Newbee: So I'm going to derail for 0.0011?

The chart says i’m derailing for 0.0011? Sounds a bit silly :slight_smile:
https://www.beeminder.com/apolyton/goals/bodyfat

2 Likes

Well, that is how Beeminder works :slight_smile: However, seeing that you appear to have started this week, your initial goal must have been either very unrealistic or maybe a misunderstanding of how to set it up properly? My feeling though is that if you explain whatever the issue is in the legitimacy check mail you’ll get after derailing, the support folks will help you, especially at this stage with the service.

4 Likes

Well, I had the extra days when I started but then I decided to live dangerously and try the retroratchet feature. I guess I am bit fuzzy about how the calculations are done. Live and learn…

2 Likes

Hey @apolyton, we’ll definitely undo the derailment and/or the retroratcheting if there was any confusion at all about how that stuff works. I don’t believe in voiding it due to the closeness of 0.0011 to zero though. (: Bright lines are a fundamental tenet of Beeminder! Imagine if we did call that close enough, then we’d have to define somewhere what the criterion for “close enough” was. It would add confusion, complexity, and fuzziness. It’s an instance Beeminder’s no-free-lunch theorem – namely, that any attempt to make Beeminder more lenient necessarily backfires and makes it harder overall to stay on track towards goals.

6 Likes

Hi @dreev, i agree with the concept, i’m not really complaining about the 0.0011 thing.

Mostly I’m curious about what happens and how the algorithm works. To be exact:
My measurement yesterday was 30.2% (it’s a body fat goal) and the mail sent after that said "You are 0.21 above the centerline and 0.028 below top edge."
That would mean that today’s measurement of 30.5 would be ok. Instead, i got the 0.0011 difference.
I tried to “fix” it by setting another value through Fitbit but i guess the system remembers the first value of the day even if you delete it (which I suppose is a good thing to avoid cheating and not a bug).

Anyway, i don’t mind derailing and adding a pledge, 30.5 was my actual measurement today. I just wouldn’t like to derail due to some rounding up somewhere in the formula.

1 Like

I do think it’s possible we have a subtle failure of conservative rounding in some cases. Like if you have 0.500003 left to go and it needs to be rounded, it should tell you “0.51”. Rounding to “0.5” is technically more accurate but it’s telling you how much you still need to do so if it says 0.5 and you do it it shouldn’t tell you you still have 0.000003 to go!

(Thanks so much for alerting us to possible cases of this, btw! Definitely important to get this right.)

But in this case it sounds correct. If you were 30.2 yesterday and 0.28 below the top edge (in a weight-loss-style graph the top edge is the critical one) means a hard cap of 30.48 which means 30.5 should be too high. (Also the road is sloping down so the next day’s limit is slightly less.)

PS: I actually don’t think bodyfat is the best thing to beemind, despite being fundamentally the metric you most care about, simply because it’s so hard to measure. More in our “Why Weigh (Daily)?” blog post. (Hover over link for relevant excerpt.)

2 Likes

I don’t think you can really measure body fat accurately without something really sophisticated like a Bod Pod: http://www.bodpod.com/en/products/body-composition/adult-children-bod-pod-gs/bod-pod

Short of that (or something similar), I wouldn’t try to use it as a metric in Beeminder.

1 Like

Thank you for the replies and clarifications.
Perhaps it would be better to make the reminder idiot proof (as in my case) and give the actual barriers (e.g. 30.48) instead of “0.28 from the top edge”.

As for fat as a metric, i’ve been monitoring it for some time now so i know it’s not accurate science. Still, i need to focus on it as i’ve been mostly caring about weight so far.

1 Like

If you care about weight why not just track weight?

I’ve heard that it is perfectly fine (and probably better than weight) to measure bodyfat as long as you only look at the relative changes, disregard absolutes and make sure to use the same measuring equipment.

1 Like

@larsivi yes, this is what i’m doing. I’m only measuring on my own scale (Omron BF511), on the same hour every day (morning, after getting up and before drinking/eating anything)

1 Like

Really appreciate this feedback! It’s slightly easier said than done because for some goal types it’s only the deltas that matter and the absolute number is just confusing. That’s probably only when auto-summing is in effect though, so… you’re right! Thanks again!

One more tip on beeminding bodyfat (if blog.beeminder.com/weighly didn’t convince you to stick to weight, though this tip applies to weight too – just bodyfat even more so):

Bodyfat is an outcome rather than an action. So keep it dialed very conservatively (probably don’t retroratchet (and let us know and we can undo the retroratchet you did – or wait for our upcoming road editor) and set a pledge cap in the goal settings to something that serves as motivation without being too frustrating if a derailment happens that feels like it was a random fluctuation out of your control. And of course add other Beeminder goals that are in your direct control and keep the pledge cap high on those.

2 Likes

Thank you @dreev, if you can take back the retrorachet it would be good to have a few more days to see how the formula works https://www.beeminder.com/apolyton/goals/bodyfat

1 Like

Done, I just made it go gradually to 20% bodyfat at a rate you’ve been achieving so far. Now I’m super curious what that perfectly linearly increasing period around the holidays was!

1 Like

LOL @dreev that’s just Fitbit auto-filling data when the user is not entering anything. I’ve left for christmas on Dec 17th and since I only weigh on my own scale, the next data point that i entered on Fitbit’s system is Jan 5. But I guess their API auto-fills the rest when you import the data.

The difference (3% up) between Dec/17 and Jan/5 is the “mama cooking effect” :smiley:

3 Likes