JP Morgan

[JMO]

Originally Posted by Dr T Non-Fan

Be my guest!

You should charge him at least $.25
JMO, of course.

[E]

It will look great hanging over the obnoxiously large, life-sized bronze Wall Street bull I have floor mounted near my doorway.

[Dr T Non-Fan]

Originally Posted by JMO

You should charge him at least $.25
JMO, of course.

Well, I've only used this transitive property twice here. Really not enough precedent to take to Trademark Court.

[Jack]

Originally Posted by E

I'm not that concerned with JPM shareholders. If JPM does stupid stuff, their shareholders should take it on the chin. If shareholders don't like taking it on the chin, they should evaluate management.

Again my problem is when the actions of one or two banks cause immediate and devastating contagion effects throughout the whole market. They are the market, and that is not good.

That's right. If they fail and only the equity and bond holders lose that's fine. If they fail and take down the whole credit market with them that's not fine.

[campbell]

It's the Frenchies' fault!

http://www.thedailybeast.com/article...ch-trader.html

The City of London and Wall Street are full of men and women from France who pass for geniuses, and sometimes they are. The key to their success is the language they speak, but not necessarily that of Camus or Molière. They know how to talk mathematics of a very rarefied kind that most of us find quite impossible to understand. And as the market for derivatives, the exotic financial instruments based on that math, exploded over the last decade, so too did the demand for the skills of those who could master the relevant equations. (And if they couldn’t, who among us could tell?)

The reason that so many came from France is that basic French education puts a strong emphasis on pure mathematics, and many French universities have developed programs specializing in “the very, very complex modeling of derivatives,” says one young French veteran of the City.
.....
Jérôme Kerviel, whose supposed “rogue” trades brought Société Général to its knees in early 2008, was a graduate in finance from the University of Lyon. “Fabulous” Fabrice Tourre, allegedly the bane of Goldman Sachs, which paid out half a billion dollars in fines related to his work, went to the École Centrale Paris, a top-flight engineering school. Bruno Iksil, the latest Frenchman to be partially but publicly implicated in disaster, this time at JPMorgan Chase, also reportedly attended the École Centrale. He was known as “The Whale” because he took such huge positions in the markets where he operated. He was also called “Voldemort,” after the nemesis of Harry Potter.
....
Most of us can see in hindsight how dangerous this juxtaposition of mathematical theory and multiple billions of dollars might be if something goes wrong, but for the traders in the middle of the action following the model and counting on it working out, that may not be so obvious. They’ve seen the numbers sort out before, and maybe they will now. The outside world hears only about the derivatives that go wrong, while careers are built on the ones that go right. Or, at least, they were.

Thing is, I doubt they had a huge amount of "pure" math ... this is just an author who figures that if you don't see numbers in most of the work, and that proofs are involved or if it's really really complicated, it must be "pure".

But this does put me in mind when I first saw the model structure behind CDOs, in an impromptu lecture by John Hull at an ALM seminar, back in 2004. People were having math orgasms, and yes, the math was lots of fun, but there didn't seem to be much thought about how elegantly the models would fail.

To wit: when they fail, they fail ugly. That's not so much fun.

[whisper]

Originally Posted by campbell

Thing is, I doubt they had a huge amount of "pure" math ... this is just an author who figures that if you don't see numbers in most of the work, and that proofs are involved or if it's really really complicated, it must be "pure".

Actually, if they knew anything - they'd understand that what was actually done is far from "pure" mathematics - it was physicist math.

It's one of the things that I rant on - that the assumptions matter. You can't take an elegant piece of analysis that works in a very specific area built on very specific assumptions and use it in another space that has about as different set of assumptions that you can.

If a key assumption is a normal distribution, if your data isn't normally distributed - you can't use that concept. If a key assumption is infinitesimal time frame, you can't use that concept over a 1 year+ time span.

To wit: when they fail, they fail ugly. That's not so much fun.

I don't know, the catastrophic failure can be fun and certainly interesting....

[campbell]

Well, you can use those assumptions... but you need to know what will happen when you're in the regions where things don't behave properly.

Sometimes it doesn't matter.

This wasn't one of those times.

[Brad Gile]

Pure math is wonderful. It was my turf. The problem is that you have to set down the assumptions and try to see what may happen when they are wrong.
This can be done only if the tester understands the underlying assumptions.
Worst case is when someone uses assumptions that make computations easy just for that reason.

No matter what you use, it just makes sense to analyze differences by actual and theoretical results and test whether the model is at fault. It makes even more sense NOT to offer a method before you have a good idea that your assumptions are reasonably realistic. It truly amazes me that many people failed to do this. Greed is probably major here.

[whisper]

Originally Posted by Brad Gile

It makes even more sense NOT to offer a method before you have a good idea that your assumptions are reasonably realistic. It truly amazes me that many people failed to do this. Greed is probably major here.

You'll never have perfect information, so there will always be some risk of model failure. With that said, their model failed to replicate the known information.


Sent from my Xoom using Tapatalk 2

[asdfasdf]

I see a bit of a hiesenberg problem, the better you model the known world, the more your model will influence other people's behaviour, so it's never possible to fully model things.