Holy Crap I've Got Abs (Kinda)

Origin Story

My story is pretty standard. I was always the fat kid in school, always prefering sedentary activities to well, active activities. I never liked physical activity much. The excuse I gave was that my pursuits was in the intellectual, not the physical, and I derided people who chose to play sport as nothing but physical brutes. Then I got bitten by a radioactive spider. Then I became obese.

Now, contrary to what the fat activists/advocates out there on Tumblr say, weight loss is simple, and it's sustainable. Understanding what works and what does not doesn't require specialized degrees in nutrition science. All you need to do is educate yourself[ref]To those who get the reference, it's been a pleasure flaunting my thin privilege at you shrimpdittles[/ref].

So what is my conclusion regarding weight loss? Simple: weight loss (or gain) is simply a function of calories in and calories out. One is highly controllable (calories in), while the other is mostly difficult to control without supreme amounts of effort (calories out).

Diet

It began with flirting with Paleo about 4 years ago, and then keto, and then various other diets. At the point I was very much into the Quantified Self movement[ref]Now I'm not so much a fan of tracking everything for the sake of tracking everything, which seems to be what QS people are all about. I'm more interested in tracking data to be used to modify my own behaviours[/ref], and it was through data collection of what I ate and the weight fluctuations that I worked out that Calories In vs Calories Out was the sole reason for weight loss.

However, I eventually still did put on weight with those diets. Because I kept track of the data, after analyses, I realized that I eventually ate larger and larger portions as my stomach (and brain) got used to the satiety from the extra protein intake.

TDEE

Having analysed and realised it, the battle was still ahead for me. To lose weight, I'd have to eat at a deficit from what my body uses. Thankfully, a quick regression analysis showed that my Total Daily Energy Expenditure (TDEE) was somewhere around 2200 kcal (when I was still obese). This was roughly corroborated by the Mifflin St Jeor equation. I was eating close to 2400-2600 kcal at that time.

Snacking

Things that I tried that didn't work included snacking at pre-allocated times, smaller plates, putting the snacks far away so I had to walk to acquire them... These all didn't work due to the general availability of the snacks. There was little to no cost of me getting the snacks to eat. Furthermore, through an audit of time thanks to RescueTime combined with a FitBit, I found out that I was simply snacking out of boredom (or when my mind is on a problem and needs to idle in the background for a bit).

After a while, my weight came down, and eventually my TDEE too changed. But I was still overweight. So I recalculated my TDEE, did some math stuff, and figured out that my TDEE was roughly 2000 kcal. To lose weight at the rate I deemed fast enough, I had to eat at 1600 kcal per day. And thus came the difficult bits: portion control.

Portion Control

I have come to learn that in Sydney, everything is ridiculously portioned, especially if you eat out. It took me a while in my early days of tracking what I eat, to figure out that sometimes the portion sizes were not standard. A standard takeaway box of pad thai, for example, serves 2 - I mean that to say there are enough calories in that one box for two meals.

Here's a half-hearted compilation:

I was lucky that I realized this early on, and I had started weighing my food on a scale. But still portion control was difficult. A large part of that was a mental hurdle.

Hunger

When you've eaten large portions for a sustained period of time, your brain gets used to it, even if your body doesn't actually need the energy. So when you you start eating slightly less, your brain sends itself hunger pang signals. I have found, too, that certain macronutrients does indeed affect hunger signals as well.

From my notes, here is how actual physiological hunger feels like: sleepy, and foggy. The brain DOES indeed receive similar hunger pangs signals as when I ate less, except I could also feel my stomach twisting (not sure if that's actually happening, but that's the sensation), of which the psychological hunger didn't feature. It also only happened about 12 hours after I last ate, for a period of 3-ish hours, and then it stopped. For the rest of the two days I felt mostly sleepy and as if a fog had came over me, like when I'm down with a flu - in fact it does feel almost exactly like that, without the fever or the bone pains.

With that experiment done, I was convinced that a lot of my hunger was mental (in fact I had repeated that experiment a few more times in the past 1.5 or so years).

Chasing Macros

I consider myself to have nailed down eating the correct calories by March this year - it had taken about a year of trial and error. By correct, I mean eating to the goal amount of calories. The remainder of the year though, was an attempt at me chasing down the correct macronutrients.

Cutting carbs was mainly so that I could get consistent readings off my scale, day in day out[ref]Sure, I could have factored in my carb intake when doing my analysis, but really that's a lot of extra murk in the analysis phase which I didn't want, it was better to consistently intake less carbs and make my readings as consistent as possible, and reduce the amount of random error, which is high enough as it is.[/ref].

That all said, I have never actually managed to really control my carbohydrate intake too long. The longest streak in keeping my carb intake under 70g per day was 10 days in August. The typical streaks would be 5 days (weekdays), followed by a couple of days with slightly higher carb intakes (under 200g per day). Now, if you're familiar with PSMF diets, it may remind you of refeeding periods, which came naturally due to the lack of discipline on weekends.

This is especially difficult when eating out, given that the amount of carbs I was chasing was low. Outside food is generally high in carbs and fats while low on the protein side. Think of your burgers, pizzas or pad thai - the majority of the content are carbs and not protein. For many people, the solution is to eat out less, and cook more at home. Or find places to eat that serves food that fits their macros. I did similar things, but I didn't entirely cut out eating out.

Amongst my favourite food to eat for macros is kangaroo meat. The protein-to-calories ratio is just nuts. Here's a favourite kangaroo recipe of mine:

Exercise

The whole idea of "eat less, move more" should really be "eat less, move a LOT more". For a while, I had a FitBit, and then a Jawbone UP and through analysis of my intakes and output, I had come to the conclusion that fitness trackers are a waste of time[ref]For the purposes of tracking output, that is. There are other motivational things a fitness tracker does to your psychology but that's for another day[/ref]. It is essentially wankery, and what I usually call "vanity statistics".

Instead, I focused on strength training. Having been considered weak all my life, it was something I was intent on changing. My goals for exercise is not to lose weight, but to gain strength. How did I do? Here's an idea:

https://www.youtube.com/watch?v=8SJm07INTas

The Plan

I'm now back on StrongLifts for a while as I've just returned from a month long holiday. I eventually plan to get back onto a PPL-style routine after I've fully ramped up with StrongLifts.

Eventually my plan evolved into this PHUL-based program:

Progression

This was how I progressed for the main lifts over time (since June 2015, due to my misunderstanding of the StrongLifts app, I had accidentally deleted all data everytime I restarted. Mehdi that's not how you design an app!):

I'm personally not particularly proud of the progress I made with the PHUL-like program. The problem is when you design your own, there is a mental tendency to half-ass it. Whereass with Stronglifts 5x5, the program says: Nope, you do 5x5, or you don't progress. That's a hard and fast rule which psychologically affects the actions I take, and the motivations I have to push through.

Optimizing For Different Things

One thing I observed, both through data, and going through the motions, is that different programs optimize for different things. StrongLifts for example, optimizes for increments in mean weights, while PHUL optimizes for increment in total volume. The latter has an effect of not increasing weights as quickly as StrongLifts.

Motivation

Lifting on a cut... generally sucks. The body hurts quite a bit. And that can dampen my motivations quite a bit. There are just mornings where every muscle hurts too much to drag myself to the gym.

The going back to the training program bits is what I dread the most coming back from my month long vacation. I have no idea how to get myself started back into the habit.

Goals

Data Science!

Acquisition and Cleaning

Date | weight | bf | carbohydrates | protein | fat | calories | Exercise1 Sets | Exercise1 mean(Reps) | Exercise1 mean(Weight) | Exercise1 sum(Reps) | Exercise1 sum(Weight) | Exercise2 Sets | Exercise2 ... | ...
-----|--------|----|---------------|---------|-----|----------|----------------|----------------------|----------------------|---------------------|-----------------------|----------------|---------------|----

• Tanita BC-541 scale - for measuring my weight, bodyfat percentage and water percentage
• MyFitnessPal - for recording weight, bf, macros and calorie tracking
• StrongLifts - for recording StrongLifts tracking
• FitNotes for Android - for tracking of non StrongLifts lifts

Anyway, I digress. Most of these apps have an export-to-csv function. Which is great news, because then I can munge the data around. With MFP, they provide an API, which you must use to build your own CSV. Here's a quick and dirty script for that (I used Adam Coddington's excellent MFP package for Python):

Munging the data from other sources is then a simple case of using Pandas. In particular, the unstack() method of Pandas' dataframe object is extremely powerful. Here's how to do it in fewer than 30 lines of code:

You might note that the StrongLifts file was already unstacked. Truth is I had a lot of time during my vacation, and had access to a computer with no development tools on it (but had Excel), so I manually unstacked the data from the StrongLifts CSV using Excel.

Figuring Out TDEE

Figuring out my TDEE was quite a challenge, but once the basic principles are understood, analysis becomes simple. I have obviously done a lot of work on this topic, and I can say that any kind of measurement of "Calories Out" will only add to the noise of the data, and you'd be better off ignoring any logged additional "Calories Out"[ref]Your body uses calories just to survive, it's called the BMR[/ref]. Instead, we want to figure out how much the body uses. It turns out the Mifflin St Jeor formula (with Sedentary as my activity level) is pretty similar to the results I get thru statistics.

Let's say you have a table like this:

To calculate the transformation, first we need to pick a number of days to compare the differences. I picked 7 (because it's a nice number - we have 7 days a week, and we can hence compare Mondays to Mondays). So we take the difference of a week for the weights, and as for the caloric intake, we'd have to get the average caloric intake within that week[ref]alternatively, and arguably clearer in narrative - you can also take the sum, but this article uses the average, and all calculations will be done assuming the average was used[/ref].

This is how it looks like when you plot it out (red line is weight difference, blue line is average caloric intake):

A very obvious thing is that there is a time factor involved. Sometimes the weight change will peak or trough before the caloric changes. However, most of the time the caloric intake leads the weight change. This is a logical result - you don't put on weight on the same day you eat. This lag or lead is mostly affected by the amount of carbohydrates being eaten in the previous days, and will vary from person to person (my wife for example, has a obvious 2-period lead).

So to find my TDEE, it's just a matter of figuring out the best fit line:

Once that's done, we just have to set Y to 0, and that's my TDEE.

$latex 0 = 0.0005506x - 1.206424$

$latex x = 2191.1 kcal/day$

3500 to a Pound

There is a truism about losing weight: that 3500kcal equals one pound. If you eat an excess of 3500kcal a week, you will gain one pound a week. Likewise, if you eat at a 3500kcal deficit in a week, you will lose one pound per week. I don't like this mainly because we're in 2015, and we should really be using kilojoules and kilograms. However, with this, it's easy to perform a sanity check on our data above.

According to the best fit line here, I am losing weight at about 0.0425kg per day.

$latex \frac{0.453592}{3500} \times 453.43 = 0.0588$

The difference in expected and reality is due to the fact that 1 pound of human body fat burn up in a bomb calorimeter to be 3500kcal. When eating on a deficit, you will lose fat and muscle at the same time (which is why I do weight training while at a deficit. Weight training promotes muscle growth, so the net muscle loss isn't as high). Add to that the various systemic errors that comes with measurements, and the fact that some macros end up undigested in poop and other miscellaneous errors that add up, I think that my estimated TDEE above is fairly decent. I could certainly improve on the systemic error bits but hey, nobody's perfect.

Macronutrients Don't Matter (for weight loss)

So now that we've empirically addressed that weight loss is mostly due to calories, we can now move on to see if macronutrients do matter. Short answer: no. The long answer is as below.

So from a logical point of view, there isn't any reason to think that a calorie released from breaking carbohydrates is different from a calorie from breaking down proteins.

Residuals:
     Min       1Q   Median       3Q      Max 
-1.14110 -0.25445  0.03073  0.25689  1.26645 

Coefficients:
                Estimate Std. Error t value  Pr(>|t|)    
(Intercept)   -0.9541062  0.2091143  -4.563 0.0000147 ***
harmean.carb.  0.0094769  0.0072935   1.299     0.197    
harmean.p.    -0.0042279  0.0078548  -0.538     0.592    
harmean.f.     0.0114412  0.0172750   0.662     0.509    
perf.cal.     -0.0002891  0.0018204  -0.159     0.874    
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 0.4687 on 98 degrees of freedom
Multiple R-squared:  0.3324,    Adjusted R-squared:  0.3051 
F-statistic:  12.2 on 4 and 98 DF,  p-value: 0.0000000436

However, once you understand what calories really are, you'll note that calories ARE indeed comprised of macronutrients, you begin to suspect that there may be some multicollinearity going on, which in fact it does:

> vif(fit)
harmean.carb.    harmean.p.    harmean.f.     perf.cal. 
     30.02082      16.94114      31.98968     126.10371 

In fact, we can regress perf.cal against the carb, protein and fat intake to have an idea of the multicollinearity:

Residuals:
   Min     1Q Median     3Q    Max
-48.48 -23.12  -1.30  20.96  47.68

Coefficients:
               Estimate Std. Error t value Pr(>|t|)
(Intercept)   -4.69978   11.53535  -0.407    0.685
harmean.carb.  3.93330    0.07663  51.329   <2e-16 ***
harmean.p.     4.03605    0.15335  26.320   <2e-16 ***
harmean.f.     9.15083    0.25255  36.234   <2e-16 ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Residual standard error: 25.87 on 99 degrees of freedom
Multiple R-squared:  0.9921,    Adjusted R-squared:  0.9918
F-statistic:  4128 on 3 and 99 DF,  p-value: < 2.2e-16

I make extensive use of MyFitnessPal's barcode scanning function. I had noticed that sometimes the values don't add up (Aldi products, I'm looking at you). So I have also added notes on whether a day's record was a "perfect" record or not.

As they say, garbage in, garbage out. The regression analyses were done based on the "perfect" subset of data. This simply acts to reduce the noise in the data. It turns out when you measure your day perfectly, the values turn out to be exceedingly close to the theoretical values.

Macronutrients Matter (for strength training)

Besides my weight, macros, calories and weight training data, I too recorded several other things, like pain and sleep hours. Sleep hours, being recorded by my fitbit, ended up generally not too useful, as I had difficulties extracting the data from FitBit[ref]There was also this[/ref]. I have contemplated using my Emotiv Insight to record sleep, but it really wasn't comfortable to sleep on.

An astounding fact is almost everyday my pain level is at an average of 1. There have been days where the pain climb to a 4 or 5, and once for about 3 days the pain levels were about a 7[ref]but that was because I suspected I had dengue fever... which is a story left for another day[/ref]

A quick regression analysis had shown that the amount of protein I eat actually does lower the pain levels. Again, the statistical power isn't great, but my gut feel tells me that this is pretty valid, up to a certain point.

There is also a truism online too, that higher protein intake prevents losing muscles while on a cut. I had no way of figuring out if this is true as my BIA based scale isn't the most accurate (with 1 decimal place). I did some work in this area, however the results were inconclusive.

Supplements and Substance

Lastly, I was curious about the effects of the various supplements and substances I had on my training. I take creatine as a supplement, and the substances I use include caffeine, alcohols (mainly beers), modafinil, and very occasionally, some trees.

Over time, I have cut out alcohols from my diet to my best effort, mainly because alcohol affects my sleep. After a night of drinks, I'd suddenly wake up at 3am, and be unable to fall back asleep until the next night. This happens even with just one beer. It's irritating.

Lastly, cannabis is a mass gainer. 'nuff said. The three 3600 kcal days were all directly attributable to weed. It doesn't affect anything else. I don't use it a lot, as it does nothing for muscle pain either (does help with joint pain though).

Photography

Other than your standard diet and exercise, another bit that is quite vital into making the picture above was lighting and photography. Of course, while the photo on the right is largely unedited[ref]There are definitely some edits: mainly to crop the image, and to change the white balance a little. There was also some frequency separation (force of habit), but no dodging and burning to make anything pop.[/ref], my muscle definition is really showcased by the lighting setup that I had. Lighting plays a large part in taking nice photographs where the abs pop. Here's how I lit my scene:

The scrims were actually frosted glass doors, and the flash was a YN460 at full power. I tried various heights before settling it at about shoulder height. f/8, ISO 400 at 1/80.

Thoughts

It's been an interesting year for me, most definitely. A lot of effort was expended getting my diet right, and I can foresee it being an ongoing thing. A tweak here and a tweak there. In retrospect, I wish I hadn't been so undisciplined, and have taken more types of measurements - like neck size, arm diameter, etc. That way I could do better analyses, but ah well, I can leave that for next year's progress reports.

In closing, calories in vs calories out is very much valid, and my own data proves it. As an additional fun fact, my rs5082 genotype is CC(or GG in 23andme terms). I would love for people to come up to me and say I have the genetic advantage of not being obese. I've not checked my genome for the FTO gene though. The point is, there isn't a reason to blame genetics (or medical conditions) for being fat. There isn't any moral reason to be fat; conversely there is a moral reason to be fit. Losing weight is simply a matter of discipline (of which I don't have a lot, and yet still managed to do it) and self control.

Lastly, I did enjoy the side effects of losing weight and getting stronger. I have more stamina, and my productivity (as measured by RescueTime) has improved a lot year on year. I also did enjoy writing SquatCoach, and learning the mechanics of various body parts. New knowledge is always welcome in my head.

• Caloric restriction works
• All the fancy diets - paleo, whole 30, keto, IF etc - are various different types of method for caloric restriction
• Figuring out TDEE via statistics is a LOT more fun than figuring out your TDEE via the Mifflin St Jeor equation[ref]bearing in mind that I'm technically a professional statistician, so I definitely find stats a lot more fun[/ref]
• These are my tricks for cutting:
  • Cut snacks altogether (by not buying them)
  • Since $5 is 500kcal (in Sydney), split your meals into $5 portions
• Macros don't really play a part in weight loss - carbohydrate intake does affect short term weight change (this is commonly known as water weight) - so use 7 to 14 day weight change instead.
• Macros play a small part in training - more protein means less pain when training while on a cut
• Creatine appears to have an effect on total volume - the statistical power is not strong, but given more data, I'm sure it will show more strongly
• How simple the analysis of TDEE, macros etc can be. Anyone with Excel can do it, and verify for yourself!
• Fitness trackers are a waste of time, and add noise to your data. Don't use them if you are not using them for motivation
• You should try out SquatCoach if you want to learn to squat with proper form and you have nobody to spot you
• Lighting and photography also plays a part in getting nice progress pics

TL;DR

Resources

• Is a calorie a calorie? - a very good paper with an overview of many other papers regarding macronutrient intake and weightloss/gain
• Ketogenic low-carbohydrate diets have no metabolic advantage over nonketogenic low-carbohydrate diets. - exactly as the title says
• Energy intake required to maintain body weight is not affected by wide variation in diet composition. - again, what it says on the tin
• The effect of protein intake on 24-h energy expenditure during energy restriction. - My own regression analysis (regressing macros against weightloss/calculated TDEEs) confirms this: protein intake has a small effect in increasing energy expenditure[ref]I edited it out of the section above because interpreting the regression coefficients require some finesse, and could easily be misrepresented if I don't explain it well, and the blogpost is at 7000 words as it is right now[/ref]
• Effects of Dietary Composition on Energy Expenditure During Weight-Loss Maintenance - A calorie may not be just a calorie afterall. This is one of the few papers out there which challenge the notions that a calorie is not just a calorie. The details are in the statistics applied and the interpretation of it. I wouldn't be quick to jump to cite this paper as proof that a calorie is not a calorie without fully understanding the statistics and measures used in the paper. It is however, a valid interpretation.
• Greg Nuckols' Strengtheory Website - he breaks downs a lot of interesting papers, and things that I'm starting to collect data for (though I would imagine a lot of it wouldn't apply to me as I'm not advanced in weightlifting in any way).