Week 49 Wrapup

I ran a tempo run on Monday, but my foot still hurt a lot after that, so I took a few days off to rest it. I did get to a spin class with my Dad on Wednesday morning. I fared much better than I did last time, which means that either I was really wiped out from the marathon last time, or I was just being smarter about it this time and not pushing so hard right at the beginning only to crash 10 minutes in.

We went down to Rocky Point for the weekend, and there was a bunch of family there that I don't get to see very often. My cousin Randy was there, and since I hadn't seen him in a while, I was unaware that he had taken up running and lost 40 pounds over the last year or so. He really looks great.

On Saturday morning, I enlisted Randy and his wife Lisa to accompany me on a long run. We left the house at about a 10:00 pace and ran down the Las Conchas road to the guardhouse, about 4.1 miles. They've both been fighting injuries the last few months, so they turned around at the guardhouse and then walked most of the way back. I continued on to the Pemex at the corner, then turned right and headed down the road that goes to the Caborca highway. I continued on that road for about another mile, then turned around and headed back, stopping at the Pemex to buy water, then stopping again to chat with my Dad, who was just headed out of the neighborhood on his bike. I ran back into Las Conchas, then stopped at the 12 mile point and walked the remaining 1/4 mile home.

It was a good run. My foot bothered me a little the first mile, but the pain went away, and it didn't hurt at all after the run. The last two miles were a little more of a struggle than the first ten, but only just barely. I think the fact that 10 of the 12 miles were on dirt helped a little so that I wasn't so sore afterward, but I also feel like I'm just getting better at this. I finished the 12 miles at a 9:54 pace, and that's even including the stop at the Pemex and chit-chatting with my Dad.

Lisa ran the St. George marathon last year, but Randy hasn't done a marathon yet. So, we all talked briefly about doing St. George or some other marathon next fall. I hope it comes together, because that would be fun.

Next week, I'll try to do a more normal training schedule, then on Saturday, Elizabeth, Chuck, and I are going to do the Bisbee 1000.


Moving Average

Carson asked how the weighted moving average is calculated, so I'll post my long-winded answer here.

First some review: I've said it before, but it's my strong belief that basing your weight loss on the comparison of daily readings is a recipe for failure. Any human will have a fluctuation from day to day and even hour to hour just from their own biological processes. The only way to directly compare one day's weight against another's would be to completely eliminate all food or waste products out of the GI tract, void the bladder, and bring the water percentage of the body to a specific predetermined value. To be really precise, you'd also have to make sure your hair and nails were at the same length, you had just showered to remove extra surface dirt, and that you blew your nose really well.

That's impossible to do on a daily basis, or at least really difficult. But, you still want sets of numbers that you can compare against each other. So, what's needed is to somehow average out these other factors to smooth out the noise introduced by measuring them day to day, leaving numbers that, while not technically an accurate picture of your weight at a precise point in time, are at least a better basis for comparison. You can do that in several ways.

You could only weigh yourself once a week, or once a month. That's what some diet places do. They tell you to not even bother to check your scale at home, and only go by what your reading is on their scale when you come in for your weekly visit. That will smooth things out somewhat, but it still has the possibility of tricking you. If one week you weigh in right after a heavy duty workout, but the next week you weigh in after a big dinner, it might appear you've gained 2 pounds in a week. In reality, your body may have burned two pounds of stored fat that week, but you have four more pounds of water and food in you than you did at your previous week's weigh-in. So, even weekly doesn't smooth out enough, and any longer than weekly and you're no longer getting useful feedback, in my opinion.

Another way to try to smooth things out would be to weigh daily, graph all the daily readings, and just see if the line is generally going up or down. That's easy to see once you graph enough readings, but kind of harder with just a few data points.

A more specific form of this would be to actually plot a trend line through your data points, plotting a straight line or a curve that fits the line such that half of your points are on one side, and half are on the other, and they're generally evenly spaced from the line.

That's too much work, I think, and although I could have Excel plot the trend line for me, a straight line plotted through all my readings isn't informative enough about whether I'm losing or gaining in a short period, like a week. And, the curved lines that Excel plots are not "curvy" enough to be useful for short periods as well.

So, I look to some sort of averaging method to tell me whether I'm really gaining or losing. For obvious reasons, I don't want to just average all the readings I take. If I did that, my current average weight (taking all readings from 11/07 until now) would be around 200 pounds. Yes, that's less than the average would have been in December 2007, but it's not a useful number. I could average all the readings in a particular month or a particular week and compare those, but I'm obsessive enough that I want a number I can compare on a daily basis.

That leads me to the idea of a moving average. Simply put, a moving average is just an average of the last x number of readings. I could calculate a 20 day moving average by just taking my last 20 days worth of readings and averaging them. Then the next day, I'd be counting 20 days back from that new point and calculating a new average. Each day, as I make a new reading, the oldest reading drops off and I make a new average from the 20 most recent days. This is cool, because it gives me a number I can compare on a daily basis, but has a little bit of a problem in that one big spike or dip has the same effect on the average on the day it happens as it does 2 weeks later. If I cram 6 pounds of donuts into my donut hole, then weigh 6 pounds more the next day, that abnormally high reading is going to pull my average up for 20 days, even if I pooped them all out and had constant readings since then. Then, when it drops off the list on day 21, my average will go down significantly, even though my weight was 100% constant during the intervening time. I didn't like the idea of something I did 2 or 3 weeks ago affecting my average as much as something I did yesterday.

Everything I've written here is pretty much an accurate depiction of my train of thought as I started to think about weight loss, and how I wanted to track whether I'm gaining or losing. Then, a couple of months into my health quest, I read The Hacker's Diet, and specifically the "Signal and Noise" section. Everything in there made perfect sense to me, specifically the argument that the best way to visualize a data trend hidden behind a lot of noise is to use a moving average that assigns a higher weight to more recent readings than to less recent ones. Once I got a handle on that, I converted all of the previous data in my logs to be able to chart a weighted moving average from it.

The actual formula used is an exponentially weighted moving average with 10% smoothing (or a smoothing factor of 0.9). There's something in the Excel Analysis Pack to do the incredibly complex (for me) math involved with making an exponentially weighted moving average, but the "Pencil and Paper" chapter in The Hacker's Diet taught me a little trick to make it easier. Basically, when using 10% smoothing, you can get the same numbers by just taking your current day's weight, subtracting it from the previous day's moving average, then multiplying the result by 0.1 and subtracting that result from the previous day's moving average.

So, in my spreadsheet, I punch in that day's weight, then the next cell takes that weight and subtracts it from the previous day's average. Then, it takes a tenth of the difference, and subtracts that from the previous day's moving average to calculate the new moving average. (Then, the number it displays is rounded to the nearest tenth of a pound before display in the cell. So, in it's calculations, Excel's keeping whatever precision it has for the individual cells where the weighted moving average is kept. If I were doing this by pen and paper, I would round off after calculating the weighted moving average, or perhaps even after multiplying the weight by 0.1. If I did that, I might find my average to be off by a tenth of a pound on a few days here or there compared to the computer version, but it would generally match what my spreadsheet comes up with.)

So, here's an example: On October 7th, my weigh-in was 180.2 pounds. Subtracting that from the previous day's moving average of 187.0 gives me a difference of 6.8 pounds. Multiply that by 0.1 and you get 0.68. Subtract that from 187 (the previous day's moving average again), and you get a result of 186.32. Round to the nearest tenth, and 186.3 is the number I post for my weighted moving average for October 7th.

A couple of other things: To remove as much variance as possible, I always weigh in naked right after getting up in the morning, but after voiding my bladder and (if possible) moving my bowels. I also will wait until after a shower if I'm taking one in the morning, because for some reason, my scale pretty consistently measures me 0.2-0.4 pounds lighter after a shower than before. I can't possibly think of why. The only possible theories I can come up with are these: 1. I just have that much grease in my hair and dirt on my body. 2. Being in the shower pulls more water out of me by making me sweat or through come other mechanism. 3. Being in the shower accelerates my respiration and metabolic rate so much that I just burn through the food or fat in my body. 4. The humidity in the bathroom somehow affects the scale to read lower. My money's on 2, although it's still weird.