Stigler Diet | Hacker News Reader

Stigler Diet

en.wikipedia.org

111 pointsvinchuco68 comments

Here's a story of a guy who used the same method to come up with the optimal diet for himself: https://www.gwern.net/docs/statistics/decision/1990-dantzig....

The first try he ended up with a diet that required 500 gallons of vinegar a day. There was a mistake in the data that give vinegar 0 weight.

The second try ended up with 200 bouillon cubes per day. He didn't put a limit on salt content.

Then it gave him 2 pounds of bran per day. And when he put a limit on that, it changed to two pounds of black molasses. His wife put a stop to it after that.

Optimization is very dangerous when applied in large quantities. The vast majority of good solutions of any high dimensional problem will end up in weird edge cases.

Clearly he was not optimizing for overall individual welfare - there are strong diminishing returns for any single ingredient.

Culinary and psychological satisfaction is a valid target for optimization, otherwise the individual will drop the diet and endanger the quality of nutrition, given a limited budget. Or he might outright commit suicide after the second week, not healthy at all.

It's even better than that: "At this point Anne got tired of the whole game. Speaking firmly so that I would know who was boss, she said, "I have been studying the various menus the computer has been generating. There are some good ideas there that I can use. I'll put you on MY diet." She did and I lost 22 pounds."

And for those who don't know anything else about George Dantzig, the author of that article:

There's an urban legend about a college student who shows up late for a math class, sees some problems written on the board, assumes they're homework, and solves them over the weekend, but it turns out the professor wrote the problems on the board as examples of open problems in the field.

This urban legend is true, and George Dantzig was the student.

https://www.snopes.com/college/homework/unsolvable.asp

It's funny - the author isn't just some guy, it's George Dantzig. Relevant because he is one of the fathers of linear programming. He and his simplex algo are actually referred to in the wiki page

Optimization is not dangerous, it's just that coming up with the correct model is hard.

No model is perfectly correct. That's the problem. Any model that approaches perfection would be unimaginably complicated.

Ironically, this diet isn't too far off the diet I've formulated for optimal nutrition, which is based on substantial literature review.

In terms of optimal nutrition, the core of the diet I've come upon is:

* Lentils/split peas - the core of the diet, high in fermentable fiber, protein, vitamins and minerals

* Yogurt - inexpensive source of high quality protein, vitamin d and calcium, the basis for smoothies

* Sauerkraut - cheap, filling, acts as a probiotic which when paired with lentils ensures a lactobacillus dominant microbiome.

* Eggs - cheap source of high quality protein, omega-6 essential fatty acids, vitamin d & e.

* Seeds - mostly chia and hemp, but sunflower and pumpkin are really good as well. These satisfy your omega-3 and omega-6 essential fatty acid requirements and are loaded with minerals. If you skip out on spinach and buy cheap eggs with low vitamin E content then sunflower seeds are excellent.

I also eat a lot of berries (frozen, in smoothies), whey protein and lean pork/chicken, though these are less important in my opinion.

An additional health hack that I highly recommend is to consume a lot of vinegar. The putative health benefits of alcohol are almost certainly due to its metabolism into acetate. By consuming vinegar directly, you skip the production of acetaldehyde (which has been linked to DNA damage). My favorite way to consume vinegar is by pickling vegetables in a vinegar solution, then using them as garnishes to add flavor to meals.

While I understand many people don't put a lot of stock in anecdotal reports, this diet paired with a bodybuilder lifestyle keeps me around 220lbs @ ~12% bodyfat which is quite exceptional considering I'm 6' tall, and naturally obese if I don't watch my diet.

Very cool. I just did a similar exercise also based on a lot of research (posted it on this thread). Interested in your thoughts.

Becoming 'fat adapted' reduces this cost significantly as you can get approximately 120 fat calories in a tablespoon of high quality coconut or olive oil for about 10 cents in bulk based on my last trip to costco, and as long as you supplement that with an inexpensive protein source you can get by on a very cheap diet. I splurge on high quality veggies simply because we don't know enough about phytonutrients' effect on health to rule them out. This in combination with skipping meals (intermittent fasting) causes the food bill to drop significantly and has some potential health benefits.

If you were to blend the benefits of services like Soylent (nutritionally balanced meals) and BlueApron (delicious delivered recipes), you arrive at a service where you pick from a list of healthy ingredients and the service uses linear programming to build "solutions" aka recipes for meals that balance the ingredients such that they provide proper nutrition and then provide you with the list of ingredients to purchase (or ship them directly to you).

We'll have IBM use Watson to provide the solutions -- call it "Big Iron Chef".

I would love that, particularly if it let one set different preferences for different days (e.g. for Catholics fasting on Fridays).

It'd be awesome to mix in some machine learning to figure out new recipes one might like, too.

I just did my own version of this diet with a nutrient-dense low-carb approach.

Ground beef: $3.50/lb = $2.19/day

Sardines: $9.99 for 6 - 4.375 oz = $1.67/can = $1.67/day

Fresh spinach: $3.99/2.5 lb (.35 cents/g) = $.84/day

Eggs: $7.99/ 5 dozen (13.3 cents/egg) = $.40/day

Sunflower seeds: $2.49/lb = $.36/day

Butter: $9.99 for 4lbs = $2.50/lb = $.12 cents/day

Total cost: $5.58/day

This is 1800 calories and is 65% fat, 30% protein, and 5% carbohydrate. I used http://nutritiondata.self.com to formulate it and it meets or (far) exceeds every vitamin and mineral requirement except for only 50% DV of Thiamin (not worried about this).

I basically do a version of this but with more expensive boneless short rib (very marbled and delicious) and am in perfect shape (you can find evidence here if you scroll down a bit https://twitter.com/FoodLiesOrg )

Spreading this menu over 2 meals (one at noon and one at 7:30pm) will give you a nice intermittent fast. It will also train your body to burn fat instead of sugar. You will be very full off of only 1800 calories and should lose a lot of weight.

Disclaimer: I'm not dispensing nutritional advice as I am not a doctor or dietician. I've just researched this stuff for years and am making the documentary http://foodlies.org

This can all be bought at Costco with the exception of Sunflower seeds on Amazon.

"Eggs: $7.99/ 5 dozen (13.3 cents/egg) = $.40/day"

Wait, does that mean 3 eggs/day? That would be asking for troubles (cholesterol).

The hypothesis about dietary cholesterol causing heart disease has been debunked.

"Previously, the Dietary Guidelines for Americans recommended that cholesterol intake be limited to no more than 300 mg/day. The 2015 DGAC will not bring forward this recommendation because available evidence shows no appreciable relationship between consumption of dietary cholesterol and serum cholesterol, consistent with the conclusions of the AHA/ACC report."

Source: https://health.gov/dietaryguidelines/2015-scientific-report/...

(Office of Disease Prevention and Health Promotion, U.S. Department of Health and Human Services - ODPHP)

Dietary cholesterol is actually not bad - check out this article for a good overview https://chriskresser.com/the-diet-heart-myth-cholesterol-and...

Have been researching this for a long time. It was hard for me to believe at first, but the science is sound.

> It was hard for me to believe at first

Nina Teicholz has a lot to say about the garbage state of nutritional science, the AHA and just how much expensive taxpayer-funded research has been buried over the past 50 years.

[0] https://www.youtube.com/watch?v=2_kEmYTiHtc

I was being coy. I've read Nina's book and heavily researched all of this. I'm glad you and other people know about this stuff. Let's make this common knowledge!

Vegetarian alternative possible?

I'm not a fan of a vegetarian diet because it leaves out the most nutrient dense foods.

I haven't done research into it specifically, but I would avoid carbs and sugar (and all processed foods), and eat a lot of avocado and protein from non-carby foods (beans and quinoa are ok, but still have a lot of carbohydrate).

google "vegetarian LCHF"

If anyone is trying to lose weight and wants to try this simple and cheap diet (or similar) I would love to document it for the film. Contact me at hi@foodlies.org

FWIW, when adjusted for inflation, $39.93 (the total annual cost of Stigler's 1939 diet) is approximately equivalent to $712 in 2018 US dollars(1). Still a paltry sum, though, for a year's worth of food for most people in the US, I think.

1. http://www.usinflationcalculator.com/

Or delivered on a pallet for about $1000/person-year: https://www.costco.com/THRIVE-1-Year-Food-Storage-31%2C500-S...

Note that that's $1000/person-year for 1300 calories per day, vs. 3000 calories on the Stigler Diet!

The diet I created and posted here is only $2,000 per year and includes fresh, nutritious foods that don't detract from your health like most of those foods loaded with carbohydrate and sugar would in your example.

Obviously they aren't comparable, I just think it's cool you could live in perfect health for a year for $2,000

(the irony of avoiding my own noprocrast settings)... For a university course, I managed roughly 600$ a year (local prices) with a (few-constraints-)diet of peanut butter and rice sandwiches. Even lower price for directly consuming spoonfuls of flour. Thus the added commentary:

- you manage what you measure (linear programming is really powerful, but also simplistic for this model)

- the human body is great at survival (general requirements are easily achievable, but surely something is wrong with the spoonfuls-of-flour approach)

Haha, the Stigler diet leaves out a couple of important items. Salt and yeast turn flour into bread, which is damn tasty. They're cheap, but not free, and should have been accounted for.

If you have a sourdough starter, then you don’t need to pay for yeast (though in my experience, I always end up pouring off excess starter, ie wasting flour).

Me too. Someday I hope to have a kitchen, life, and workflow that allows me to bake sourdough every week, but I'm not there yet.

It doesn't even need to be sourdough: you can make a starter for regular bread, too!

I mention this because I'm not typically a fan of the sourdough style, but I am a fan of starter-leavened bread.

I'm not sure that food is so cheap anymore.

How much would the items actually cost at today's prices, though?

Edit: Okay, I figured it myself. Around $1850 Canadian, where I live. Still works out to $153 per month, which is almost exactly what I was spending on food during my frugal student days, about 3 years ago. I ate better than the pretty boring diet described here (though not by much) and I'm also quite a bit smaller than the human described in the question. Still, it's good to know that I couldn't have done much better than I did.

Numbers and assumptions: https://imgur.com/a/udCL3

Aren't the inflation adjustments entirely based on the cost of common goods like flour?

Yes, but also a lot of non-food items.

https://en.wikipedia.org/wiki/Market_basket#Consumer_basket

There are a lot of ways to measure inflation. It's tricky because some categories of goods or services didn't exist decades ago, but do now.

For boring but optimized meals for nutrition and cost more people should really look at what bodybuilders eat. When you're a competitive bodybuilder every calorie counts and every meal is planned and carefully calculated. Now, most of you don't need the same volume of food and the protein amount is going to be higher than what your average sedentary person needs but still provides a great baseline for getting in the right nutrients and balanced amounts of fats/carbs/proteins.

> a moderately active man weighing 154 pounds

> 3,000 Calories

It's really striking how much heavier and how much less active Americans have become.

This bit stuck out to me too:

  I called my doctor and asked him how come the nutritional requirements didn't
  show a limit on the amount of salt? "Isn't too much salt dangerous?" He
  replied that it wasn't necessary; most people had enough sense not to consume
  too much.

As I started preparing home-made kefir, I realized that full-fat milk has a balanced protein-fat content, and is rich in some other nutrients. Fermentation to kefir reduces the lactose content while increasing some of the good things (vitamins, amino-acids).

In my calculations, about 3L of kefir per day should satisfy the energy and protein needs of an adult; for under $3/day.

If that'd work, it would be the first complete diet made from only one ingredient!

3L of Kefir is only about 2000 calories, which is not enough for most adults.

fun fact kefir is like 0.5% alcohol so that amounts to 15 mL of alcohol for 3 L of kefir! Not much at all, but probably adds up combined with other drinks

Now that's what I call a complete meal, the drink included :)

Here is a meal from these ingredients by a Google chef: https://gsuite-developers.googleblog.com/2014/09/sudoku-line...

> Google Chef Anthony Marco took it as a challenge, and we’re calling the result Foie Linéaire à la Stigler

Extremely cool until even the smell of navy beans makes you wretch.

Also, given the generally poor state of even contemporary nutritional science this strikes me as something akin to a particularly beautiful set of solutions for a universe in which the luminiferous aether is real.

BUT it is still a really neat optimization problem, and that is likely the spirit under which it was submitted.

Bookmarking this for later use. I've always wanted to find a recipe somewhere between Soylent and Lembas Bread that can be used for long hiking or canoe trips.

Unfortunately the beans and cabbage hardly sound appetizing for a bread. The obvious ideas are wheat, nuts, and dried fruits.

Any good resources for this?

Not a bread, but...

https://en.wikipedia.org/wiki/Pemmican

I love pemmican. Just be careful on the berries though, as they can vastly reduce how long it stays good.

Refer to my other comment. If you must, any constraint optimization solver will do. Here's two: [1] Example http://apmonitor.com/online/view_pass.php [2] HN Thread https://news.ycombinator.com/item?id=15530630

There's a difference between food good for survival and food that is good for you.

You can buy modern high calorie food bars that will last ten years and contain 3000 cals in a tiny package, but they're mostly carbs. Again, great for survival but not something to live off of while on a fun trip in the backcountry.

I usually pack precooked meats (frozen if needed), dried meats, firm biscuits or crackers, and some nuts and fruit. Granola and oats are good too of course.

If you have conveyance (a canoe or kayak, an atv, etc) then you can afford to bring much less dense (and usually better tasting) food.

> If you have conveyance (a canoe or kayak, an atv, etc) then you can afford to bring much less dense (and usually better tasting) food.

Can confirm, many years ago went on a canoe trip on the Shenandoah for a week and we brought about 400lbs total food and equipment per 2 man canoe. We ate like kings, and mine and my boat-mate's canoe was the only one that didn't capsize on the trip (everyone had fully waterproof bags, so no one lost anything, it was just a matter of pride).

Logan Bread https://sectionhiker.com/logan-bread-recipe/ (various recipes out there, but I've tried this one)

or Mealsquares: https://www.mealsquares.com/

The Logan bread was tastier, though the Mealsquares are really pretty good just a bit dry.

PB&J are great if you don't want to break camp. It sounds lame but they're basically fruit nuts flour and salt.

You could try fermenting the cabbage to get a lot more flavor.

There's a guy in the UK who eats nothing but beans on toast... Apparently, not super healthy, but not too bad either. http://www.dailymail.co.uk/health/article-3089167/Could-surv...

I created a diet optimized for maximum dietary fiber and protein with minimum calories (optimized separately). Highest fiber/calorie ratio is generally cruciferous vegetables; highest protein/calorie is fish and chicken (skinless boneless breast meat). Using these two food groups I can satisfy my macronutritional requirements on 600 calories/day. I add a multivitamin in case I missed anything.

Based on these parameters I can judge any food based on whether its fiber and protein can match (or beat) a linear combination of chicken and broccoli. Basically if (2F + P) * 20/3 <= C then it's on the diet. This includes non-fat yogurt, wheat bran, eggplant, and some types of beans.

Not all fiber is equivalent. The fiber in legumes tends to be highly fermentable, and provide far more in the way of health benefits than the fiber in vegetables such as broccoli.

Lentils are pretty much nature's perfect food when considering cost, protein and fiber content. Additionally, lentils are low in methionine. I don't put much stock in methionine restriction for human longevity, it does work in rodents, so if you want to be a guinea pig (couldn't resist) they're good in that respect too.

Eggs are also excellent for their high essential fatty acid content, and this may sound counter-intuitive but if you eat a lot of beans dietary cholesterol is actually beneficial. High quality eggs are also a really good source of vitamins E and D.

Finally, though they're unfortunately not so cheap, hemp seeds provide you with both omega-3 and omega-6 fatty acids and minerals. A mix of sunflower or pumpkin seeds and chia seeds works well and is cheaper if you're optimizing for cost.

The only vitamin that I think is worth supplementing is niacin. Niacin has been shown to increase BDNF levels in animals which can help with mood and memory formation, and it has also been shown to increase lifespan by 10-15%. For people who don't go outside and don't consume much milk vitamin D might also be a good idea.

> if you eat a lot of beans dietary cholesterol is actually beneficial

Interesting, I have found some references to beans lowering cholesterol but not this stronger statement?

I've read that the link from dietary cholesterol to body cholesterol levels is genetically variable, but that there's no test at present for determining individual risk factors.

An extremely high fiber diet has been shown to negatively impact testosterone production. Cholesterol is also important for nerve myelination, and high cholsterol intake has been show to magnify strength gains from resistance training.

it would be interesting to try to constrain / regularize this problem such that the solution is palatable. User could answer some questions about ingredient preferences to build approximate set of weights for the LP objective. Also easy to add some constraints that prevent huge amounts of one food. Tools to allow the user to "critique" the solution which generate new constraints or modify the objective.

Also would be interesting to run the same LP over historical ingredient price datasets and see how the optimal diet changes as ingredient prices fluctuate.

“No one recommends these diets for anyone, let alone everyone.”

But surely there must be some hacker out there who has decided to try it.

Yes, I did :D

Huh, I did a similar thing a few years ago with much fewer ingredients to come up with a low budget diet (I also used multivitamins to fill in the gaps). I just threw a simple evolutionary approach at the problem, because modern computers reward laziness.

I'm sure we could attempt this today with a larger dataset of foods and nutrients. I know that there's nutritiondata.self.com and the usda's nutrition database, anyone know of an average price dataset for food?

http://nutritiondata.self.com uses the USDA database.

I made a diet with only 5 foods (plus butter) that is ideal for human health and weight loss (posted on this thread).

https://www.completefoods.co/diy/recipes

> a 154-pound male

That's so XX century it hurts.

Reply on news.ycombinator.com