Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Joleran

Pages: 1 [2] 3 4 ... 11
Investors - LC / Re: Lending Club witch hunt has started (automation)
« on: September 17, 2014, 12:07:58 AM »
- I then launched 3x the threads under different names, after paying the people a pittance for their ID and credit history.  And STILL AM.  I only mention this so LC knows how frigging pointless their BS is.

The number of browse's was variable.  Certainly FAR less than 1 per 15 secs.  But guess what, I log in and manually do searches and I get no nasty calls.

Something doesn't make sense here.  If you're an experienced programmer, you can create a program that visits the webpage in the exact same manner as a user. LC can obfuscate this and make it difficult to track down exactly what they expect of a normal browser user, but it can be replicated without question.  What should happen is a different manner entirely of course, but either you've screwed up something fundamentally that allows them to differentiate between screen scraping and normal browser usage, or you've been experimenting with the brown acid.

Investors - LC / Re: Loan Information Asymmetrical Timing
« on: September 16, 2014, 11:59:08 PM »
The obvious problem is that as soon as the asymmetry goes away, manual buyers suddenly get fucked as hard as the API buyers are today, but worse and in reverse.  Today, a person frantically f5'ing the note listings can nab a good deal of high-desirability notes (that are funded in <5 seconds of listing) and filter through them at their leisure.  Tomorrow, this person will be 100ms behind a computer controlled program and will get nothing.

Did that guy say Monte Carlo?  Isn't that a car?!

Seriously?  It's a casino in Vegas.

Investors - LC / Re: Auto sell?
« on: August 21, 2014, 11:15:23 PM »
Where does all this magic stuff come from?

Right click -> View Page Source

You're going to have to learn how to do other things like setting HTTP headers for an HTTP request if you want to automate this.  Every few months LC will make a breaking change to their code and you'll need to dig back in and adjust.

Investors - LC / Re: Worrisome autopick
« on: August 12, 2014, 04:49:32 PM »
I think that in a marriage type situation, you're going to have mostly joint debt anyway, so having joint income reported makes sense to me.

Investors - LC / Spamming credit score updates
« on: August 05, 2014, 10:14:07 PM »
The score updated 21 times last month for this note:

Someone at LC spamming the refresh button for this note or something?

Investors - LC / Re: Blank Collection Log
« on: July 30, 2014, 01:25:29 PM »
Looks like someone at LC reads these forums and updated the note.

Investors - LC / Re: Loss / Interest Earned percentage.
« on: July 11, 2014, 10:32:21 PM »
Options come to my mind include but aren't limited to investing in higher grade notes that may be less likely to default, hone an investment strategy that historically has reduced the % of charge offs that occur, and develop a selling strategy to move the note to someone else.  Possibly a combination of all 3 may be the most effective.  I don't plan on investing in the higher grade notes and I've got my investment filter pretty respectable so perhaps I need to look at a selling strategy.  I'm curious for people who typically do 'limited' selling, do you find it improves your position or the scenario I'm thinking of, you sell more loans at a small to moderate loss but holding them all and having only a handful of them default would come out on top.  Make sense?

Well, the question is, are you mentally challenged or tax challenged?  That is, are you going to irrationally turn down a higher rate of return because it also involves more defaults, or do you shun defaults because you're not able to take advantage of (what is commonly held to be the "correct" way of deducting losses) capital losses, resulting in an actual lower rate of return?

I fully believe in selling notes that look to be going bad, preferably before they actually have a loss, but I'll sell 31+ late for a dollar or two after they've missed two payments in a row too.

Maybe I'm missing something, but I don't think you can get credit score detail for every month of the loan from any bulk CSV file - the only way to get it is from screen scraping the foliofn note detail page:

Just compare each measure independently to defaults and see if one stands out.  You could try to do something more complicated with multiple drops, but probably not worth the complexity.

Perhaps.  Even getting the data is pretty rough.  I don't think LC would take kindly to me requesting the note detail page of every note ever issued, but I guess I could do it over a period of days or even weeks.

I assumed PeerCube used a proper database. I was wondering if any individual investors had gone to this trouble or if everyone was conducting their analysis in Excel or third party tools? I have upload and used your site.. thank you.

I do actually have a program that analyzes the drops for me, and it looks at several different types of drops - overall (start -> now), trend (trailing X months), and cliff (last month).  That said, deciding whether to sell these notes, and for how much, is entirely manually done by me.  This equates to selling 20-40 or so notes each month out of ~1000 held, and deciding to hold around 10 notes identified as problematic by the program.

The component I'm missing is good risk analysis, so instead of going off of gut feeling, I would be able to say "well, the FICO dropped 10 points 3 months in a row, and historically this led to default 40% of the time so I'm selling it immediately", or maybe the default rate for this behavior was only 10% more than the population as a whole and dropping it would be premature. 

There's no way to tell without analyzing the credit score curves over time, which is computationally difficult.  For example, I can't just look at defaults, each one of which may exhibit multiple types of drops and multiple examples of the same sort of drop, but I have to look at the credit score curves of all notes to find notes where the same type of behavior didn't lead to a default.  It's frankly a task for a small team of risk analysts with a statistical background, so it's more of a pipe dream at this point.

To put it another way - I don't want to just sell notes that are obviously bad to anyone who looks at them, I want to sell notes that look good but are statistically bad.

Investors - LC / Re: Here's a new one for Log Entries.
« on: July 01, 2014, 03:33:06 AM »
Best status ever.  You can still sell it on Folio?

As rawraw mentioned, I previously analyzed the data and published results. FICO Score Trends and Defaults for Lending Club Loans

Thank you for the link, some very interesting stuff.

Think of this way.

Some interesting thoughts, but I don't think you can say originating new loans are the same thing as holding on to a loan you already have.  You have different type of information on the two loans, the new origination has up to date data such as income, delinquencies, and credit score, while the existing loan has up to date payment history and credit score.

I think that the credit score should be a good variable to decide whether to dump a loan in an overall buy-and-hold strategy because other than payment history, it's the only up-to-date variable we have.  But just how good is it, and how to analyze the data?

Has anyone run the numbers on this?  It seems logical that a lower credit score indicates the borrow is statistically less likely to continue making payments, but if the borrower has no other negative indications are they really a higher default risk for a loan already made?

I'm pretty sure most people would agree if the credit score drops from 760 to 575, you should dump the loan very quickly as there is something seriously financially wrong, but what if the score goes from 680 to 630 over the course of 2 years?  How much noise is there in a credit score - what's a statistically significant movement that should be actioned?

I've been selling notes with large credit score drops for a while, and it seems to be working well for me with a well-seasoned 15% XIRR on a mostly buy and hold strategy.  I'm not sure how I can get the data regarding notes I've sold performing/not performing over time - any suggestions?

Pages: 1 [2] 3 4 ... 11