after how long will you reoptimize the parameter

Discussion in 'Strategy Development' started by robinxing, Jul 9, 2007.

  1. for an intraday trading system

    after how long or after how many trades will you reoptimize the

    parameter

    now I reoptimize the parameter after 2 months

    or after 200 trades

    keep the parameter for one month

    what do you think of it
     
  2. KS96

    KS96

    does it work?
     
  3. it works

    but I am not sure wether it is the best way
     
  4. KS96

    KS96

    there is no best way
     
  5. Parameters fit nicely into about four categories.

    They do have to be adjusted and the reasons for adjustment vary from parameter set to parameter set.

    There are also valid ways to check out your adjustments to see if they are proper.

    You may have noticed that many people do not use intraday parameters for an assortment of reasons. These reasons are closely associated with their personal failures to get parameters adjusted and keep them adjusted. A secondary reason is that they fail to understand how to get signals.

    A lot of my adjustments are quant oriented since I front run quants from a parasitic orientation. The quant type results are "gating" type logic considerations.

    The four parameter areas are:

    1. Indicators

    2. Market technical conditions

    3. Variation from standards.

    4. Trader orientation.

    You are restricting your inquiry to optimizing for day trading. I am assuming you are in the ball park.

    The recent example example of Batchelor and Ramyar (SEP, 2006) shows how being out of the ball park causes that wrong artefacts of the parameters destroy any possible potential for results. ranfiel's comments show he does not understand this failure symptom; he often also demostrates he is out of the ball park as were the authors above cited.

    The parameter use and its signals are the function that drives optimization. Market conditions and the passage of time do not, except for the reasons cited below. Long term longitudinal studies always fail because they do not take into condition the causal factors involved in parameter adjustment over time.

    the primary causes of parameters going out of whack are as follows:

    a. technological advances.

    b. data transmission advances.

    c. data processing advances.

    d. failure in aspects of EMH.

    e. seasonal changes

    f. standard changes

    g. Human factors that are a consequence of broad life style changes

    h. human factor that are a consequence of collective feelings.

    i. the shift in how things are measured by prominent players.

    The techniques for dealing with optimization fall into the following categories:

    A. arithmetic adjustments

    B. periodicity adjustments

    C. regression adjustments

    D. moving adjustments using A, B and,or C above)

    What fouls most mediocre traders up, they tell us, is that their trading effectiveness comes and goes. Usually it is a blame game small or large. By simply looking at the subject of the blame, it is easy to see where the failings lie vis a vis optimizing parameters.

    A neat way to stay sharp on these considerations is to track an assortment of methods and tune them up. you can also play the game of "what would take the method or person to the next level of performance".

    there also comes into the picture which method to use to validate the optimization. Using a common method that applies broadly is the easiest, especially if the person doing the optimization has not acquired a lot of skill.

    One footnote: All of this requires that the independent market variables be used in the optimizing process.


    Optimizing is a three step process: What; Why and How to adjust.

    I'll do one for you (an intraday trader) that is a quant front running cool deal. I''ll do two since they are closely related.

    Quant based arb'ing is a threshold thingy. They do not see it until it is present to a minimum degree. Quants do not trade, they tell associated traders how their model is detecting stuff; the trader uses the detected situations to do what he does. You take the feed of the most common arbitrage parameter and do an MLR on it to determine what is happening to it. To get your screen to always be unbiased, you feed corrections to the screen display (this is optimizing the display so you have a consistent precisely the same signal which others are no longer getting); and you run a count on the corrections and the frequency of the corrections.

    The second example is the "settlement correction" on the most common arbitrage parameter. This is a "jerky" situation and it has a damping optimization correction. you get to use an ajustment ahead of the market that is over, under or critically dampening the non bias you are maintaining.

    These are both daily (streaming actually) optimizations and people who are not making multiples of the h-l range do not bother. Such people usually do experience lousy afternoon trading as a consequence and they will never know what, why or how.

    Arbitrage has a hedging aspect wherby the arb trader fed by quant modelling recognizes one thing he is in can't be fixed because he is tardy so he does the next best thing. he tries to cover the difficulty by hedging in a place expected of him by those observing him. This timing offset which is endemic to hedging arbs is an advantage to others facing this person. optimizing this dynamic is done on the display of the optimizer.

    A slow example is to watch Pring. He is very conservative in his optimization efforts. Look at two places: the What is an indicator; the why is transmission advances; the how is arithmetic. Also look at (this is a pending one he has not as yet invoked): the What is market technical conditions; the why is data processing advances; the How is that he has to change his starting point for a trendline, a periodicity adjustment in the phase variable.

    In the SPM thread you can see that the adjustment Pring finally made still has not been made by the SPM OP. Same for the charts of InvestTools and the same for the charts of Joe Pepper 2001. What happened that they all missed at first? Batchelor, Raymar and ranfiel, above cited also missed it. thw what is technilogical change cause periodicity changes in the psychology of traders. Why it got out of whack was that the original defaults of indicators "bridged across the shorter comtemporary cycles and an adjustment had to be made to get the indicators to go back to their prior signalling. This was done arithmetically by adjusting the defaults. The 12, 26, 9 of MACD does not work since the PC was invented and came into common use.

    The seasonally affected parameters most also be handled by each season. It is not abrupt so moving adjustments must be made. People fail to change their lookup tables seasonally in automated systems.

    Do the What, Why and How for the FOMC. You have three cascades and a damping to the regression on an FOMC day. each has to be delt with BEFORE you go through the trading opportunity. The FOMC periodicity is like a clock and the look ups for each cascading have to be optimized and so does the asymptotic damping to the appropriate regressions for the day. this is an example of one of the shortcomings of EMH.

    Lets say that a person is just learning to strade and he doesn't do automation very well. Lets say he doesn't have a good mentor who can help. Let's say he just reads books and stuff to learn. None of this is related to optimization.

    optimization only come up later for someone who is accomplished. Take a look at readon metal who commented on my prose yesterday. What are the parameters he has to deal with now? Why are his parameters out of wahck since late last year? How is he going to shap[e them up to get out of drwadowns that equal his periodic gains? If you are a person who has everything being opptimized either manually or automatically (the better way, I can assure you) then give him a hand or two.

    The obvious stuff I did not get into would have make this post too long.
     
  6. To make a long story short: my basic parameters have never been changed since 1990 and they still work very well.

    The art is to create parameters that automatically adapt to the markets. Within certain limits i can even change my parameters without having significant differences in results.
    I don't trade on 1 parameter or on 1 indicator. I trade a system where the different parts keep each other in balance and compensate for the not always optimized indicators separately.

    I strongly believe in behavioral finance.

    Optimization is optimizing your system for the PAST period. It will never give you a guarantee for the future.
     

  7. That's a given for indicators in general, optimized (i.e., curve fitting) or not.

    st
     
  8. Yes, but there is a substantial difference between all of them.

    The number of my winning trades has always been more or less at the same rate. The system has always been profitable, no matter what market we were in.
    So the probability of the outcome of my next thousand trades will be far more accurate than those of the indicators that have to be re-optimized on a regular basis, with the risk that they will stop working after a while.

    I did lose however lots of money before i completely understood how to manage my system. It was not a failure of the system, but a failure of the trader.
     
  9. per spike 500:

    "Optimization is optimizing your system for the PAST period. It will never give you a guarantee for the future."

    The future is not where trading takes place.

    Optimization takes the future off the table.

    Optimization is oriented to the present only and that includes all your leading parameters of price.

    I agree that behavior finance is where much trading improvement may be made.

    It is more important to keep one's self optimized compared to the automated or manual trading parameters a person is using.

    Not using displays where the scaling is oriented to filling the picutre is an example. Standard scaling standardizes and optimizes the viewer.

    EQ and IQ go hand in hand.
     

  10. Personally, I don't use indicators. I found them to be nothing more than a derivative of price, and therefore, always lagging. Using indicators rather than price itself to determine setups of entries and targets, in my humble opinion, is a losing proposition, and thus the reason so many traders fail. Raw price action never needs to be optimized, it's always there, showing you the way. But I couldn't agree more that traders themselves are the weakest link of any system, not the method itself. We can be our own worst enemies.

    Regards,

    st
     
    #10     Jul 9, 2007