Mathematically Predicting the Future?

Discussion in 'Automated Trading' started by MarkBrown, Oct 27, 2007.

  1. Where is Marks performance outputs posted for all his system?
     
    #21     Oct 28, 2007
  2. nidarian

    nidarian

    I guess you just explained to all why you dont make a shit load of money.
     
    #22     Oct 28, 2007
  3. nidarian

    nidarian

    What do you mean by perfect trading method?
     
    #23     Oct 28, 2007
  4. MarkBrown

    MarkBrown

    they use to be on my web site i kept them up for about three years in real time - did you miss them? i was averaging a million hits a month with the free oddball site.

    i had spread systems, daytrade systems, the original oddball and some other eod system which is far better than aberration for free etc. all those trades, pl's and charts as well as the code and free assistance was provided by me for years. i also provided patch data for ts among a plethora of other services i can't remember all.

    soon i will have all my old stuff and loads of new stuff up for anyone who wants it. mb

    ps i have no beef with bright - they would do well to embrace who i am and what i do than constantly pick at me. ask any futures industry professional - look at my nfa record. read my accolades and tell me how many people have they made millionaires? i have several who would testify to that, in fact it has been documented at a cme class i taught on trading.

    look, good thing i have no interest in stocks or stock brokering..
     
    #24     Oct 28, 2007
  5. Kuran47

    Kuran47

    fatrat wrote:

    "You have some fundamental misunderstanding of what it is mathematics are doing in the world of trading.

    Trading and market outcomes are random variables. All individuals are doing is modeling randomless and assessing likelihood of a certain outcome."

    I think YOU have some fundamental misunderstanding of the markets. The markets are not random. Just because it's unpredictable (for the long term) does not mean it's random.

    For the markets' movement to be random, it has to conform to the Normal / Gaussian distribution. It does not. There are too many outliers for it to conform to the Normal Distribution. If it can't conform to the Normal Distribution, it can't be random.

    Mathematically, if the markets' distribution is normal, the kurtosis and skewness must be zero (they are not). The Hurst Coefficient must be 0.50 (it is not). The market collapse of 1987, Jan of 1994 and 1998 should not have happened (but they did. So, what are you going to believe: the reality of the markets or the delusions of the efficient market theory?).

    EMT was created by Eugene Fama in 1960, based on a paper written by Louis Bachelier in 1900. Since they did not have any computers at that time for number crunching, it was easier for them to assume that the price action was random, aka Brownian Motion. If it's random, with the Normal Distribution, the math they can use becomes easier. You can use the mean, mode, median, standard deviations, z scores and such. Without the Normal Distribution, the mathematics becomes harder (there's no mean, mode, standard deviations, etc. and you'd have to cpmpute the 3rd and 4th moment about the mean - except here the mean is just a location parameter M with scale parameter C).

    EMT, and its derivatives (CAPM, VaR, Portfolio Theory, etc.) that are taught in business school is basically worthless. Even Fama recently conceded as much.

    The best distribution that the markets conforms to is the Stable Pareto Levy Distribution (SPL). One of the implication of the SPL is that the short term is "predictable", the long term is not. This is why you never have any accurate long term weather forecasts. Both weather systems and the markets are non linear dynamical systems, where the short term is "predictable" (if you keep it rough enough) but the long term is not (remember the "New Economy" of 2000? Dow 36,000?).

    You can argue all you want about how the markets are random, except the mathematics prove that they are not.
     
    #25     Oct 28, 2007
  6. MarkBrown

    MarkBrown

    i guess a perfect trading method would be one which has found a edge which works and is dependable and consistent.

    not random, not luck. a mathematical method which can define what the market will do with uncanny accuracy. no excuses of why it didn't work. no words, no explanations of this and that just cold hard facts.

    "it's like a trading machine which is mindlessly milling from a solid block of ignorance, something artistic which mankind has not the comprehension to appreciate." mark brown
     
    #26     Oct 28, 2007
  7. Thanks, looking forward to it! Please keep us posted on when your website is updated.

    (For whatever my opinion is worth, I do not know of anybody who has contributed more solid free trading material than Mark Brown.)
     
    #27     Oct 28, 2007
  8. fatrat

    fatrat

    Random numbers do not have to conform to a normal distribution. Where do you get this?

    Random numbers can come from any probability distribution. All the distribution tells you is that the randomness is parameterized. Normal forces a mean and some variance. Poisson forces some parameter lambda. Binomial forces parameterization on n and p, and so on and so forth.

    Furthermore, what's the results of goodness of fit tests on daily market returns for normal vs. SPL? If you can show me those results, I'll believe you. If you're suggested that SPL accounts for those odd tail densities for unlikely events, I'll be amused -- I've never used SPL.
     
    #28     Oct 28, 2007
  9. MarkBrown

    MarkBrown

    if you honestly mean that - you will not belive what comes your way soon. i am starting the next year publishing like a mad man.

    mb
     
    #29     Oct 28, 2007
  10. Dude, WTF are you talking about for the market to be random having to be Gaussian? What a load of bollocks. There are a phreakin zillion non-Gaussian distributions, and guess what - random variates obtained from them are just that - RANDOM.

    eg, look at the LGD distributions in the credit world - definitely not Gaussian.
     
    #30     Oct 28, 2007