Thanks for the update.

It appears that 20% is the average percentage of borrowers who are going to be late consistently. If they are not there now based on their origination date, then they are inching towards it month by month.

It looks like we are getting to the point where we can see the effect of cutting off the bottom of the HRs - it looks like about a 15-20% improvement on the overall performance (that is, the new late rate, and presumably default rate seems to be 80-85% of the previous late rate) That is good! Although definitely not good enough to make the overall prosper portfolio profitable, and certainly not good enough to justify bidding on Es or HRs.

Has Pninen given up?

I was looking forward to a new graph...

But I guess all we gotta do is get out a ruler and extend the lines until everything has defaulted.

Great data Pninen, thank you. And fairly consistent curves each month, and their slope is not very attractive...

How about an update Pninen!

I wouldn't lend to an HR with your money. :blink:

I know that I will get lots of abuse for saying this but I will anyway: "I have 0% delinquency on lendingclub after 3 months of lending." And it seems that they have 0% delinquency for ALL loans. Can other lendingclub pioneers share their data?

If Prosper does not take active measures to implement suggestions of the community, it has no future. Prosper, are you listening???

I have 0% delinquency here after almost 5 months of lending -- even that is too soon to tell, and your 3 months is way too soon to tell. Let's see where you are in another 6 months or so.

I certainly agree with this.

So that means those first couple loans have made their first 2 whole payments? OMFG!! Sign me up! I'm mortgaging the kid! Untold riches, here I come!

To do some of my own analysis, I had to recreate pninens charts to be able to build mine. I'll post them later tonight if I have time, if pninen doesn't beat me to it.

I 0% dq here after a year...so what.

If you scroll down on this page you will see a chart I've created similar to pninen's already. My graph does not include any repurchased loans unlike pninen's though.

Cool, Z, thanks. That look like mine. I'm glad to see yours at the 22% mark like mine. Can I post my new view to that same screen? I'm pretty wiki illiterate as to how that works!

Feel free to add to the existing pages or create new ones. If you want some detailed instructions, feel free to send me a PM.

Me too, 40 loans almost a year and the longest a loan has been late was 3 days.

poof

This graph shows april 06 finally getting ahead of the previous month, the magic is over for that month I guess. Is that just a result of the difference in pninen's graph and yours since you handled defaults differently?

The 09/15/2007 update...

(IMG:http://img.villagephotos.com/p/2006-6/1187065/prosperlate-2007-09-15.gif)

and with each month's curve slid to a common origin...

(IMG:http://img.villagephotos.com/p/2006-6/1187065/prosperlate-2007-09-15-A.gif)

These curves are based on data from Prosper's "performance" web page. For discussion of methodology, see my posts near the beginning of this thread.

Thanks, pninen. Some scary curves there! :o

A little financial analysis and a lot of patience is necessary to lower your default risk. Commonsense judgment may also help.

It would be cool if Lendingstats or Ericscc would allow us to generate individual late loan curves and overlay it with the overall platform.

This looks too much like someone has released balloons into the air on a day with a little bit of wind....

...we need more wind (and gravity). :rolleyes:

Prosper would rather you started graphing the Prosper Select Index instead :rolleyes:

I think this (2nd) chart is missing Mar '07 thru June '07?

Yes. In Feb'07, Prosper changed the rules, increasing the credit score cutoff below which borrowers would not be accepted. This got rid of many of the worst borrowers, so should shift the slope of the all-prosper curves for months after Feb '07. As expected, you can sorta see that in the first chart. If I plotted these curves on the same common origin chart, they'd just blur out the envelope. Therefore I thought perhaps I should put them on a separate chart.

I am sure that they would.

Thanks for your work on this pninen. I think Prosper should post this graph on their performance page.

The 10/01/07 update...
(IMG:http://img.villagephotos.com/p/2006-6/1187065/prosperlate-2007-10-01.gif)

These curves are generated using data from Prosper's performance web page. For explanation of methodology, see my posts near the beginning of this thread.

Thank you for the chart, pninen! However, I have the feeling that the charts show we went from catastrophic default rates to just disastrous default rates. :(

Edited to add: nah, just a little delay and then on to catastrophic. :angry:

This post has been edited by Mark12547 on Oct-9-2007 06:52 PM

Is catastrophic worse than disasterous? I must have missed school that day. I'm not sure.

Seriously, the most recent data points do not seem to show any improvement. You remember this comment from the Doug Fuller interview?

40% increase sounds like a big change. Taking Doug's comment at face value I would expect to see some effect in the data points for 9/15 and 10/01 but I don't. Why not? Soon we'll see the 10/15 data and again will look for evidence of improvement.

A 40% increase is only a big change if the initial number was large. Penncro currently has 1094 accounts (according to Prosper). So that is 1094 borrowers they should be trying to contact. If pre-September, Penncro managed to only contact 1% of them (11 borrowers), then a 40% increase in September (up to 15 borrowers) would not have noticeable results in terms of the late loan stats.

Presumably (hopefully) Penncro's pre-September contact success rate was a lot better than 1%. But given the stories we hear about multi-month late borrowers having received no or 1 or 2 calls, and given what we know about Penncro's previous calling patterns (calls made during the same 1.5 hour window), I bet their prior contact rate was pretty bad. Thus, a 40% increase may still leave the contact rate far lower than it should be.

Moreover, this asserted 40% increase was simply in the contact rate. While contact is certainly necessary, it doesn't in itself mean money came rolling in as a result.

Well, you're right.

This is the age old problem with percentages. Question is: Percentage of what, ie what is in the denominator. I assumed it was +40% of total accounts, but as you point out it could have been +40% of accounts contacted last month, which could be a very small change.

Perhaps we need to get them to talk about what the absolute contact rates are.

Stronger change is required.

I still wonder whether anyone at Prosper has looked into the growing list of borrowers who have volunteered statements saying they are not being contacted.
http://forums.prosper.com/index.php?showtopic=30284 There's never been any response from Prosper on that thread.

Quite honestly I couldn't care less if they manage to contact 100% of the lates. I want to know the % of late money they are bringing in. Period. Any other number figure does nothing to my bank account.

I hear ya. When we discuss esoterica like contact rates we've dragged ourselves somewhere down into the inner workings and the muck. I only mentioned it to point out that Doug Fuller had described this great-sounding process improvement, and yet I couldn't see improvement in the results to correspond to it.

Sometimes it is appropriate to discuss the inner workings, if only to expose whether same are a competent execution.

Most recent charts showing % of late money they are bringing in can be found at ...
http://forums.prosper.com/index.php?showto...ndpost&p=379870

Of those loans that go 1 month late, about 13% get cured, and the rest default.

It is interesting to compare Prosper lates with lates for subprime mortgages:

http://calculatedrisk.blogspot.com/2007/10...prime-data.html

Thanks for the data.

Looks like if a loan goes more than a month late you have one in 50 chance of a cure. :(

To me it looks like many of the more recent months are going to shoot past the first few months of data, in other words, this chart is looking to get quite a bit uglier in the next few months though the very newest loans may improve if prosper manages to do something with collections but that won't be visible for probably six months on the chart where in two months the chart will need to be scaled again to fit the new high default rate.

Every time I think the months are finally leveling off, they take another tick upward...

And who woulda thunk those pre-pre-data loans from early '06 would turn out better than the pre-data of the next couple of months?

The only positive I can see is that the post-Prosper Days loans do look a tad better. Nice what eliminating sub-520 borrowers can do...

Compare with the mortgage data. Mortgages from early 2006 are performing much better than more recent mortgages.

Getting about past time for the November bad news. I can hardly wait... :(

Fast-forward one year, and referendum's port still looks good. LS ROI > 10%.
http://www.lendingstats.com/memberProfile?...erId=referendum

The November 1 2007 update...

(IMG:http://img.villagephotos.com/p/2006-6/1187065/Prosperlate-2007-11-01.gif)

You know, the slopes on those lines don't look very good for lending.

But on a positive note, they look like they'd be fun to ski down. :rolleyes:

~X~