Morgan Stanley Traders Lost $390 Million in One Day in August

[Equalizer]

Quote from PointOne:

Well said. It's hard to remember when you didn't know something because once you walk across the line and get it you can't imagine how anyone can still not get it.

One way to conceptualise it rather than getting lost in conditional probability maths is to consider this:

The host removes all the incorrect doors except one, so you are left with two doors: your door and another. When there are only 3 doors it is intuitively difficult to see what has just happened with the introduction of this new information.

But if there were 1 million doors and we apply the same logic the host removes all the incorrect doors from consideration except one, leaving two doors including the door you originally chose. Now it is patently obvious that your original choice is vanishingly unlikely to be the winning door, so you switch without hesitation. (Of course there is still the possibility that your first choice was correct all along.)

The key to all this is that the host will leave the correct door available in the second round - this is crucial to understanding the logic. This is the revised information.

Well put!. Although this has already been mentioned with 100 and 1000 cups, the "intellectual giants" amongst us beg to differ.

Now wait for the next village idiot on ET try and tell you how that view is wrong - or you are forgetting about this and that - or decision trees BS, blah blah...

Quite frankly, any in-duh-vidual who cannot see this - after so much discussion and many examples - should not be entrusted with using even plastic cutlery.

But I firmly believe they should "play" the markets. Someone has to take the other side of the trade. Next!

 

[tradestrong]

All right..one last post. If I know that the host knew where the coin was, the probability at that exact point that the coin is under the cup in that game is 99%. I have a 100% probability of making it to the point of being able to even make this choice in this game. The probability that the coin is under my cup is 1%.

If I knew that the host was randomly selecting cups and he made it to the end, the original probability of it being under either cup is 1%. That doesn't change. I had a 2% probability of making it to the point of even having a choice in this particular game. So, it's 50/50 at that point. My odds are exactly the same whether I switch or not.

Quote from sprstpd:

There is a story. In the story, you play the game only once. The story mentions nothing of whether the host randomly turns over the 98 cups or actually knows that those 98 cups do not have anything under them. You just know that all 98 cups turned over do not have the coin. Do you switch? Yes, of course you do.

You are telling me that in this one game, that if I told you that the host was randomly guessing which cup to turn over or if I told you the host knew where the coin was would make a difference in your final decision?

 

[heidegger]

Quote from ASusilovic:

One more reason to never, never ever invest into "quantitative market neutral statistical arbitrage" HF´s !

That's exactly what my cab driver told me the other day. Isn't it supposed to be a sign of something when your cab driver says things like that?

 

[IluvVol]

Quote from dividend:

Ahahah... Morgan Stanley loses $400 MILLION DOLLARS in ONE day and what is the big hubbub about?
Lets make a deal (door switch) probability question.

This is why the US is great and the rest of the world hates us.


By the way, this is not an easy problem.
By the way, when this question first came out many PhDs couldn't figure this one out.
The person that famously solved the problem had the world's highest known IQ at that time.

It was and still is one of the harder puzzles to conceptualize the answer.
It is however easy to copy the solution from Google. And that is what 99% of the thread is.


So to the person that got this question wrong, don't feel so bad. It's a difficult problem.
Also, good chance the person asking you couldn't get it without studying the Teacher's Edition first.

a) I dont get your logic about why someone would "hate" you because losing 400mn is not that big of a deal for big blue in the US. I think you are forgetting that many parts of the world have by far reached/outpaced the US' wealthiest. A Mexican is the richest man of the world now, the top 20 richest people contain a number of Russians and Asians. So, you sound as if the rest of the world is still nibbling crackers while Americans savor Caviar. Even Thai dont anymore stutter pronouncing "1 billion". If you suggested another logic then please enlighten me as I then really did not get it.
b) The thread (most of it) was not about solving the problem but understanding the solution.
c) The person who got this question wrong interviewed for a quant position at an investment bank. Working myself as rates trader at a an IBank I have no clue how he got this interview, judging from his attitude and how he described his conversation with his interviewer. Fact is, he did not do his homework. EVERY quant in the world who applies for a quant position SHOULD KNOW this puzzle and its solution. The point of half the puzzles in interviews is to test whether the interviewee has done his homework or not because the puzzles are very hard/impossible to solve in such short time even by very smart people. It comes down to whether you know the puzzles or not. The other half of the puzzles test conceptual thinking and analytical problem solving skills.

 

[IluvVol]

Quote from tradestrong:

I'll take a stab. Perhaps I'm thinking of this from the wrong perspective, but here's my take.

Black comes up 60% of the time. Figuring the 99% confidence interval on 1000 games, the range is from roughly 55%-65% that black will come up in those 1000 spins. This also means that red has a worst case scenario of coming up 45% of the time.

So, if I pay $55 per spin, it will cost me at worst $30,250. On the other hand, I should profit $45,000 on the winning spins. So I should walk away with $15,000 net profit.

Of course I'm sure I'm missing something.

Nope. Hint: This problem is about risk-neutral pricing. It is very related to how you would price an option, using binomial trees.

 

[virtualmoney]

Quote from ASusilovic:

Oct. 10 (Bloomberg) -- Morgan Stanley, the world's second- biggest securities firm, said its quantitative strategy traders lost $390 million during a single day in August as their computer models failed to account for ``widespread'' investor selling.

There is no paradox here. It is an issue of transparency:

A number of ex-college mates who are engineers, accountants, sales managers,etc. asked me if I could trade for them since they have no time to monitor their trades and could only invest long term in mutual funds/ETFs/Bank Structured products(which included subprime CDOs).

Some proposed a 40%:40% scheme, meaning at the end of a period (3 mths or 1 year), I take 40% of the net profit or compensate 40% of the loss if it is a net loss.

A gambler I met, after hearing this, straight away told me it's a fantastic deal. I ask why?

"Just tell them you lost everything and you get to keep 60% of whatever they gave you"

So did these Billions dollar funds really lose that much?

Was is clearly stated that the cup taken aside did not have the coin under it? If not, then it remains that every cup still has the probability of 1/3, so you DO NOT need to switch your decision.

But the real question is: if a new person comes along to play the game when there are only 2 cups to choose from, he could be the patsy because the coin could be under the cup taken away.
How do you know if there is a coin under any cup in the first place?

 

[ASusilovic]

Quote from heidegger:

That's exactly what my cab driver told me the other day. Isn't it supposed to be a sign of something when your cab driver says things like that?

Heidegger, I like cab drivers ! Point is, I am just waiting for the next 25-standard-deviation-we-aplogize-for-this to-happen-because-we-thought-to-be-smarter-then-the-rest-of-the-market-letter !

Translation : Sorry for being Quant !

 

[ASusilovic]

Maybe we should give this thread another "dimension" and discuss about a much more interesting topic, i.e. chaos theory !

In mathematics and physics, chaos theory describes the behavior of certain nonlinear dynamical systems that under specific conditions exhibit dynamics that are sensitive to initial conditions (popularly referred to as the butterfly effect). As a result of this sensitivity, the behavior of chaotic systems appears to be random, because of an exponential growth of perturbations in the initial conditions. This happens even though these systems are deterministic in the sense that their future dynamics are well defined by their initial conditions, and with no random elements involved. This behavior is known as deterministic chaos, or simply chaos.

Question : Did we experience a butterfly effect during August ?

 

[sprstpd]

Quote from tradestrong:

All right..one last post. If I know that the host knew where the coin was, the probability at that exact point that the coin is under the cup in that game is 99%. I have a 100% probability of making it to the point of being able to even make this choice in this game. The probability that the coin is under my cup is 1%.

If I knew that the host was randomly selecting cups and he made it to the end, the original probability of it being under either cup is 1%. That doesn't change. I had a 2% probability of making it to the point of even having a choice in this particular game. So, it's 50/50 at that point. My odds are exactly the same whether I switch or not.

So you are telling me that if you played this game once, and the host turned over 98 cups and no coin was underneath and he did not mention whether his actions were based on knowing or luck, that you would not have enough information to switch even though the physical result of his actions is exactly the same?

 

[gbos]

Quote from sprstpd:

and he did not mention whether his actions were based on knowing or luck

This is a totally different problem definition. Each definition has a different solution.

1. Host knows --> you have to switch
2. Host doesn’t know --> It makes no difference switching
(I have explained why a few pages ago)

3. You are uncertain about Host Knowledge --> start with a prior belief and after each turn update posterior belief with the new evidence