trend following delusion shattered

Discussion in 'Trading' started by hank rollins, Mar 15, 2005.

Thread Status:
Not open for further replies.
  1. From New Market Wizards:

    During the interview I asked, "Do you trade the markets fundamentally or technically?"
    That got a good chuckle. He answered, "We trade technically."
    I responded by asking, "Is fundamental analysis dead?"
    Dennis answered, with a smile, "We certainly hope not."


    Hopefully those so eager to bash trend-following appreciate the fact that their profitablility probably relies in no small part to the large amount of money still going into mechanically traded trend-following funds. The question is why they would go so far as to try to convince everyone else that trend-following is dead -- counterproductive in a way, no? :)
     
    #851     Mar 25, 2005
  2. Let us abandon reality for a moment. Assume for the purpose of the following academic exercise that we have an instrument with the following underlying market characteristics :

    Its price curve (discrete) is randomly generated, giving it either an upward or downward bias for n time periods, where n is also randomly generated. This process is repeated for the duration of the entire trading period.

    Let me illustrate with pseudo (not exactly, but please try to understand it) code :

    Code:
      [color=blue]/*   Price Curve Generation                                         */
    
      /*   Rather than using variables, actual figures will be given.     */
      /*   This is to allow for easier understanding of the process, and  */
      /*   other figures are just as acceptable                           */[/color]
    
    
      no_periods = 300     [color=blue]/* total trading period */[/color]
    
      curve_subscript = 1
    
      curve[curve_subscript] = 1170 [color=blue] /* starting price */[/color]
    
      tickvalue = 0.25
    
      
      while curve_subscript < no_periods do
        begin
    
          generate rnd_bias, a random number (real) between 0 and 1  
         [color=blue] /* < 0.5 = long bias ; > 0.5 = short bias */[/color]
    
          generate rnd_run, a random number (integer) between 1 and 20    
         [color=blue] /* length of "trend" */ [/color]
    
          if rnd_bias <= 0.5 then   [color=blue]/* long bias */[/color]
            begin
          
              for i = 1 to rnd_run do
                begin
                  
                  generate rnd_actual_bias, a random number (real) between 0 and 1
                 [color=blue] /* < 0.65 = trend continues */[/color] 
    
                  if rnd_actual_bias < 0.65 then
                      curve[curve_subscript+1] = curve[curve_subscript] + tickvalue
                  else
                      curve[curve_subscript+1] = curve[curve_subscript] - tickvalue
    
    
                  curve_subscript = curve_subscript + 1
    
                end
    
            end
    
          else                   [color=blue]/* short bias */[/color]
            begin
    
              for i = 1 to rnd_run do
                begin
                  
                  generate rnd_actual_bias, a random number (real) between 0 and 1
                  [color=blue]/* < 0.65 = trend continues */[/color] 
    
                  if rnd_actual_bias < 0.65 then
                      curve[curve_subscript+1] = curve[curve_subscript] - tickvalue
                  else
                      curve[curve_subscript+1] = curve[curve_subscript] + tickvalue
    
    
    
                  curve_subscript = curve_subscript + 1
    
                end
                        
            end
    
        end
    
    It doesn't get better than this, right? So I'm awaiting your algorithm for exploiting such a market, ignoring slippage and commissions.

    Disclaimer: This is an intellectual exercise only; any similarity between the above model and real markets is purely coincidental. For example, it would be very easy to add the possibility of no price change, but why?
     
    #852     Mar 26, 2005
  3. Hello:

    "Let us abandon reality for a moment....."

    Just want to point out that at least two of this thread's participants already exist in an alternative universe many light years away, where the basic rules of cause and effect do not apply. :D

    Also just for the sake of argument, is there room for the concept of "stationarity" in your example?

    Happy Easter Everyone!
    Lefty
     
    #853     Mar 26, 2005
  4. Do you actually understand the pseudo??? and his point???

    Well... Happy Easter to you too. :D
     
    #854     Mar 26, 2005
  5. Pseudo? pseudo.... hmmm. No, doesn't ring a bell. I do think however that at points n, n+1 and n+2, an interesting phenomena may occur. I wouldn't dare give it a name.

    By the way, would there be room in your example for pseudo-stationarity?

    Thank you, I am enjoying this Easter holiday already.

    Lefty.
     
    #855     Mar 26, 2005
  6. 1. Nice post. With all the people mentioning market being random in "different" terms... it's a nice "model"... simple and to the point. Hypothetically, there can be other models added like Brownian but it's nice and simple.

    2. Obviously, after extensive runs on this model, we'll come up with a very clean distribution of the "most" quantitatively available measure, with a clean bell curve.

    3. Adding 0, would change the distribution, of course... then it can be interpretted in different ways... noise for one.

    It really depends on the timeframe and market on which you compare the "model"... The model actually is more trendy than some markets. (I'm saying some markets, not all)... Like intraday ES... there's going to be more noise than the model...

    ...

    Well, another model would be to Monte Carlo any market's price series, to create a "relatively" close model...

    Anyways... I'm sure there's other oppositions that people can come up with...

    But I get your point, and a great point. :D

    For the sake of simplicity... can you develop a trading system under a (pseudo)-random data? If so, what kind???
     
    #856     Mar 26, 2005
  7. #857     Mar 26, 2005
  8. Happy Easter to Lefty, TSG and Equalizer!

    Forget about stationarity and everything else. An internet site is holding a trading competition, and your cousin, who happens to be a programmer at the site, divulges the above algorithm with which prices are to be computed. Will this enable you to get an expected return above 0%?
     
    #858     Mar 26, 2005
  9. I don’t want to disclose the system that i use for trading, but i just want to show “proof” of the fact that one can predict the future based on past prices only.
    The market always moves in waves up and down. I have 1 indicator that can mesure how long we minimum will stay in that direction ( I’ don’t dare to use the word “trend”). So to me that’s a proof that I can predict the future moves in a certain way.
    The blue arrow gives the signal that we will go up; the arrows indicate the zone where we have to look to go short. The indicator should go from 0 to 100 and back. Profitability of this indicator only, without combination with other things is over 80% on the long term, with periods of over 90%.
    Don’t interprete this as buying at 0 and selling at 100, because this indicator is only a small part of the total system that I use. The only purpose was to show that as we went up from 0 you should stay long at least until the indicator reaches 100 without disclosing my whole system.

    How can you define if trend following works?
    If I show real trades that give me each week over and over again at least 25% return on my investment and I make more than 75% profitable trades, would that be a proof or not?
    Or is such a track record just purely luck?
     
    • doc1.doc
      File size:
      33.5 KB
      Views:
      396
    #859     Mar 26, 2005
  10. I am tempted to say that if one got a return greater than 0%, it would be due to random chance.
     
    #860     Mar 26, 2005
Thread Status:
Not open for further replies.