Log in

A presentation on how algorithms are all around us


  • JohnLast 3785 days ago

    Yeah I know this presentation. Great, in fact   Jaguar the same presentation that in a private message some weeks ago.

    However look this link, it looks like the big boys make huge mistakes. Is it inautorized trading from a human? Hm, I do not think so, more likely it is some bot going mad.



    "Options volume in UBS AG was the heaviest in two years as traders staked out bullish bets its stock can turn higher after big losses accrued by a "rogue trader" sank shares.

    Shares of UBS tumbled as much as 12% Thursday, to a 29-month low, after the Swiss bank said one of its London-based traders lost roughly $2 billion from unauthorized trading and was arrested on suspicion of fraud. The steep losses cast doubt over the bank's quarterly profits and prompted fresh questions about its ability to manage risk.

  • JohnLast 3785 days ago

    Is there something to do with what happened with the CHF. I think so because many high frequency algorythms may be trapped in that kind of movement.

    That may happen because out there exist other specialized algorythms that are designed to benefit from market crashes. In fact those algorythms are constantly at red alert for a market crash. Do they benefit from it or contribute to it. 

    So a specialized range bound high frequency may eat the dust if a crash happen.

    Maybe that can be called a new doji formation. High Frequency Grave Doji Formation. LOL.

    In fact I was expecting this to happen very soon. I remember some times ago when the flash crach happened I was communicating about the possible involvement of high frequence long before it was conformed in the medias. 

    I mean I discuss the way what happened in CHF not its fundamental reasons (I have heard the fundamental explanation but it does not explain what really happened, the fundamental explanation can explain a normal trend not that kind of market singularities). 

  • JohnLast 3785 days ago

    Of course, I may be wrong I am not a high freuquency guru and I may be perfeclty wrong. But still I cannot help myself to stop thinking in this direction.  

    Sooner or later  High Frequency Grave Doji Formation is about to come.

  • CamaRon 3785 days ago

    You bet!

  • JohnLast 3785 days ago

    Yeah otherwize we are dead traders.

  • JohnLast 3785 days ago

    We cannot beat high frequency. As Einstein asks the biggest question. Is the Universe a friendly place.

    I ask. Are the high frequency algorythms  making the market a friendlier place to us?

  • jaguar1637 3785 days ago

    Some equations for working on a time series

    Mean = SMA(x,n)
    MAD = SMA(abs(s-Mean)),n)
    Variance =n*SMA(x-mean)^2)/(n-1)
    Skewness = SMA(((x-mean) /sigma) ^3)
    Kurtosis = SMA(((x-mean)/sigma)^4)
           with (sigma = SQRT(Variance))

             Shannon Entropy H(X) = -n*SMA(pi*log2(pi))

          The  volatility (Historical volatility Sigma)  can be calculated by :

    Sigma^2 =T*Sigma(-1)^2 +(1-T)*y^2

           Iam wondering how to calculate Sigma ? any ideas ??

                T = 0.94 is best pre-defined value
                Y is latest value of bar volatility
    Mean Deviation Ratio = MathSqrt(2n/pi) where n is the number of steps that the market makes on average to reach a certain price change. One can use it to determine the exact skew and build the algorithm to exploit
    With the Skewness = SMA(((x-mean) /sigma) ^3), wa got
    If some could continue to investigate
  • JohnLast 3785 days ago

    Hi i think somwhere on tsd there are indicators which calculates in real time some of the parameters.

    I prepared the entropy function indicator both on tsd and forex  factory but this is only binary entropy for the real entropy I think from mql point should be used arrays as they are used in the calculation of stanard deviation. I mean for the Sigma. However do not forget that replacing iCustom functions and formulas is all i can do.


  • jaguar1637 3785 days ago

    In practical terms 

    Skewness =   n / (n-1)(n-2)    *  For  i= 0 to x=n (((xi - |x) / s) *MathPow(3)