Exposing source of logged activity and detecting 'cheating'

[keeth]

Hello Fitbit folks,

Our app is based around group challenges and rewards, so we have a strong preference for tracker data over manually logged data.

With the recent arrival of the GPS tracking feature we've seen an increase in logged activity, which results in an increase in user confusion when the tracker-only total in our app doesn't match the tracker/logged combination total in the Fitbit app.

I guess I'm wondering a couple things, 

1. Would it be possible to expose the source of an activity in the API?  i.e. if we knew that an activity was generated by Fitbit GPS, or by another first-class app, we'd more likely trust it than a purely manual log.

2. I'm curious if/how others have dealt with this problem.  Our biggest source of user frustration is when their step counts don't match between our app and Fitbit's.  So even though it's important for us to accurately validate steps, simply eliminating logged steps may cause more confusion than it's worth and we may need to come up with something more sophisticated, like allowing for some amount of logged steps while trying to detect cases of flagrant cheating (which are sure to be very rare, but even a small amount of cheating could have a large impact on our system).

Thanks!

Keith

[JeremiahFitbit · Accepted solution]

At this time, the Fitbit API does not have a way of identifying the source of non-Fitbit tracker activity, though this is something we are considering.

In your particular use case, I would suggest only trusting steps from the tracker. This can be done by looking at the activities/tracker/* time series.

That said, there are many ways to cheat such a system and it's very difficult to prevent such methods. Fitbit trackers are very good at detecting real human movement, but they are liberal in what they accept because they assume that the user isn't cheating themselves.

[keeth]

Thanks Jeremiah,

We are considering reverting our change to use the activities/tracker/* time series... users get too confused/upset when the steps don't match the Fitbit app. We'd just have to assume that most people are honest, and be on the lookout for logged step counts that seem way out of line with tracker steps.

Keith

[tray123]

I'm starting to get very suspicious of a few people at work, so much so that I'm considering talking to them about it.  I'm really upset that I'm working my butt off trying to meet my goals while others, I feel are making a mockery of this challenge.  I've been keeping track of their numbers and they just don't match up with the time one would need to get so numbers. 

Would it be wrong to ask them to take a run with me?  I'm sure they would have an excuse to keep themselves from being exposed.

Ugg!!!!

[fitfiz]

I totally agree with the cheating thing.. I have been in a few challenges and some people seem to just jump 10000 steps every now and then.. I wondered how.. Then watched a youtube vid explaining how to log manual steps.. Surely manually logged steps shouldn't be taken into account on the challenges due to the unfairness.. Anyone can just log 100000 steps..

[Doug]

---

@fitfiz wrote:

I totally agree with the cheating thing.. I have been in a few challenges and some people seem to just jump 10000 steps every now and then.. I wondered how.. Then watched a youtube vid explaining how to log manual steps.. Surely manually logged steps shouldn't be taken into account on the challenges due to the unfairness.. Anyone can just log 100000 steps..

---

Jumping 10,000 steps every now and then is very easy to explain.  Jumping 100,000 is harder but the same principle applies (if legitimate).

All it takes to jump 10k steps at a time is to turn off "All day sync" on your mobile device and not get within range of a USB dongle during the time.  When I power walk, I get about 8500 steps per hour (I will set a metronome app at 142 beats a minute) so I'd hit 10k steps in 71 minutes.  By disabling All Day Sync -- or not even having a phone with you -- it is very easy to increase 10k "every now and then".  

People who jog can get the 10k steps in an hour -- that only takes 167 steps a minute and many people pace between 170 and 180 steps a minute.  So unless I know the person and their capabilities (or lack of them <g>), I wouldn't consider 10k step jumps suspicious.

I've been in challenges where people, by their own admission, are intentionally turning off sync and removing their USB dongle until close to midnight.  Then during the day using web profile pages to try to see where others are at, if they have not also disabled syncing.

But here is what you can do, if you don't want to confront them.  At the start of a challenge, visit their user profile web page and make a note of their lifetime step count.  Compare to their lifetime count at the end of the challenge.  Unlike challenges and leaderboards, daily and lifetime badges and counts are based only on tracker accumulated data.  Manual logging is excluded from the counts.

If you are all in the same time zone, then even better is just noting the number of profile steps in each day of the challenge.  Since challenges are based on the time zone of the creator, this can't be compared directly using daily profile counts.  And you have to be careful of the online lifetime counts you see near the start and end times of a challenge.  They are based on the values current at the last time the user synced, which is not necessarily close to the time you are checking.

But it should give you a good idea.

In terms of manually logged steps being unfair in a challenge, consider the reverse.  If someone is wearing a Fitbit during a high vibration activity like riding a lawn tractor, off-road driving, some motorcyles, etc then you will get false steps and manual logs are the proper way to negate those for that time period.

[JeremiahFitbit]

The activities/tracker/steps time series reports steps only from Fitbit devices and MobileTrack.

See https://wiki.fitbit.com/display/API/API-Get-Time-Series