sharetrader
Page 377 of 465 FirstFirst ... 277327367373374375376377378379380381387427 ... LastLast
Results 3,761 to 3,770 of 4649

Thread: Harmoney

  1. #3761
    yeah, nah
    Join Date
    Mar 2017
    Posts
    491

    Default

    I meant to add that the tiny peak at the 36 month point, where a couple (...) of loans actually get to is, well, tiny...

    Just thinking about the Payment Protect numbers - possibly not meaningful as they include a large portion of loans when PP was not available? Could probably redo this for loans after whenever PP started...

    Quick look at the data shows that the large 'jump' in defaults was due to a heap of loans debt being sold off early February 2018.

    The 'Enquiries (6 months)' chart has a bar with no label (far right - count of 404), I'm guessing that this info was not recorded in early days, so was just blank - I'd suggest just ignoring this bar.

    Everyones loan set is different, so these numbers can only be considered a guide, nothing more.
    Last edited by myles; 07-10-2018 at 11:07 AM. Reason: Adding thoughts...

  2. #3762
    Member
    Join Date
    Sep 2012
    Location
    christchurch
    Posts
    386

    Default Thanks Myles!!

    Fantastic work Myles. Many thanks for the effort. Thanks too to the investor for sharing.

    One quick conclusion is that it is much better to invest in rewrites. None of my almost 400 defaults were rewrites too.
    Last edited by Cool Bear; 07-10-2018 at 09:29 AM.

  3. #3763
    Member
    Join Date
    Jan 2018
    Posts
    48

    Default

    Quote Originally Posted by Cool Bear View Post
    Fantastic work Myles. Many thanks for the effort. Thanks too to the investor for sharing.

    One quick conclusion is that it is much better to invest in rewrites. None of my almost 400 defaults were rewrites too.
    Quite a lot to digest in there- you're right though Cool Bear the Re-Writes is absolutely staggering.

    Another that stood out to me was Home Improvement loans defaulting at just a tick over 1%.

    One of the selection criteria that I use, is I will only go into a Home Improvement loan if the residential status "Fully Owned" or "Paying Mortgage". Would you mind plotting this Myles?

    y-axis = number of Home Improvement Defaults
    x-axis = Residential status

    Huge thanks to both the investor & Myles. Fantastic effort!

  4. #3764
    yeah, nah
    Join Date
    Mar 2017
    Posts
    491

    Default

    Quote Originally Posted by alundracloud View Post
    y-axis = number of Home Improvement Defaults
    x-axis = Residential status
    I think this is what you want? You're on the money - looks like most are mortgagors anyways.

    defhome.jpg

  5. #3765
    Member
    Join Date
    Jan 2018
    Posts
    48

    Default

    Exactly what I was after, and looks similar to how I expected. Thank-you!
    I have some data which you might be interested in also, how best to get it to you?

  6. #3766
    Member
    Join Date
    Sep 2012
    Location
    christchurch
    Posts
    386

    Default

    Quote Originally Posted by myles View Post
    I think this is what you want? You're on the money - looks like most are mortgagors anyways.

    defhome.jpg
    Thanks again.

    The single default in the "boarding" resulting in the 6.25% (1 out of 16) could be an anomaly due to the low population size. Similarly too for the "Living with Parents" and "others". However, when you combine them, it is a total of 10 defaults out of 88 or 11.36%. And 88 is a credible population size. Even if you add the 2 "unknowns" and 8 "supplied by Employers", it is still a high 10.1% compared to 0% for fully owned, 0.5ish% for mortgagees and even the 'high-ish' 3+% for renters.

    Valuable confirmation, Myles.
    Last edited by Cool Bear; 07-10-2018 at 01:31 PM. Reason: clarity

  7. #3767
    Senior Member
    Join Date
    Apr 2002
    Location
    , , New Zealand.
    Posts
    726

    Default

    Quote Originally Posted by myles View Post
    Below is another data set from a long term Harmoney Lender - total loans is just over 10,000. Many thanks for sharing
    Many thanks Myles and InTheRearWithTheGear for sharing. Interesting to see your lending volume up substantially in the last six months. Confidence from past lending results and learning, no doubt.

    Thanks also Cool Bear for your insights
    Last edited by beacon; 08-10-2018 at 09:36 AM. Reason: Additional thoughts

  8. #3768
    yeah, nah
    Join Date
    Mar 2017
    Posts
    491

    Default

    There are a few people interested in pooling data and I've got most of what needs to be done to merge and clean the data. I'll not include any data I currently have so if you have given me some previously can you please generate a new report and upload that.

    The good thing is that the data in the orders-report-###.csv file provided, is free of lender info - so it will be safe to upload the csv and not have the data associated with the lender. However, the name of the csv file does include an ID number - for anonymity it will be best to rename the file to just orders.csv.

    Please generate a new report, rename the file to orders.csv and submit that as it is (don't zip or anything else).

    When asked for a name and email address to upload the file just use a fake one, I typically pick on poor old fred:

    fred.png

    I'm just using a Drop Box file request, which is secure and safe:

    Upload orders.csv


    There are some issues with timeliness of data i.e. it becomes 'stale' if not updated with any new details of defaults. So I'll only keep this link live for the rest of this week (late Sunday) and then shut it down.

    Then there is the question of what to do with the data. I can clean it up and make it available for those who want to do their own analysis and perhaps come up with one 'tidy' set of charts and summary data for those who don't/can't. Some thought needs to be put in around older loans and the 9 month default gap which effects the 'real' values - I'll think about that a bit more.

    The 'mechanics' of doing it aren't a problem I don't think i.e. once I've merged and cleaned the data I'll just provide a link to download the compiled data as a csv in the same format as it currently is.

    If anyone has suggestions/thoughts please jump in.

  9. #3769
    Member
    Join Date
    Apr 2016
    Posts
    83

    Default

    You may wish to strip borrower information from it as well - use your own key.

    ie LAI-00140637 becomes some sort of new id tracked across mutliple data sources - that way loan is not identified.

    But At the end of the day, we cant beat what they provide which is the loan grade as a loan risk, this loan grade takes into account all the hidden data points ie credit score.

    You would be wrong to associate risk with a individual loan dimenson such as "board" or "living with parents"

    A better way to visualise things use kmeans clustering on each dimension - there will be python tools to do it -

    And then you could spin your graphs by dimension

    https://www.youtube.com/watch?v=9qnMHAZaD7k

    https://www.youtube.com/watch?v=l98GVNwdTks

    is an example.

    Its kinda interesting stuff and food for thought.

    (you would also need a loan from harmony for the pc specs to work it out)


    We should call your data dump idea the #TheHarmoning you know after the #thefappening
    Last edited by IntheRearWithTheGear; 09-10-2018 at 08:47 AM. Reason: too cold, cant afford heating

  10. #3770
    Senior Member
    Join Date
    Apr 2002
    Location
    , , New Zealand.
    Posts
    726

    Default

    Quote Originally Posted by myles View Post
    If anyone has suggestions/thoughts please jump in.

    Neat idea. Thanks for making the time to do this, Myles. I have some pooled data too, which I am happy to upload to help with this initiative.


    Data labels on bar charts can be very helpful. Eg., in plotting default by coborrower, you currently include sample sizes on x-axis (1079 co-borrowers and 9909 single borrower). Both bars are between 0 to 5%. Could you also include labels on top of the bars like 3% or better still 32, if including them is no trouble to you.


    Protect loans might be better compared against non-protect loans (control) after the date protect loans became available, to be more truly representative. So if I were plotting these, I'd begin counting the control sample from the day protect sample began. This means control sample will be a subset of Harmoney loan population and control population will be less than Harmoney total population. Of course, the result may not not be too different.

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •