Tolerancing bug is fixed on iOS and released to Beta

Discussion in 'Official V1Driver Discussion Section' started by mswlogo, Jan 20, 2018.

  1. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    This is a pretty bad bug, sorry. Worst that would have happened is a few lockouts stopped working. The code that handles frequency drift was broken.

    New code works much better. Obviously I didn’t test it enough. I did the statistics work at one time and the drift work at another time and the two were not compatible.

    It will automatically do some repairs to your database on the next “promotion”. This includes resetting all error statistics. You should do a backup before installing for good practice. I could have skipped the repairs and it would have self corrected but taken a long time and generate a lot of confusion.

    Promotion is when I process all the “visiting” for unlearning. This happens once a day. And only does it upon a connection.

    I also now display that “promotion” time in debug settings so you can have some idea when it will or has last done it.

    I will probably release this soon to the App Store because of the severity of the bug. Probably Monday morning.

    If you are super curious you can set DebugMask to 16 under Debug Settings and it will display all frequencies collected. They will be wrong until it applies the repair.

    This does not affect Android since that feature is not released yet. Probably release Android late tomorrow. Still yet more Bluetooth work.
     
    WJS, 918dad, Deacon and 1 other person like this.
  2. 918dad

    918dad Learning to Fly Beginner User

    Messages:
    60
    Likes Received:
    12
    Country:
    United States
    Joined:
    Dec 30, 2017
    What are strength and ratio measuring?


    Posted from my iPhone using the RDF Mobile App!
     
  3. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    I knew that was coming ;)

    Strength of the signal 1 to 8 (float)

    Ratio is the ratio of the strength of the front antenna relative to the rear.

    1.0 is 100% front, which never happens in the real world.
    -1.0 is 100% rear, which also never happens.

    0.0 is exactly equal front and back.

    Side is really somewhere around 0.2 to -0.2

    The V1 is NOT really front, side, rear. That is a limitation/simplication of the V1 physical display.

    Nor are strengths just 8 levels. There actually 255.

    V1Driver takes full advantage of this and the reason for the “bogey” map display existing. And you’ll see the strength on the V1Driver display isn’t limited to 8 levels.

    These 2 values are basically the raw information from the V1.
     
    cihkal likes this.
  4. 918dad

    918dad Learning to Fly Beginner User

    Messages:
    60
    Likes Received:
    12
    Country:
    United States
    Joined:
    Dec 30, 2017
    I thought that might be what they were for, very cool.

    All of my gpsMuteTol are gone with the new beta. They're just plain gpsMutes now. They're saying stuff like 18 hits but only two samples.

    Is that what you expected to occur? IMG_8044.JPG


    Posted from my iPhone using the RDF Mobile App!
     
  5. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    Yes, that is correct. I'm surprised you had any gpsMuteTol so quick. How long have you been running V1Driver? and did you change any Settings around this from the defaults?
     
  6. 918dad

    918dad Learning to Fly Beginner User

    Messages:
    60
    Likes Received:
    12
    Country:
    United States
    Joined:
    Dec 30, 2017
    I've been running it since around 12-22-17 I think.

    I've been a little bit of a nerd and driving at night in areas I haven't run as much in town to try and get my numbers up.

    I've tinkered with the numbers but I've set it back to 14 for statistics and I changed my "day" from 23 hrs to 18 hrs.

    Posted from my iPhone using the RDF Mobile App!
     
  7. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    You can lower the stats to take a peak on what pins are close, but don’t leave it low. You need good stats to calculate the tolerance accurately.
     
  8. 918dad

    918dad Learning to Fly Beginner User

    Messages:
    60
    Likes Received:
    12
    Country:
    United States
    Joined:
    Dec 30, 2017
    Yep. I set it back to 14 after looking at it the first day.


    Posted from my iPhone using the RDF Mobile App!
     
  9. 918dad

    918dad Learning to Fly Beginner User

    Messages:
    60
    Likes Received:
    12
    Country:
    United States
    Joined:
    Dec 30, 2017
    Are my old hits still valid or is it redoing all the hits since the update?


    Posted from my iPhone using the RDF Mobile App!
     
  10. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    WYSIWYG - What You See Is What You Get.

    The Hits you have should be good and are being used.

    It is constantly self adjusting.
    [doublepost=1516721007,1516604847][/doublepost]FYI. The fixed version for dynamic tolerancing is in the App Store.
     
    Deacon likes this.
  11. HeartDoctor

    HeartDoctor PSL + 5 never more;) Intermediate User

    Messages:
    230
    Likes Received:
    152
    Country:
    United States
    Joined:
    May 30, 2016
    Location:
    SE Florida
    Installed new release, linked to V1 to get promotions. All GPStol pins except 1 gone. Pins show ratio and strength as "not set" which I presume will change as new data is added. Out of curiosity, set Debug to 16 to view the 1 GPStol pin. I think I've found a bug in the new release in that after viewing, I had only 2 choices, "Delete" or "change to whiteArea". "Done" was not there (probably crowded out of the box). To avoid taking either of the 2 choices, I killed the app. It may be that this problem only occurs with a specific number of data points or perhaps with more than a certain number of points. BTW Ka pickers now work correctly.
     
    Last edited: Jan 23, 2018
  12. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    The pins you looked at for ratio, might be old. That was added later. Filter map on newer pins and see if it shows on any pin.
    I forget if I update that field or not. So it might have to be fresh pin to have that.

    What phone you on?

    You can scroll that dialog if it gets chopped off.
     
  13. HeartDoctor

    HeartDoctor PSL + 5 never more;) Intermediate User

    Messages:
    230
    Likes Received:
    152
    Country:
    United States
    Joined:
    May 30, 2016
    Location:
    SE Florida
  14. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    Can’t view attachment.

    Both the top area and button area are scrollable.
     
  15. HeartDoctor

    HeartDoctor PSL + 5 never more;) Intermediate User

    Messages:
    230
    Likes Received:
    152
    Country:
    United States
    Joined:
    May 30, 2016
    Location:
    SE Florida
    D5214AB2-C610-4489-A766-EA2D10ABE519.
    [doublepost=1516734679,1516734552][/doublepost]Scrolling works. I checked recent pins and found strength and ratio not set. None of the pins are since software update today but are as new as 2 days ago. Phone is iPhone 6.
     
    Last edited: Jan 23, 2018
  16. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    The button area scrolls. Flick it up.

    I also don’t worry how “user friendly” “debug” features are. Which is why it’s under debug.
     
    HeartDoctor likes this.
  17. HeartDoctor

    HeartDoctor PSL + 5 never more;) Intermediate User

    Messages:
    230
    Likes Received:
    152
    Country:
    United States
    Joined:
    May 30, 2016
    Location:
    SE Florida
    Looking at the above data it seems the tolerance is too wide. All data is within +/- .015 and I would expect tolerance to be no more than that. It was set to 2 SD’s prior to update which reset it to 3 SD’s.
     
  18. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    First, that data is no good, but the calculation of Std Deviations from that data should be correct and didn't change.
    I didn't clear out the error history on demoted pins, probably should have, but not a huge deal, it will still self correct if that pin gets promoted again.
    I suspect when correct values on active pins are accumulated the "wiggle" will be a lot less than what you see there.
    You don't have to wait until 14 samples to take a peek now with the Debug hook how stable they are.

    Keep in mind that V1Driver doesn't use +/- any where. It's tolerance is always the total width (not +/- half of it)

    So you say it should be 0.030 and V1Driver calculated from those numbers it should be 0.060 (WITH 3 standard deviations of certainty).

    With only 14 samples (which is a fairly small set), what is the PROBABILITY of the next hit being within +/- 0.15 (0.30)?

    This is how you calculate that probabilty. 68–95–99.7 rule - Wikipedia

    If you want a 99.7% probability that it will land within some tolerance that tolerance needs to be 3 Standard Deviations.
    If you drop to 2 standard Deviations then it's a 95% chance (which is probably close to your gut feeling of +/- 0.15).

    We had this discussion before and you dropped it to 2 Standard Deviations (without seeing the data back then), and you'd probably be happy with that value again (try it of you want).
    But I reset Standard Deviations because you based it on incorrectly collected data.

    I'm happy folks are looking at this closely and cross checking me, but I don't have time to punch your numbers in to Excel to double check what V1Driver came up with (it's pretty easy to do in Excel).
    I'd be thrilled if someone wants to do that, hint hint.

    It's not a simple, does it fall into the Min and Max. It only takes 1 Outlier to be outside the 3 Std Deviations for it to create a new Pin.
    We really don't want that to happen because it will create a new Pin centered on that new outlier frequency and start at the "Full Tolerance" and increase the total width locked out (until it reaches enough statistics to narrow it down again).
    So we want to be fairly certain it stays on the correct PIN. If it's outside 3 Standard Deviations there is very good chance it is something else and rightfully creates a new pin.
    [doublepost=1516770158,1516755830][/doublepost]Also don't forget, if the stdev calculation is larger than the global one it won't make it any worse (wider). It can only improve things (narrower).
    My hunch is most pins will be significantly narrower. But my hunch has yet to be proven because of the bug and we are basically starting all over with this "feature".
     

    Attached Files:

    Last edited: Jan 25, 2018
    cihkal and HeartDoctor like this.
  19. Rabbit212

    Rabbit212 Learning to Drive General User

    Messages:
    11
    Likes Received:
    17
    Country:
    United States
    Joined:
    Jan 30, 2018
    :confused::confused::confused: you lost me at tolerancing.
     
  20. mswlogo

    mswlogo Premium Member Advanced User Premium Member

    Messages:
    2,754
    Likes Received:
    2,395
    Country:
    United States
    Joined:
    Nov 1, 2014
    Location:
    MA
    Think of it like this.

    Rather than use ONE tolerance for every pin.

    It sets a tolerance on each pin based on how much the frequency wiggles on subsequent days on each pin.

    If it wiggles a lot, it sets it wide.
    If it wiggles a little it sets it narrow.

    The narrower, the better (smaller chunk of the k band is locked out). Reducing the risk of locking out a Leo. Which is still possible.

    It will also self adjust if the frequency gradually drifts over time (rather than create a new pin). That’s what I broke in the earlier release.
     
    Gowski, HeartDoctor and Blizzard like this.