Moneyball of Weight Loss

November 15, 2012

I want to use what I learned from reading Moneyball by Michael Lewis & watching Nate Silver (fivethirtyeight.com) using his unique analytical skills to accurately predict elections.

So, as an ode to their genius, here’s my own amateur attempt at predicting when weight loss will happen using my existing data.

First a few caveats.

This is highly nonscientific. I’m not measuring my weight every day. I usually do it whenever the mood strikes, which turns out to be around 3-4x a week.

I also don’t weigh myself at the exact same time or under the same circumstances.

Even though I took statistics in college, I’m not very proficient in it. I’m just going to eyeball what’s happening looking at trends w/ the naked eye rather than thru formulas.

Finally, I’m adding new weight loss techniques frequently, so the rate at which I lose weight should theoretically increase.

With that out of the way, I wanted to just have some fun to see if I can model how my body likes to drop weight.

Data & Charts:

I have a few weapons to keep me motivated when weight stagnates week to week or even goes up.

I look at trend rather than actual weight loss.

Even though I only officially keep weight loss score once a week, I still like to weigh myself multiple times a week unofficially just to see if I’m losing weight.

This helps me know if the trend is heading in the right direction or not.

And I do this at 18 data point intervals (just an arbitrary number I picked). I say 18 data points rather than 18 days because I didn’t weigh myself every day.

For example:

The first set of charts show the period where I was eating a lot of paleo friendly treats like plantain chips, dehydrated organic cookies made w/ almond flour/butter/honey, element bars, etc.

This is the actual weight fluctuation for the those 18 data points.

Just by looking at that, we kind of get an idea of what’s happening compared to if I kept no data. But it’s still not that clear as to what’s really happening.

So let’s take a look at the trend of those 18 data points:

The trend shows that weight was going down steadily for a while, then reversed course and started creeping back up.

When I saw this data, I immediately stopped ordering & eating any more paleo treats.

Once I stopped, I wanted to know what exactly would happen next.

Here’s weight chart for the next 18 data points:

It looks like weight is going down and the trend paints a much clearer picture of what’s really happening:

I was happy to see this trend which in turn kept me super motivated to keep doing what I was doing.

Lastly, this is the most recent data set where I’ve added a lot of new tools of weight loss (high dose fish oil, bulletproof coffee, earthing, meditation software, sea salt + water in morning, eye mask + ear plugs while sleeping, etc)

There was a huge initial drop, then kind of a lull the past few days.

The trend shows:

The weight is dropping at a slightly faster rate so far and is looking similarly as 18-32. This is a great sign even though this week, I gained +1.1 week over week.

See how having this data is super helpful?

In the past, I never kept track of data like this and it was very demoralizing for my psyche to see weight scales go up even +1 lb, even if all else was going well.

Predicting when huge weight drops will happen:

I want to play a poor man’s Nate Silver and try and see if I can predict when huge drops in weight will occur for my body.

There seems to be a pattern. I’ll lose a large amount of weight one week, then a few weeks of stability, followed by another week of huge drop.

This is what I came up with. I compared actual weight and looked at when the drops happened. Then I just projected forward when the next one should happen.

My definition of a big drop (or gain) is basically +/- 2lbs. So I ignore the weeks where weight went down -0.9 or went up +1.1.

Next, I just counted all the weeks where weight didn’t have a big drop. It seems like I have a big drop every 5 weeks. The only exception so far is in week 14 where I had back to back big drops.

If we take that possible outlier out, the standard of losing big every 5 weeks stays true.

Also, it should be noted that the first 4 weeks of data being tracked, I wasn’t really dieting. I tried random things like green smoothie fast etc but for the most part, I just ate fairly normally.

I started doing paleo diet full force starting 8/7/12 and week 5 reflects the first weight loss due to that.

Anyways, I roughly predict the next big drops to occur on week 21, 25 and 29.

Why did I do all of this?

I needed to find a system to take emotion out of weight loss. Emotion was the overriding factor for most of my past failures over the past 16+ years of yo yo dieting.

If I can step out of the day to day scale fluctuations and even get out of the week to week changes to take a look at the big picture, I can keep calm and move forward with what’s working so far rather than panic.

So, my two analytical tools (trend graphs & big drop predictions) will be super helpful in losing weight & keeping a calm mind.

The trend is like a compass and accurately shows if I’m going in the right direction. If at any 18 data point set, I see the trend going up, I can immediately reassess exactly what I’m doing wrong and fix it before it’s too late.

Big drop predictions help temper my high expectations of losing large amounts of weight every week. I can see how my body actually reacts to dropping weight and plan my emotions accordingly.

This is fun. Hope all of this made some sense!

Source: Idea & formula for calculating trend thanks to The Hacker’s Diet (see “pencil & paper” section)