
- Alice Medrich
- Kayjay (in comments at Megnut)
- Debbie at Parents Need to Eat Too
- Wes at gastronomyblog.com
- Sol (in comments at Megnut)
- David Lebovitz (in comments at Megnut), a recipe from his Great Book of Chocolate
- Toll House
- Megnut’s “Mean” Cookies
Phase Space of Recipes
For a few years, Harvard University hosted a series of public lectures about the science of food and cooking. The lectures covered a wide variety of topics — including emulsions, baking, chocolate, proteins, fermentation — and were delivered by some top talents — like Harold McGee, Ferran Adria, Grant Achatz, David Chang, Dominique Crenn. They are currently available on YouTube and from their podcasting service (look for Science and Cooking from Harvard University).
Each lecture is preceded by a short talk by one of the faculty members. In the introduction to the lecture called “Bakistry” (delivered by Joanne Chang from Flour in Boston), the speaker brought up the subject of “phase diagrams of recipes.” Or, in other words, plotting one ingredient against another, just like what I did above.
A group of Harvard students that included Elaine Angelino, Diana Cai, Madelaine Boyd, Shelby Lin, Julie Monrad, Naveen Sinha, Jessica Stanchfield, and Alec Dah-Wei Yeh were ambitious. Instead of looking at a dozen or so recipes, they wrote programs that downloaded around 40,000 recipes from AllRecipes.com, extracted the ingredients, and complied them into a useful format (did the team ever publish their results?).
Then they started plotting. The figure below, which is a screen grab from the YouTube version of the lecture, shows the flour-sugar “recipe space” for brownies (pink) and sugar cookies (blue), with the speaker’s favorite recipe marked with a star. Only a small region of phase space contains brownies, and only a small region contains sugar cookies. What explains this? Is it history or science? The speaker proposes two possibilities: 1) History set the basic recipe, and everyone following tinkers. 2) The “laws of physics are brutal”: all good brownie recipes to date fit into the phase space, and a brownie that doesn’t fit is not good to eat. I suspect it is #2: there is a lot of chemistry and physics in butter, sugar, flour and baking, and tripling the sugar in a brownie will turn it into something completely different (like fudge).
The discussion of bakery phase diagrams runs from 2:29 – 11:00 here: Bakistry, lecture by Joanne Chang (starting at 2:29) on YouTube.