P2P Lending / NFT Lending Forum

For awhile now I've noticed how each LC loan grade is spread across a wide range of FICO scores.  Justifiable, I suppose, because Grade is a better indicator of LC loan loss than FICO IMHO.  On the other hand investors seem to go nuts over that FICO up/down indicator on the secondary market, as if it were a very sensitive indicator.  I dunno.  I don't have a conclusion here.  I just present a chart...



Those are mostly 5 point buckets.  The software I used didn't want to play nice, so it made a couple of 10 point buckets over on the right side.  Sorry about that.  Doesn't change the picture much tho.

I think you are focusing on level. Early on in my financial career I was told to always look at level and trend. So level is the FICO score. It is a measure of risk. So that level is spread across grades. So what about trend?

Other than payment, FICO is the only variable that updates. If the loans were regraded each month, this would likely be more predictive. But given its all that we have, what information does the trend convey? I believe it conveys how risk has changed since the loan was made. Credit scores have a monthly standard deviation of like 20 points IIRC. I only care about trends outside this normal range. And I believe it conveys very meaningful information, especially for debt consolidation loans, both in the first few months (showing evidence funds were used as disclosed) and over the life as the borrower goes delinquent on other debts, takes new debts out, etc. 

I made my initial decision based on a level of risk. If that risk is gotten worse, then there is a chance the yield is no longer good enough. If I was presented with that loan today, would I still make it? If the answer is no, that loan is sold without hesitation.

Sent from my SAMSUNG-SM-G935A using Tapatalk

from: rawraw on October 31, 2016, 07:27:42 AM

LendingRobot presented criteria across grades in another way that I liked some time ago.
FICO is one of my note purchasing criteria.  I have a floor for each grade I invest in. To me it shows insight into a historical behavioral pattern. Yes, I realize it can be manipulated but I go with the bell curve.

Image: http://i.imgur.com/K0PIELI.png

Funny, but I just happened to run an interesting test yesterday that seems to be relevant here. It was a quick and dirty test and may have been flawed so take the result with that in mind! Caveat Emptor. Still ...

The hypothesis tested was that a drop of N FICO points below FICO edit of previous month for a Never Late loan could serve as a leading indicator to sell that loan. Since the loan is Never Late it should be possible to sell it on Folio at a discount no worse than -9% and avoid a potentially much larger loss if/when the loan charged off. From the pmthistall.csv file I extracted all vintage 2013 Q1, grade D term 36 month loans. All are fully matured now (either Fully paid or Charged off). Bottom line was that I could find no value for N that resulted in improved profitability. I did not search for a sales discount less than -9% that would make this a profitable strategy and think it unlikely one exists. It wasn't even close.

Have you read this yet?

https://www.peercube.com/blog/post/fico-score-trends-and-defaults-for-lending-club-loans

And I wish there was a way to automate the FICO spike detection. When my account was only a few hundred notes, I monitored it and sold those that didn't spike or started to lose the spike. But it's impossible with thousands of notes.

Sent from my SAMSUNG-SM-G935A using Tapatalk

from: rawraw on October 31, 2016, 10:53:03 AM

I wonder if trying to determine an n level of FICO drop of each month for each grade would be a better determining factor, so say a loan is 6 months old w/ an a score at origination of 670, at month 7-n what would the score have be to indicate a possible default?
from: Rob L on October 31, 2016, 10:01:57 AM

Bigger FICO drops can happen for a reason and not necessarily because of missing payments.  FICO is very sensitive to running a high balance on a CC.  I did this earlier in the year with a balance transfer card.  (Was probably really dumb to do as I do not run balances monthly.)  But I got a limit of 1k on a 0% balance transfer CC and topped it off.  My FICO dropped 47 points.  My overall utilization did not change but because I had such a high utilization on just one card it made my FICO go nuts.  The crazy thing here though that when I made a $25 payment to take the utilization under 80% my FICO jumped over 20 points.  How silly is that?  That is the problem with FICO scores is it does not tell you exactly what is going on with them. 
I do not sell my notes just because of a FICO drop, not sure where I would start selling them due to a drop.  When I buy notes on FOLIO I would not touch a down FICO of let's say 50 points in the past couple of months even with a 20% discount.  Now toss in the fact that the notes might have an IGP recently on it and that note becomes toxic for me.  I do not buy many notes on FOLIO so I can be very picky, but even with being picky I have had FOLIO notes with up FICO and perfect 15-month payment history go bankrupt out of the blue.  (Seeing more go IGP recently which has me more than a bit concerned.)
Rob makes a point where he looks at the FICO not from origination.  If I am looking at a 60 month note that the FICO score was destroyed in the first year, do I care what the original FICO score is?  Or do I only care about how the FICO has rebounded after whatever devastated the score?  I wish I knew what killed the score at the time.  CC utilization is way more favorable than missing or defaulting on payments. 
These are things that make FOLIO notes tough to buy and where the poor pricing makes it even tougher to deal with. 

from: Rob L on October 31, 2016, 10:01:57 AM

from: rawraw on October 31, 2016, 10:53:03 AM

Quote"> from: Fred93 on October 31, 2016, 04:21:10 PM

from: fliphusker on October 31, 2016, 03:37:05 PM

from: Fred93 on October 31, 2016, 09:03:57 AM

from: Rob L on October 31, 2016, 10:01:57 AM

For the secondary market, IMO, loan "Status" is a more significant measure of risk than FICO score or trend.

Quote"> from: Fred on November 01, 2016, 03:32:11 AM

Rob I have a Citi too and if they want to give me cash for using their card and not carry a balance I am solid with that.  I have numerous CC that do the same thing and I use them for different reasons during the year.  Citi is very good about upping my limit when I do not ask for it.  My credit is what is called in the garden and just growing.  I have nothing that I am going to use my credit for anytime soon.  So I am ok with whatever monthly hits I take.  If you are taking a monthly hit and utilizing that much you need to be asking for an increase.  They bumped me out of the blue to 5500 and not asking for it.  But ya see where buying on the FOLIO market is crazy as your own FICO swings. 
Ya sell on FOLIO you know it is not easy. 
Quote"> from: Rob L on October 31, 2016, 04:56:15 PM

Quote"> from: Rob L on October 31, 2016, 04:56:15 PM

from: Fred93 on November 01, 2016, 05:25:47 AM

from: Fred93 on November 01, 2016, 04:32:14 AM

Quote"> from: Fred93 on October 31, 2016, 05:25:03 AM

I put considerably more work into the problem and feel somewhat more confident in my results.
Hypothesis; sudden drop in monthly Fico score of a NeverLate suggests a potential sale on Folio to limit losses from
probable charge off is a smart move. My analysis of 2013 Q1 Grade D loans does not bear this out.
One is never better off to sell, even if a "sucker" is found to buy.

EDIT: THIS CHART IS IN ERROR AND CONTAINS A MAJOR FLAW. IGNORE IT AND SEE CORRECTION BELOW

from: Fred93 on November 01, 2016, 05:25:47 AM

from: Rob L on November 03, 2016, 07:29:08 PM

from: Fred93 on November 03, 2016, 07:40:14 PM

Quote"> from: Rob L on November 03, 2016, 07:29:08 PM

from: rawraw on November 04, 2016, 07:58:44 AM

I think that's way more useful, because it makes no assumption on price (the 0% line).  Is that still sequential change? 

How are you calculating return?  The problem is FICO drop loans have already paid interest and principal in most cases.  Not sure if you thought through how to address this

from: rawraw on November 04, 2016, 02:03:05 PM

Hypothesis; Drop in monthly Fico score after loan origination of a NeverLate suggests a potential sale on Folio to limit losses from
probable charge off is a smart move. Again my analysis of 2013 Q1 Grade D loans does not bear this out.
Other Grades may perform differently.

EDIT: THIS CHART IS IN ERROR AND CONTAINS A MAJOR FLAW. IGNORE IT AND SEE CORRECTION BELOW



Not what I would expect... Go figure...

Is this analysis assuming you sell and hold cash vs holding notes to maturity or default?  It seems like from your methodology that is the case.  And these are no longer month-to-month changes in FICO, right?  Thanks for posting

from: rawraw on November 04, 2016, 08:31:08 PM

Quote"> from: Rob L on November 04, 2016, 09:13:17 PM

from: rawraw on November 05, 2016, 05:53:56 AM

Quote"> from: Rob L on November 05, 2016, 10:39:41 AM

from: rawraw on November 05, 2016, 11:03:10 AM

The interesting question, that I never thought of fully, is the assumption the game continues. So this has been helpful to my thinking. My returns always decline when I stop selling fico drops and I used to track it a few years ago to see if I was avoiding more defaults. But I always implicitly assumed the game continues forever. So appreciate the conversation

Okay, so I slept. There is a part of my mind that works "in the background"; not a conscious process, but solutions to problems seem to just pop out from nowhere. Thus it was this morning when I woke up and instantly knew I'd made an awful mistake in this analysis and I even knew exactly what it was. It was plainly a stupid error and I'd just as soon not tell the world how dumb I'd been. Let me instead correct the charts I've posted and provide replacements. They certainly seem more reasonable to me; not to say there's no more dumb stuff in there! Once again Caveat Emptor! None of the revised charts addresses the "time value of money" after loans are sold which "rawraw" has quite validly suggested as a factor that should (possibly) be part of the analysis. LOL! if my analysis continues to screw thing up so badly that will be small potatoes.












Good the Patriots had a bye week or I'd never had an opportunity to correct my mistake.

This is awesome work. I'm not sure I fully understand the data though. It seems like your first chart, from the most recent set of charts, shows that a FICO change of -90 to -100 which are sold for 1) a 5% discount 2) at par or 3) a 5% premium all have a 0% difference in return when compared to buy and hold. I don't see how that would be possible. It seems the return for selling the note at a 5% premium should be at least 5% higher than a note sold at par. Maybe I simply am not understanding the data you're presenting.

from: SLCPaladin on November 06, 2016, 11:22:26 PM

Am I to understand a note with -20 month-to-month change in FICO sold at a a 5% discount would result in a -2.2% return vs. simply holding the note? Likewise, a -20 month-to-month change in FICO that is sold at a 5% premium would net a 1.5% increase in yield vs. selling the note on Folio? Of course, we're tlaking about 2013 Q1 36-term D notes that you analyzed. I just want to make sure I'm understanding the chart. If I understand correctly, your data -- at least for this subset -- implies that it is almost always better to sell a dip in FICO score as long as you are selling for par or at a premium, is this an accurate summary?

Yes, exactly; you got it. Breakeven at 0% Mkup/Disc (including 1% service fee) appears to be a 25 point FICO drop. Who woulda thought ...
Now all we have to worry about is that I wake up tomorrow with another epiphany invalidating my results published to date. Caveat Emptor!

Don't necessarily think I screw up more than most (certainly not much less), but I freely admit when I do.
Think that's been an important contributor to whatever success I've had in my other endeavors.

However, I'm sleeping better and don't have anything rattling around in the back of the head except "rawraw's" time value of money.
That would argue the results are even somewhat pessimistic.

Keep in mind we are talking about D grade notes here! I'm sure the results for less risky loans (lower interest rate) would be quite different.

I've never tracked it, but it seems anecdotally that you are correct.  I'm not as concerned with a FICO drop from 800 to 750 but I am extremely concerned with a 660 to 610.

Yeah, I think you're exactly right.
It's not often mentioned but odds versus FICO score is not a straight linear relationship.
Actually the log of odds (i.e. fully paid's / charge off's) is very close to linear FICO scores.
Straight line on a log graph and all that. Or so I've heard ...

BTW: I chose D grade loans to analyze simply because they make up 65% of my portfolio.
Maybe I should be doing a bit of Folio selling again.

I never did get my free toaster.  Just be sure to price those never late notes at -5% for me. 
I have not read this thread closely will do so later as it is quite interesting stuff.
Quote"> from: Rob L on November 08, 2016, 05:09:03 PM

Quote"> from: Rob L on November 07, 2016, 07:00:57 PM

As you proceed from left to right on the chart to which you refer above, the amount of FICO drop required to initiate a Folio sale increases. There are fewer and fewer loans that meet the sales criteria, hence more loans held. The buy and hold total receipts is always constant, $27,221,921.10. The graph shows simply (total receipts of loans held or sold at a given FICO change) / (total buy and hold receipts; a constant).

Here's a table including numbers of loans sold or held:



So, the dollar difference contribution of Mkup/Disc for the 286 loans sold at -100 point FICO change is small relative to the buy and hold total receipts constant of $27,221,921.10. Maybe this is a poor way of displaying the results, but that's what you're seeing. If you have a suggestion of a better way I could possibly look into it.

BTW: This also explains why the month over month FICO change graph converges to zero benefit or loss since there are almost no loans with a month over month change of -100 or more FICO points.

I'm still amazed that the threshold is so low.  I always looked for 50-75

from: rawraw on November 09, 2016, 05:15:17 PM

Quote"> from: Rob L on November 09, 2016, 05:59:34 PM

from: fliphusker on November 08, 2016, 05:18:06 PM

I have more or less validated your results.  I say more-or-less, because I never seem to do anything quite the way somebody else did it.  For example, I ran all completed 36 mo loans originated after 2011, whereas you ran just 2013.  I ran the "test" only in months where the loan was current, whereas I presume you ran it in months where the loan was current or late.  I did this because I thought it would be better to look for current loans to sell, as late loans sell at very poor prices.  There are probably more differences, as there are many details.  In any case, I get curves of the same shape as yours, although the numbers are a little different.  So I call that validated -- more or less.

But then...  This analysis depends mightily on what fraction of loans in the universe are detected.  In other words, the set you intend to sell.  I believe you used the word "sold" to describe them, as certainly the simulation presumes they are sold.

For grade D  and a 60 point FICO drop (presuming I am understanding your numbers), you see 27% detected.  633/(633+1704)  I see 20.5% .  Ok.  Given our differences in algorithm and data set, those are similar numbers.

But then... I ran this same test on my current portfolio.  It only detects 4.2% of my loans!  Yipes.  That's way different than either 27% or 20.5% !  Red flag.

Now you might think, well the fraction detected just affects the level of profitability, so I would have less profit from a selling program.  Yea, but if the number is this far off, then the other stats, like the fraction of detected who go on to default may be way off too.  This leads to possibly bogus selection of threshold, etc.  I feel like the robot from Lost in Space is waving his arms and yelling "Danger danger". 

Here's what I think is going on.  When I use all loans in the history file, that's a really broad set of loans.  I broke them down by grade, but nothing else.  My portfolio quite different than that, because when I bought those loans, I used a bunch of filter criteria beyond grade, such as: inquiries, DTI, years of credit history, income, and so forth. 

What I'm saying is that it seems likely that historical info from the full payments file is not representative of my portfolio, so statistics gathered on this broad unfiltered set do not apply to my portfolio. 

So maybe the problem is even harder, and we must filter the history, selecting only loans that pass some filter criteria similar to the filters we used to select the loans in our portfolios, and only then will we have statistics that we can use to guide us in how we should expect our portfolio loans to behave.

Doing one grade at a time was a good idea.  The statistics are different for different grades.  Maybe it just wasn't nearly enough differentiation.

Yes, if you're really choosy up-front, you won't have so many that drop and need to be sold.  (I heard you say, in nine paragraphs.)

Another way to say that might be:  "For every up-front selection criterion, there is an uniquely optimal back-end sell-signal trigger."

(And you can go more granular from there, until it stops being worth it.)

from: nonattender on November 10, 2016, 07:30:17 AM

from: Fred93 on November 10, 2016, 05:03:23 AM

from: Rob L on November 10, 2016, 10:02:07 AM

from: Fred93 on November 10, 2016, 05:00:03 PM

from: Fred93 on November 10, 2016, 05:00:03 PM

Another bit of (imperfect) validation...

Instead of running the test on current notes in my notes.csv file, with the sampling problem we discussed above, I realized that this file has all my Charged Off & Fully Paid notes as well, so I ran the test on them instead.  Now this isn't exactly the same as running the test on the history file, where I have updated FICO every month.  In this case I only have FICO at time of that final event.  Better than nothing, and I think final FICO and lowest FICO along the way probably aren't that far apart most of the time. 

The number  I got for FICO drop>60 detection was 22.4% (all fully paid & charged off in my portfolio, ie mix of grades), which very closely matches what I got on the history file.  So maybe my selection of notes doesn't change things so much after all.

Quote"> from: Fred93 on November 10, 2016, 08:42:34 AM

Quote"> from: Rob L on November 10, 2016, 06:24:25 PM