'Discretion' in discretionary trading

Discussion in 'Strategy Building' started by bearmountain, Sep 12, 2010.

  1. While your premise is attractive to those with analytical minds, myself included, I believe it's false. And I believe a couple of examples will illustrate why.

    First, though, we need to define what "mechanical" trading is. For simplicity's sake, I'll use a two part definition:

    1) A trading system sufficiently well codified that you could hand the rules to someone else (or a computer), and the trades they would generate would be the same ones you would have if you executed the system. Obviously there will be some variation in fills.

    2) The above system had been tested (forward and/or backwards) and on the basis of those tests is believed with some statistical confidence to be profitable.

    I'll just define discretionary trading as trading that lacks the properties of mechanical trading.

    On the basis of this, I would suggest that most "mechanical" systems are in fact not entirely mechanical, even if they initially appear to be. Their discretionary components usually manifest in two ways:

    A) Dealing with rare events. Consider for example the flash crash. Numerous mechanical systems during the crash made the mistake of selling at the exchange's minimum allowed price ($0.01) or buying at the maximum price ($10,000 or whatever). Clearly these systems in fact did not meet test 2) above - had they even once been tested on similar stimulus, such bugs would have been eliminated long ago. Point being, even systems that are supposedly mechanical often behave in ways that the people trading them never expected when exposed to events that were not in the test set used to develop the system. Since the markets are constantly doing new things they've never done before, this problem never goes away no matter how careful the system designer is.

    B) Built-in assumptions & decisions. Most systems have aspects built into them that are never tested. Consider the following decisions:

    - what instruments/exchanges will be traded (or what universe will they be picked from)?
    - what times of day will be traded?
    - what brokers/exchange access, data feeds, execution platforms and programming environments will be used?
    - what kind of losses will cause the system to be turned off? Will it be discarded at that point, modified and re-started, or re-started later as-is?
    - will the system be turned off when certain types of rare events (see above) happen?
    - how will position sizing for the system work?
    - what space of possible systems will be searched for profitable ones? How will over-fitting be avoided?

    Now, anyone who has traded knows that those decisions have huge impacts on the bottom line. And yet, for a variety of reasons the decisions I listed are rarely the same from trader to trader (violating 1)) and are rarely tested either backwards or forwards (violating 2)). In place of testing, these decisions are often made by the system designer who chooses values that seem reasonable. As such key elements of supposedly mechanical systems are in fact discretionary.

    Now obviously my argument could be countered by expanding the definition of mechanical trading to encompass all trading, or alternately arguing the human mind is "mechanical". Such word games aren't really interesting though.
     
    #21     Sep 12, 2010
  2. heech

    heech

    Yes, absolutely, there are various examples of human endeavor in which we can't quantify rules for success.

    Can you design analytical rules for describing beauty? How about analytical rules for composing music? Can we design an artificial Mozart? What was his secret sauce?

    I'll give you another example. It remains incredibly difficult to draw animated movements correctly... we can render complicated solar systems with 100 different light-sources + 1000 different textures... but try to write a program that correctly shows a baby crawling. It's an incredibly difficult task, and exactly why motion-capture of actual humans remains the standard method for animating movements in video games / animated movies. On the other hand, Disney's animators were able to draw animated movements by *hand* decades and decades ago; they (and we) know by instinct what looks right, and what looks wrong.

    Keep in mind I'm absolutely an analytical person. I majored in computer science, and I develop automated trading systems. I've also studied/implemented artificial intelligence systems as part of my graduate course work, and I know there are many limitations to our ability to imitate the human mind.
     
    #22     Sep 12, 2010
  3. sure, I think I understand what you are saying. I would argue that designing an artificial Mozart, is a limitation of our current knowledge of human neurology, programming etc. Would we be able to design an artificial Mozart 200 years from now, in 2210?

    what are your thoughts on the secret sauce in discretionary trading? what is it? what role does luck play? Thanks.
     
    #23     Sep 12, 2010
  4. its really not that complicated...there are a number of things that a discretionary trader can benefit from that a fully automated trader cannot...here are a couple of simple examples:


    1) How is the market reacting to certain news events? We all know sometimes the market reacts really badly to bad news, and sometimes, surprisingly, it doesn't. A discretionary trader can take advantage of the insights this gives...i'm not sure we can effectively incorporate that into an automated system yet.


    2) Understanding what key factors are driving the market at any given point....case in point -- this year the market has pivoted back and forth between being driven by corporate profits/earnings, and macro-global events and conditions (and risks). Again, an insightful and capable discretionary trader can take advantage of this knowledge whereas its going to be difficult or impossible to do so with automation.



    I'm sure there is a long list of these things, but the two above are quite simple to understand, quite important with regard to their influence on the markets, and probably impossible to effectively automate.
     
    #24     Sep 12, 2010
  5. Thank you all for your replies. This has exchange has been very beneficial to me, help clarify my thinking.

    Instinct: Sea turtles, hatched on a beach, automatically move toward the ocean.

    "This is perhaps similar to a successful discretionary trader who can't articulate with completeness their decision making logic."

    George Soros is perhaps a good example and study of this.

    “I rely a great deal on animal instincts,” he wrote in his 1995 book, Soros on Soros. “When I was actively running the fund, I
    suffered from backache. I used the onset of acute pain
    as a signal that there was something wrong in my
    portfolio. The backache didn’t tell me what was wrong –
    you know, lower back for short positions, left shoulder
    for currencies – but it did prompt me to look for something amiss when I might not have done so otherwise.”

    Soros finds the flaw by combining theory and instinct.
    This is puzzling because theory and instinct are usually
    considered mutually exclusive. This might be one of the
    few matters on which economists and traders agree.
    Economists think that gut feelings are irrational. And for
    most traders, instinct and method contradict each other.
    When traders do use their instincts, they’re overriding
    their methodology. But in Soros’ method, theory and
    instinct are inextricably linked. Soros himself, however,
    could not explain how.

    I think the question is how to link trading method and instinct?
     
    #25     Sep 12, 2010
  6. heech

    heech

    No one knows with ANY kind of certainty. Very informed people thought we'd have rudimentary but authentic artificial intelligence 20 years ago... but now they believe we are further than ever. The current thinking is that we need quantum computing before we have the ability to even start thinking about the problem.

    But one thing is certain: it's not an if/then/else kind of problem.

    That's a different question! I'm a horrible discretionary trader.... and I'm by default suspicious of those who claim to be good discretionary traders. Like I said, I'm an analytical person. :)
     
    #26     Sep 12, 2010
  7. I agree.And would add. Lack of experience would lead to WHIPSAW- snide remark about discretionary trading
     
    #27     Sep 12, 2010
  8. ammo

    ammo

    you teach your children that killing is wrong and to turn the other cheek, but if a situation came up,we would kill to save them..theory and instinct
     
    #28     Sep 12, 2010
  9. I might compare disciplined discretionary trading to flying a jet airliner, in terms of the mix of discretion and "coding".

    You (the pilot) have your strategies, rules, guidelines, stops/targets, etc. (flight plan, regulations, procedures, ... even autopilot to use at some point). But before taking a trade (taking off), you still have to account for news, recent volatility, S/R on larger timeframes, trendlines, MA's, etc. In other words, no matter how the tower instructs you, you have to make sure the runway's clear, the controls are working, the co-pilot hasn't just spilled his coffee all over the console ... etc. etc.

    And I would agree with the idea of experience being the key. When the plane loses its hydraulics or the cabin suddenly bursts into flames - and trust me, if you trade long enough the cabin will at some point spontaneously burst into flames - the one who's going to survive (and possibly even come out a hero) is gonna be the one with that extra experience.

    I'm not saying this is the only way to trade, but in discretionary trading this is kind of how I view things.
     
    #29     Sep 12, 2010
  10. I think this is the bottom line -- where intuition, gut feel, the "secret sauce", all stems from: experience.

    Can it be translated, automated? Who knows. I do know there are alot of qualitative nuances that would be damned difficult to teach a program to recognize, that a human trader could process in an instant.
     
    #30     Sep 12, 2010