Nah, it was much worse.
Some people have questioned the “proclivity of UBS for getting involved in fiascos in which the bank believed it was taking relatively little risk but ended up losing large amounts of money” and a risk management policy that (really!) consisted of “still don’t lose any money, but do more.”
You might feel justified in those doubts about UBS risk management after reading (1) that stuff about the guy who lost all the moneys and (2) this:
Oops! UBS’s maximum 95% value-at-risk in the second quarter was 98mm CHF, or around 113mm USD at current exchange rates [updated, ah, math]. So if its returns are normally distributed and that’s a one-tail confidence interval it should have daily losses of over $68mm less than 16% of the time, $113mm less than 5% of the time, $205mm less than 2.3% of the time, $155mm or than 0.14% of the time …
You see where I’m going with this. A $2 billion loss is, um, 29 standard deviations. That exploded Excel’s brain but goofier methods suggest that a loss that big should occur about once in 10^185 days. Or the odds of it happening in the history of the universe are one in a trillion trillion trillion googols. Or so.
To be fair, people seem to think Adoboli’s inverse miracle came from being long a whole crapload of CHF, which had a 20 standard deviation move in September when the SNB announced that it would defend a cap. Which also, if you believe in normal distributions, should never happen – although relatively less never than a loss of 20 times your VaR.
Ibs suckd
I don't understand most of these words
-UBS Quant
Shoot me, but I'm gonna go long UBS. I'm into fat tails.
The Merkel post was yesterday.
I suspect the issue is that their VAR calculation is restricted to the lower of actual economic exposure or the total value of the UBS franchise. Frankly, I am surprised they valued themselves that highly. In a perverse way that is how this is playing out, sure they lost $2B, but does anyone really think any less of UBS as a result? The math was correct.
Over-reliance on VAR will do this to you every time.
"UBS’s maximum 95% value-at-risk in the second quarter was 98mm CHF, or around 85mm USD at current exchange rates"
ummm matt – $chf is at 0.8746 – so you divide, and not multiply….so 98 chf is $112, not that oswald, gunter, hans and kwokie give a rats @ss about this.
Be fair, Matt. The $2bio loss didn't HAPPEN on one day, it was merely REVEALED on one day. They occured over months or even years. Therefore, it is unlikely that UBS suffered more than a 1 in 1,000,000 year event on any day. Therefore, VaR is still totally reliable and robust
What has Google to do with all this???
-Confused UBS MD
Yesterday was bad for UBS. It was worse for the readers of Dealbreaker who had to suffer staring at 1 post for over 6 hours.
Glad you're back Bess.
Matt, don't act like you won't cum in my mouth.
-Rebecca Rickwood
only 15 years old bro, have some mercy…
This is even getting old for me.
-Hakuna Matata guy
Oh, come on. No one here wants to read reasonable, well thought-out arguments supported by fact.
Except for all the people who do?
But please, feel free to speak for the readership.
i hate you matt
Hey…I resemble that statement!
-$hunny
So, they cut 3,500 people to save $2.1 billion…seems like they could have cut 3,501 people and saved $4.1 billion. Just sayin'.
Var is a friend of mine.
Matt, I love you.
-nerd
And where is the Gundlach trial update?
"VaR of 98mm based on trades we knew about, 6Tn if you include the ones in account 88888"
Please don't look into account 88888, there is nothing untoward to see in there. Thank you.
What is standard deviation?
- UBS Risk Manager sweating profusely at his desk today
What is VAR?
-UBS Risk Management
Dear UBSers,
While yesterday may have been the worst day ever, today (through Sunday) are the BEST days ever. Why, you ask? It's Joseph A. Bank's famous $199 sale, of course!!!!
That's right, today through Sunday you can get any of our Signature Patterned 2-button suits for $199 plus with the purchase of any suit, blazer or sportcoat you can get 2 dress shirts or 2 ties or 2 sport shorts or 2 cotton pants for only $49!!!
Come visit us (open 'til 9pm tonight) or click on the link below.
http://www.josbank.com/menswear/shop/SubCategory_...
–Sincerely, Jos. A Bank, the official "tailors"
P.S.: We promise not to tell you that you suck
It moved. The distribution that is.
-O. G. UBS CPO
Charts and graphs? Double the spite? I wish I had 4 hands so I could give this post 4 thumbs down.
Matt has gone crazier than he did when participating in the Guadalajara donkey show.
It's ok boy, Bess is back and we can handle you horseshit since she is the light in the world of DB.
How are your anal fissures today, Matt?
This is all George Bush's fault. I'm still working on connecting the dots, but as a Nobel Laureate that's just a formality at this point
I wish you had four hands to. That way you could shove four fists up your own ass.
@28 shut up loser, it’s getting old.
What is profit?
-UBS Accountant
You wish your Honda Civic was only 15 years old, bro.
Thank you. I was thinking the same thing as I was reading this.
We've never really seen eye to eye.
oooooo burn!
it moved. …. AND I have a video camera!
You should get together with Honda for co-adver deal. Synergies left and right.
Wow what a dumbass – can't spot a clear arbitrage opportunity. The odds of this incredibly improbably event happening again are zero, lightning never strikes the same place twice. A smart person would sell them tail insurance.
~AIG Quant
so much for the diversification effect (bs/bs=bs)
- A person who believes his own
"you're going to like the way you look"
So you're telling me there's a chance…
- Lloyd Christmas, CFA
Lol, tho yur forgetting page 8 of the dress code specifically precludes 2 button suits and sportscoats.
Is the U on your pad broken.
U bs IT
Ill be up in a while, some kinda glitch in delta one i gotta check. Probably nothing, those guys are aces.
What do pirate risk managers care about?
Their VaRRRRRRRR
If VaR is so great, how did all you smart fucks nearly kill the American economy?
Nice.
-Guy Who Understands the "Account 88888" reference.
Because its wreckable ! I took another look at iy and changed my mind.
And you want to be my pirate joke guy….i don't think so.
Floyd
You know, all of those guys in khaki pants sitting around in risk mgmt should be fired. Sitting there in their silly golf polo shirts and pretending to ask hard questions.
How much you wanna bet the final loss is a lot more than $2B ?
Aryeh Bourkoff sucks. His fucking fat head gets stuck in the door frame and he tries to sit there and dis all of securities when his own team has sunk from #5 to #12
Matt, you have this all wrong!
If you back-out the normal assumption and put in a t-score with 1 degree of freedom it would probs happen about 2 times per billion years.
Var is the most stupid method to measure risk. See Nicholas Taleb on black swans.
Banks and hedge funds will NEVER get it.
I know only three firms which profit on "unexpected" trends: Bridgewater, Medallion and autosystemtrader.com
all others play against the trend, average down, bet on "normal" distribution, trade correlations which ALWAYS break out.
every one knows VAR is useless but CEO and CFO and CRO all want to see just 1 number at the end of the day to sleep on and VAR does that. It also proves how less they know what is going on that VAR is the all encompassing feel good number for calling it a day at work.
If you have every stepped foot into UBS then you know what their systems are like. Crap. Every trade goes through this old ass system called….wait for it….Geronimo. The thing is that these trades have to go through like 12 different systems before even getting to GR. The effect of this is pretty interesting. UBS spends all its time reconciling all these systems. UBS mistakes risk management for reconciliation. It would not take a genius to spoof the system.
The market and UBS's risks are not normally distributed: the marginal distributions display fat tails, potentially some degree of skewness, and the copula displays larger than normal dependence among extreme events. Typically, the distribution of the risk factors is assumed normal and the views are processed using a Bayesian formalism. For non-normal markets, one can choose an ad-hoc parametrization. An alternative is CVaR. Obviously we are looking for the VaR-level, or equivalently the number of scenarios below the VaR-level, such that the posterior has minimum-entropy distance to the prior.
Actually since the VaRs are lognormal not normal-normal, it works out to only about 1 day in about 2500 years.
Just being pedantic…
Yup. Sure. Aha.