Fooled by Randomness

Discussion in 'Trading' started by johnny88, Aug 19, 2009.

  1. johnny88

    johnny88

    Wow! Great insight guys. Thanks to all who chipped in.
     
    #41     Aug 20, 2009
  2. This is pure fantasy...
    Just a bunch of AI jargon tossed out for effect...
    You must be < 20 yo.

    Do you go to a live doctor...
    Or do you have yourself diagnosed by an "expert system"?

    "Expert systems" of whatever AI approach you may suggest...
    Are too primitive to even reliably mimic...
    The 100s of routine decisions a doctor makes daily.

    Many decisions made by Pro Traders...
    Have far more external variables and are more complex than most medical diagnoses...
    Especially in the ** extreme markets ** this thread addresses...
    Where Static Systems are crushed by Expert Traders.

    Most Algo Systems just do simple, repetitive tasks very fast...
    Just endless variations on classic Market Making...
    Hardly rocket science...
    And are shut down in fast markets.
     
    #42     Aug 20, 2009
  3. johnny88

    johnny88

    As far as CAPM in business school goes, it's a good starting point isn't it? I mean how else would you introduce this concept? Anyone can look at the assumptions and say this is not how real world works. But, is there an easier way to introduce this idea of risk when you have to find intrinsic value of equity? Setting up balance sheet, income statement, and cash flow statement can be easily explained for forecasting. But when you have to find intrinsic value of equity how do you introduce this concept of risk? This makes sense to me as far as I can see. Risk free rate + the additional risk you require for the cash flows in the future. The risk you require is where capm comes in. I just want to know what the alternative is, and if anyone has experience doing this for a living. Friend of mine recently had an interview for a investment bank and they made him do this. Project cash flow and than discount it back to find the intrinsic value.
     
    #43     Aug 20, 2009
  4. With all due respect, RC, I disagree... I haven't seen a single concrete suggestion from NNT.
    You may very well be right, T-dawg, sophisticated maths might not be the answer. Maybe then answer is biology or neuroscience. Still it's worth looking if you ask me.
    Agreed, but, then again, our analytical tools progress and get increasingly sophisticated. Maybe CAPM, like Newtonian mechanics, is just the first step.
    How do you know that? I, personally, was under the impression that we don't even know that the electrons actually "go". For all we know, they're everywhere all at once. As I said, I am not disagreeing, maybe math is not the answer.
    That's why he's an idiot. Imagine a general that says that he doesn't have a plan for what to do in battle, in case majority of his troops are wiped out. As to t-bills and options, this is utter silliness. Do you have any idea what t-bill yield is when extreme events happen? Just for your education, it's negative, so you'll be paying interest, not recving it. Try buying options with that. Moreover, if I'm an investor in NNT's options strategy, what's the return I can expect? Can you tell me how this strategy performs under different scenarios? What's the "big payoff"? Does NNT offer answers to these questions? 'Cause if not, guess what, it's a sh1t strategy
     
    #44     Aug 21, 2009
  5. I agree, johnny...

    CAPM is the first baby step and we're not even sure it's in the right direction, no more than that. But that's how science works...

    To me the two most important concepts that are extremely significant and not well-understood in any modern model, including CAPM, are risk-neutral assumptions (I already mentioned) and liquidity (aka completeness of mkts). As soon as you violate the relevant assumptions, things go pear-shaped...
     
    #45     Aug 21, 2009
  6. This thread has all the elements that an intellectual discussion should have including civility. I hope I can add to what you guys have already built.

    I believe a huge chunk of human behavior is not only predictable but is among the most predictable of variables in the equations we calculate to live our lives.

    On the most simplistic level if a human being's hand should touch the hot surface of a stove his next action (reflexive or not it is behavior) is nearly 100% predictable. PREDICTABLE BEHAVIOR

    On a slightly higher plane if you tell me where on the politic spectrum someone stands and he is at or near one of the extremes I can tell you how he reacts, votes, thinks and donates (if he does) on an astonishing wide range of issues in excess of 80% of the time and with many at the absolute extremes, I -- or you -- will hit or exceed 90%. PREDICTABLE BEHAVIOR

    It needs to be noted that predicting whether or not he backs up his opinion by donating to political parties, candidates or causes is a much less certain(predictable) exercise. NOT PREDICTABLE BEHAVIOR.

    Let's continue up the ladder and get into hard market intelligence. Let's consider a chart of a any security --stock, debt instrument or whatever -- that has been in a long decline, six months or longer, and that decline is unrelenting on the weekly chart yet looking at the daily chart we see it has gone into congestion a handful of times -- or more -- and that each time it has come out the bottom of that congestion (sometimes ticking a point or two above the congestion for a short period) with some conviction.

    Trader X opens a short position as it comes out of congestion on the daily and trader Y initiates a long position at the same moment and in fact is the other side of X's trade.

    In the eight weeks following these trades the security changes its previously predictable behavior pretty dramatically. Three times in the eight weeks the weekly chart shows advances. It no longer congests on the dailies at all. It is either swinging higher or lower but does not make even one of those regular "stops" it did in the proceeding months.

    If I ask those reading this thread which of these two market participants is more likely to close his position everyone will come to the same conclusion. Trader X will close it without a second thought (if he had not closed it a week or three earlier) and trader Y might close it, might buy more or might think it is the perfect time to execute a sophisticated option strategy he read about last week online.

    The reason X is so predictable and Y is not is simple. The question posed -- who will close out -- is a proxy for my asking which one of these guys is more likely to continue take risk when he looks at movement that has morphed from consistent to erratic.

    Which one of these guys is likely to take dollars off the table when he no longer understands what is going on.

    But ultimately I have asked which one of these guys is the better trader.

    We can predict behavior and opine on each traders relative skill at trading from a very limited set of inputs and we can be right most of the time. And maybe 80% of the time.

    The mantra of the numbers guys for many years has been that human behavior is sufficiently unpredictable we must either ignore it or not let it exert to great an influence on our calculations.

    Of course others think they can tell me what I will have for dinner on Tuesday.

    The reality is not only quite different but it can also be incredibly valuable. If one spends even three minutes doping out what type of behavior that relates to what we want to know then we can either predict or come to the conclusion, as Rumsfeld was fond of saying it is unkowable.

    To tie this in to an earlier post of mine anyone who thought about whether heavy hitters on Wall Street would be wrecking havoc with the reliability of the inputs at a moment in history when tens of billions were on the table needs to be ashamed of themselves if it took them a whole three minutes to decide whether their behavior was predictable and EXACTLY WHAT THAT BEHAVIOR WOULD BE.
     
    #46     Aug 21, 2009
  7. Hehe... your assumptions about me are as uniformed as your reasoning is. Hopefully your reasoning skills in other areas of life are better developed.

    Suffice to say your (anti?) algo bias keeps you from questioning your base assumptions.

    That and your post is a series of opinions that I could disprove easily... but, maybe in another thread in another forum.

    Geez guys, how many of you here actually know what stochastic calculus is and *why* it was created in the first place? Bueller... Bueller ... ??

    Mike

    P.S. Good systems thrive in crazy markets. Where do you get your info, Motley Fool? Or, maybe you read a book by J. Cramer or some other bloated ego?
     
    #47     Aug 21, 2009
  8. Much of the problems stem not so much from mathematics and the precision it enables, rather, the problems are due to lack of observable variables input to the system being modeled. TDOG (inadvertently?) stumbled upon a good pragmatic maxim, in that generalization is more important than precision when looking forward. An issue that is already being addressed by those who deal with these ideas from a modern approach.

    Anyways, it is similar to physics, in that if you have a closed system and are given all of the external variables that influence that system's state, you can clearly define some model to predict the output of the system (kind of like Laplace's demon, for anyone who's interested). The key is "knowing" all of the input variables, i.e. having them be observable and extractable. The reality is no one knows all of the input variables; some know more than others. Therein, I believe lies the fly in the ointment to your type of discussion regarding predictability of human nature. On a large scale, unless you are privileged to observable precise information regarding the 'cause', your estimation of the 'effect' will be contaminated by noise. Thus, we refer to simpler and more general models to try to extract some of the true signal that is drowning in a sea of noise.

    This is also the basis for much of AI. Regarding the earlier poster, in some areas, AI has easily exceeded many professional experts in classification type problems. Something many would be wise to think about. Of course, the biggest problem I have encountered, is not so much the I/O and mathematical processing, but finding and selecting the variables and relationships which are useful to process (and there are a myriad to choose from). I.e. a major challenge in trading is simply explicitly defining the problem to be solved, which is perhaps, more difficult than actually solving it.

    Regarding fat tails, they can be modeled and dealt with; hint: excessive leverage is not the key, unless you are implicitly guaranteed bailout, which is likely not the case for us less connected traders. But just because gaussian models are not an accurate representation, does not mean they should be thrown away. Fundamentals and classical knowledge should always come first; then be built upon.

    2c
     
    #48     Aug 21, 2009
  9. My God man get a grip on yourself! I have asserted that tens of billions of dollars were left laying about unguarded and those in a position to control the inputs inserted the flip side of what they knew to be reality.

    Disagree with my assertion if you think it incorrect but I can assure that none of what I have postulated can be explained away by noise or any adjustment to the model.

    The inputs were lies. The assumptions used to evaluate all manner of packaged and then distributed assets did not relate to any known reality in this solar system.

    Step back from the bell curve, fractals and all of the brilliant boys from Leipzig. Look at what I have said. I have said that three minutes of real thought would have been sufficient to conclude that the assumptions underlying trillions of dollars of assets that were sold to financial institutions and investment pools -- pension funds to managed accounts -- were flat out fraudulent. They were lies not mistakes.

    My contention is that long before you get to worrying about noise you need to worry about people. The Titanic didn't sink because of a rounding error; inferior design compounded by the shipyard using lower grade rivets sank the ship.
     
    #49     Aug 21, 2009
  10. It's too bad you didn't really understand anything I said; if you did, you would see that it did not contradict anything you mentioned above. As I said, "few" have access to information, wealth, and power that others don't. The problem is that if you are not a member of said "few," the actual observable information that filters down to yourself is masked as noise. Therefore, it does you little good to try to ascertain and model the human "cause" and "effect" behavior, other than to try to describe it in hindsight. If you are looking forward and have limited access, then precision is not the key, as the very inputs themselves, as you so astutely pointed out are contaminated by garbage.

    One entities signal is filtered and masked to the next entity as noise. Your objective is to navigate through the sea of garbage, and find something that is general and reliable over time; whilst simultaneously guarding yourself against poisoned water holes and land mines along the way.

    And by the way, the re-packaged garbage leading to crash is nothing new under the sun. Read Partnoy's 'f.i.a.s.c.o' for some history. The same fiascos happen over, and over, and over. The game doesn't change, only the players (well, actually the major "face ripping" players from f.i.a.s.c.o are pretty much the same).
     
    #50     Aug 21, 2009