Applied Nate Silver – Failed experiment: A swing-and-miss is still a miss

In science they say a failed experiment is still a success because it reveals what doesn’t work and thus eliminates that possibility. Fine, true enough, but in my proximate case a fail is a big disappoint and setback. What am I talking about – you say? Well, if you’ve read preceding articles you’ll recognize this “money” graph I’ve been updating every week as my long-term experiment proceeds:


The point back in week 11 (post-xmas) was the first substantial deviation from the trendline indicated by the red points. Breaking that spell has put me in the position of trying to get back on plan. In week 12 a super effort showed improvement, then reinforced by the result in week 13. But now in week 14 another setup although it’s not so obvious but actually has more interesting (and disappointing) underlying issues.

This single points are the product of about 100 hours (real-time) of concentrating on a goal. And it does take concentration, almost every hour of that 100 hours. So after fairly intense focus to then fall short of plan, while a “useful” experiment, is terribly disappointing. Again, you say, what are you talking about. Well, remember this from last week’s post:

214.4 (2.8lbs loss)

So stay tuned for next week’s report to see how good I did on prediction as well as sticking to my actual regimen.

Well, here’s the sad news:

215.2 (a miss by 0.8!)

Now 0.8 (out of 215.2, i.e. 0.4%) might not seem like much, but for this experiment it’s huge! Put in different terms it is 411 calories/day, either too little exercise or too much eating and that’s a fairly huge adjustment to make.

To put it a different way, look at the extrapolation. I’ve now lost about a week and two pounds by the report two weeks from now.

So what happened and what went wrong?

  1. I decided to go exclusively with BodyMedia’s system to monitor and predict loss (this is the “experiment” since I’m hoping this is my long-term solution to maintenance),
  2. I shifted more of my exercise time over to stationary bicycle instead of treadmill (another long-term plan since bicycle is more joint-friendly and I may need that in the future)
  3. As a corollary of #2 (and due to shoes not fitting pedal clips) a significant portion of my treadmill time was low incline and slow pace
  4.  As a corollary of #1 and more careful calorie consumed record keeping I loosened my eating discipline and allowed myself to go right to the limit the system allows
  5. As a result of both my interest in bread and the false belief that somehow whole grain bread doesn’t have the impact of white bread I ate too much bread
  6. As a result of the notion of more sustainable diet I had breakfast instead of largely fasting all day until dinnertime

It’s difficult to decide which of changes factored (at all or most heavily) into my miss but these are the changes I have to address.

#1. The advantage of the BodyMedia system is measuring (in theory, accurately) my calorie burn rate thus replacing my rather approximate estimates before. Now, while I haven’t done a complete “calibration”, I do know the sensor and [secret] analysis algorithm over-reports “routine” exercise (come on, washing dishes doesn’t taking as many calories as fairly hard treadmill workout) and under-reports actual exercise (elliptical and bicycle don’t tickle the accelerometer into believing I’m taking steps (I can determine this) and I believe (just a guess) the algorithm thus doesn’t count as much). Of course, not taking “steps” (jarring on the joints, vs smoother motions) is exactly the point. But all this indicates that, at least for me and my routine, the claims of 5% accuracy on calories burned just isn’t true.

This is disappointing, not just now, but for long-term. I’ve already encountered the problem (which Nate hasn’t responded to yet) that my scale is woefully inaccurate (its standard deviation is equal to about half week of weight loss) and therefore its “noise” is significantly hiding the signal. This matters, not so much now because while in weight loss mentality I can make up a shortfall, to the long-term since I believe weight maintenance has to be a “stop them at the beach”. IOW, at the first sign of any uptick in weight immediate corrective action must be taken and the gain is immediately reversed. If you let the little upticks just go and assume somehow you’ll make it up that turns into a pattern and little by little you’re going the wrong direction. By the time you realize (or admit) this it will not be hard to correct. So never let the gain get started.

But that’s where the scale error matters (it’s not an accurate way to detect that first small fail) and hopefully the BodyMedia system would solve the problem. I’ll show you how. The ActivityManager provides an export of data into Excel. From that I can compute: a) sum of calories burned (22,773 for this week), b) sum of calories consumed (12,805 this week), c) deficit/3600 (2.97 this week, for < 7 days, just adjust to weekly value). So this actual numbers (and as I measured them during the week) showed me not only on track to achieve 2.8lbs loss but that I’d actually beat it. That gave me false confidence and thus eroded discipline, mostly on the eating side. So according to BodyMedia my total deficit was   10,688 or 1527/day (well over the target in ActivityManager of 1000 and over my adjusted target of 1400).  So according to BodyMedia I should have weighed in at 214.2, not 215.2. That’s a big miss meaning I have to adjust the BodyMedia prediction to 64% of its calculated value, or, put differently, my actual daily deficit was only 977. Now had I known it was only 977 (earlier in the week) I had a couple of opportunities to decrease food intake and ramp up exercise that I didn’t do because I was doing the experiment (btw, I had the feeling I was going to come up short simply because I didn’t feel exhausted/light-headed as much this week as I usually do, so I did some diddling last night and came up with revised prediction of 214.9 (0.7 more than BodyMedia, but still the disappointing 0.3 less than actual (a shock this morning during weigh-ins)).

So my conclusion is that #1 doesn’t work, or that I have to adjust by a “fudge factor” to remove the bias in their calculations.

#2. Now of course any good science experiment isolates the experiment to just one changing parameter, but life is short so I did more than one. We got the recumbent bike (for both of us) to address the issue of aging and needing to put less stress on joints. Fine, it’s done that (also it’s easier to read on bike and thus easier to pass the time). But I could tell it was an easy workout. Now, Dear Reader, given you’ve been following me for 8 months you know I used to bike a lot (in sunny California) and had my weight under control. So biking was sufficient (but unavailable, outside, in this miserable climate). Except for climbing up to Skyline I rarely rode at high intensity, instead cruising along at a sustainable pace. Now when you ride 600 miles/month that’s probably good enough. But for less duration you have to make it up with intensity. On real bike and stationary bike you crank up intensity by RPMs and hill climbing. It might seem RPMs are “neutral” (just go faster) but I believe you have to burn muscles harder to do high RPMs and on exercise bike that seems to be true. So it’s hard for me to push beyond my natural “rhythm” of about 60RPM (70 I can do, but the 90 racers do was always hard for me and would be impossible on exercise bike). So that means hills. The exercise bike certainly does significantly increase resistance as I crank up the levels and I feel it and the machine shows it in the watt ratings (instead of the treadmill’s METs). But about 175watts is the most I can do (and then only for a few minutes now at my present state of conditioning) and furthermore the bike is shaking like crazy at that point (making reading impossible). And assuming heartrate monitor is roughly correct I only go up to about 110 BPM (well below what I do on elliptical where I routinely hit my max (>100% for my age) heartrate). Based on huffing and puffing and sweating, I believe 110 number, but I can’t conceive of cranking up to 160BPM on the bike (elliptical uses more muscles, plus lower body strength and efficiency is the product of 30 years of that type of exercise). So, simply put, the bike isn’t going to cut it for intense enough exercise to do the weight loss I’m aiming at now. Whether it will do “maintenance” remains to be seen, but even with that goal I’m going to have to be a bit more aggressive.

Now one good thing about bicycling is the relatively long distance traveled (real or simulated). For the treadmill I had to create the fiction of hiking the Pacific Crest Trail and transferring stationary distances to trail distances and pretending I really was hiking. I need that sense of accomplishment to push myself and just spending time on exercise equipment is not enough. So having just started the bicycle I returned to one old incentive tactic I used with real biking and that was to simulate real trips. While I rode a lot I only did one bike camping trip, from Ft. Bragg to SanFrancisco (actually just Samuel P. Taylor State Park). On that ride I met some cross-country riders who were like an army, totally organized and regimented, but riding > 100miles/day, every day (I was doing about 50 miles). So while it would never have been practical I always dreamed of riding cross-country (having now seen some riders either here or in Kansas crossing dreary and hot prairies I don’t have that fantasy any more). Of course across the USA isn’t really that far, compared to annual biking, so I changed my fantasy ride to a loop around the perimeter of the country. That was a bit harder then (just using maps) but now I have yet another automated tool (Delorme GPS and mapping system) so now I’m on my way north from Omaha and now past Worthington, Minnesota with next stop at Canadian border and then head west to Seattle. So my “virtual” ride will keep me going for a few years just as my virtual Pacific Crest Trail hike has kept me going for 2.5 years.

So, on biking, I need to keep going (plus push up my maximum daily ride at least back to the old metric century value) but realize I’m not burning enough calories. So pretend my route is hilly and do some intense intervals in every ride interval.

#3. On way exercise biking doesn’t match real-world riding is that road conditions dictate stops every now and then, plus some “coasting” on the pedals (I’m not riding in pelethon on a closed course, after all). OTOH, exercise bike is non-stop and that definitely puts some kinks in my legs since I never get the chance to “stretch” any. So sometimes I’d get off the bike and have rubbery legs, so I started tossing in an interval on the treadmill to “loosen” up. But I can’t wear the same shoes on bike that I use for treadmill so I was doing treadmill barefoot (after biking) or later with just slipons. As a consequence (plus wanting to push up footsteps count in BodyMedia) I was walking at low incline and slow pace. About 20% of my calories on treadmill (which dropped by about 1200 this week) were these “lazy” calories. I think those are worthless (but BodyMedia counted them). So I can work the kinks out of my legs by just actual walking and thus focus my treadmill time on: a) wearing shoes so I skip the light workout, b) only doing workouts long enough to kick up heartrate, and, c) definitely get to higher METs during each interval.

#4. The BodyMedia ActivityManager was giving me the illusion I was doing fine and so I lapsed a bit on food consumption discipline: a) instead of maybe just one day of splurging I had three hi-cal meals last week (was fun, but I felt guilty, but did it anyway); that’s just not possible, one, maybe two, but not three; b) I had “breakfast” every day, again the delusion that somehow whole wheat bread is magical I had a slice with about 8g of jam each day – sorry, can’t do it, plus my theory is that the eat-a-little-often idea (dietitians would push) is terrible, at least for what I’m doing; c) a loaf of bread is sitting there tempting me so I also piled on two “snacks” (yet another slice + some peanut butter), sorry, can’t do it; d) I even had lunch one day (the same day I later had big dinner); e) I treated protein as irrelevant to calorie intake – wrong, those extra bits of flank steak (even though lean and none of the sauce) probably were worth 0.1lb; and, f) the real guilty pleasure (honestly recorded so deluding me it’s ok) of eating 300cal of some candies I thought were out of sight, but I found while looking for bread making ingredients.

In short, all these things added up to less than planned loss and if I’d been in maintenance mode to a disastrous uptick that would be the start of a slippery slope. Maybe any one or two of these backslides might be OK, but all of them was definitely too much. Over long term I can’t make up excess consumption with more exercise so eating discipline is still my long-term challenge. Just recording calories, even accurately and honestly, is not enough (or at least until I can get BodyMedia “fudge factor” accurately determined). If I compare my discipline in first few weeks compared to last week, last week was pretty bad backsliding and it came from cockiness that everything was going so well. So, also don’t let over-confidence become the rationalization for excess consumption – stay vigilant.

#5. I’ll try to get to this is other posts but basically I think most of the nutrition information sucks and is stupid. All these little complicated rationalizations just add up to excuses to eat too much. Life is simple, eat less, exercise more, and THERE IS NO MAGIC! Now as you can see from other posts I’m having fun with making whole grain breads. Fine, as a hobby, it’s just fine and a few of the results have been quite tasty. BUT, whole grains are a myth when it comes to weight loss. Calories are calories. What’s so wonderful about whole wheat flour? I’ve said it before – whole wheat flour is 85% that evil white flour and 15% filler, period! In terms of calories, you might as well take white flour and add 15% dirt, the effect would be the same. So instead of seeing whole grains as magical just eat 15% less white bread if you eat bread at all (and none is better than some whole grain).

Then there is the other claim that all that bran interacts with your intestines and slows the absorption of calories. Key word – SLOWS, not STOPS. For maintaining blood sugar levels, fine, this is useful, but for weight control it is useless (and I’ll argue in next point, bad) to just slow down absorption of calories. Our intestines are efficient enough at extracting carbos that bran doesn’t reduce the absolute amount (and, again, sorry health food nuts, whole grain bread is 85% carbos). As to whether bran reduces cholesterol: a) I don’t care, statins are a med I don’t mind and I know they work, and, b) any “claim” not backed with any actual proof of the mechanism is garbage (the claims about bran ADMIT the mechanism is unknown – sorry, biochemistry matters).

So this week merely proved what I already believed, miracle foods and changing diet is meaningless, only calories matter.

#6. Now I’m going to debunk (by opinion, not facts) yet another nutrition (dietitian best practice) myth. Frequent small meals are the behavioral adjustment you need for health and weight control. The best I can say is, maybe for some people, but not for me (and I’ll claim not for anyone). The old “good breakfast” is bad, even if it’s all oats and fiber. The slow release of nutrients, as well, is all wrong.

Here’s the simple biochemistry. When our blood glucose goes up (seems to only be the glucose response, not triglycerides, which, of course, are larger portion of our caloric intake, except for quinoa nuts) insulin goes up. That signals tissues to do uptake of glucose (and sorta FFAs). It certainly signals adipocytes to do uptake. Now my blubber that shows on the scale is the stored little lipid sacs in all my adipocytes (glycogen in liver is irrelevant in terms of weight, only about 2-3% of stored lipids). Conversely when blood glucose drops glucagon is released (exercise triggers lipid release largely via epinephrine) and the adipocytes enter their cycle to release triglycerides and muscles commence lipolysis. Guess what, weight loss absolutely requires this. Since the number of adipocytes doesn’t decrease (at least short term) the only body weight loss is the depletion of the lipids stored in adipocytes. Now if bran is retarding absorption of glucose in the intestines, guess what, no glucagon (probably no insulin either, since now blood glucose will just be a more steady value, a good thing for diabetics, not good for weight loss).

So glucagon and more glucagon and more glucagon is what I need. I’ve about maxed out on epinephrine (can do a little better than last week, but long-term will actually be lower). And that comes from low (not steady) blood sugar. And for me that comes from NO calories for as much of the day as I can do. So no breakfast, no snacks, no lunch. Whole grains are counter-productive and only slightly reduce calories any way. So I have to spend most of the hours of my day in fasting mode – PERIOD! No (or at least few) exceptions. Don’t be fooled by the bogus nutritional info that is more based on politically correct leftwingnut hippie-dippy “natural” ideas and demand to see the biochemistry. Hey, food nazi – show me the glucagon before you start peddling any diet. Now I suspect that is what Atkins was all about, but it fails because there is some other mechanism on fat (probably the chylomicrons are the signal, not a pancreatic hormone.)

I’ll explore this more over time, esp. at 2/3rds of my day spent fasting is probably not practical, but I’ve got to find nutrition information based on science, not fads.

Changes to make:

  1. Don’t sync the sensor during the day so that as I record calories consumed my deficit steadily decreases so I won’t believe I have a “buffer” and therefore decide to eat a little more. Make myself feel I’m doing poorly on the deficit so I try harder.
  2. Mentally adjust the calories consumed (as recorded) upwards by 10% to account for systematic under-estimation (or missing something)
  3. Mentally adjust down the calories burned (as sensor and algorithm claim) by 12%.
  4. Using 2&3 and delaying until next day determine the “true” deficit and determine if it is cumulatively on track for the week
  5. In every interval done on the bike, make sure to do at least some profile achieving at least level 10
  6. In intervals on treadmill: a) do at least 15 minutes (less seems worthless), and, b) hit > 6Mets at least for 25% of the interval.
  7. No food in the morning, or before 2pm, or until after at least 500 calories burned in intense exercise
  8. No gimmicks, don’t let any sneaky little thought enter my head and allow me to believe it is OK to put any hi-cal (or significant quantity) of anything in my mouth.

Finally, this huge post has undoubtedly lost any possible reader, but any reader who just fast forwards to the end should realize that at this point my life all this weight loss is my life (even skipped some geodashing just to do more exercise). My only shot at having a life worth living is that nearly extreme weight loss (maybe even below 185) and staying there will stave off unacceptable meds and regimens (actually having to do what dietitians say). Life on a diabetic diet is not worth living, it’s just too depressing. What’s the point of decades more breathing when I’m not really living. So either this works or to hell with it all and I go into live-fast-and-die-young mode.

About dmill96

old fat (but now getting trim and fit) guy, who used to create software in Silicon Valley (almost before it was called that), who used to go backpacking and bicycling and cross-country skiing and now geodashes, drives AWD in Wyoming, takes pictures, and writes long blog posts and does xizquvjyk.
This entry was posted in musing, prediction, whine and tagged , , . Bookmark the permalink.

4 Responses to Applied Nate Silver – Failed experiment: A swing-and-miss is still a miss

  1. dmill96 says:

    Note: I just adjusted my calories burned to account for perceived low intensity and this explains about 80% of my reduction in loss this week and that seems about right. So excess eating accounts for only about 20% and, on average, that seems about right (this week there was no big binge (like there will be next week) but just a bit for four out of seven days – don’t do it, tighter on the good days, less splurge on the bad days, just tighten a bit, at least as long as I have the option of pushing exercise harder).

  2. Pingback: Sunday QuickReaction | dailydouq

  3. dmill96 says:

    Just did 48 minutes on stationary bike and BodyMedia essentially doesn’t show that at all. Clearly I must either do steps or get to very high body temperature. OTOH, my session on bike was very “easy” so perhaps it just doesn’t register much, but it’s hard to believe that steady 50RPM pedaling at lvl7 is the same as watching TV and only the brief stretch-my-legs walks registered as workout.

    As a result, despite doing better on food my calorie will be one of my lowest and therefore my calorie deficit just barely clocks in at 1000+ (maybe, not there yet). But I suspect the 400+ less calories this Sunday is more important than any exercise I did last Sunday after the BK pigout.

    But, nominally off to a bad start for the week, on another 2.0lb/week drop, not 2.8.

  4. dmill96 says:

    It appears I didn’t make prediction last week so here’s what I have the week now finished (Note: several of calorie consumed dinners are pure WAG and I’d be lucky if they’re within 10%; And, more of the treadmill was hard, but biking still registers almost nothing on BodyMedia so let’s hope the calories are there are real which actually might help some)

    So the raw estimated (calories-burned – calories-consumed) would be: 2.6lbs/loss

    which seems very unlikely to me so I adjusted to 2/3rds and think this will be far closer: 1.7 (213.5). I think it could be as bad as just 1lb/lost but I sure hope I break into 213.x range.

    Even then it’s very unlikely I’d make it under 210 by next Sunday, which was my implicit goal.

    I’m definitely slowing down, eating a little more, exercising less. I’m also feeling hunger more (could be a good thing, sign of reduced blood chemistry) but I hate the idea of getting down to < 2lbs/week since I work so hard at this to get so little and I know that will discourage me.

    So here's hoping for good results tomorrow.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s