I can't code a simple system that works

Discussion in 'Strategy Development' started by travis, May 13, 2005.

  1. travis

    travis

    I have been working on systems with tradestation for years already, but the things that work in my discretionary trading on the minidow and seem so simple are hard to code, or at least they never seem to work when coded (maybe not properly). The result is that if I am not in a good state of mind, I lose, because I haven't coded my ideas, and the recipe for a good trade.

    On the other hand, after months of work, I always end up with a very elaborate system, with a lot of inputs, which is not what I originally had in mind, and, something I don't even understand, trust or that I can use, even if it seems to make money on paper.

    There are just a few simple concepts that I consider when trading and that I can't code in a simple way which at the same time will show good results when tested, such as:

    - buy rises in an uptrend, sell declines in a downtrend, buy/sell breakouts in a range - decide the trend/range based on a 30 minutes graph

    - reversals tend to happen at 9.35, 10, 3.30, 3.55.

    Can anyone help me, work with me or give me advice?
     
  2. ewile

    ewile

    PM sent
     
  3. Let me ask you this, can you code a system that works absolutely horribly -- consistently? If so, just flip your buys and your sells...(obviously it's not always this easy when you start considering order types etc.). I always love to find a consistent money loser -- that way I can consider the opposite side of the trade.
     
  4. That would be nice but 99% of 'loser' systems accomplish that by the old 'death by a 1000 cuts' and when commissions and spreads are properly accounted for can never be reversed and profitably traded.

     
  5. travis

    travis

    That's correct. I have already tried reversing systems, and they lose both ways.
     
  6. This is what I do (Don't know if it works)... Like I approach any programming problem, I have to put the finger on the core of the problem first:

    1) Focus on your favorite discretionary setup... Stuff that you really like and excited about.

    2) Come up with a hypothesis on why your setup worked... what are the dynamics going on that cause the traders to do what they have to do.

    3) Once you capture the psychological core... Stick with the essence and try to find good ways to approximate it in the code. I think people probably failed to capture the core... they would just say... oh breakout... oh pullback... oh overbought/oversold... oh it got to make money with that one... Everybody looks for the same thing... what makes yours so special?

    4) When you were trading your discretionary setup, what market reads gave you the conviction... I mean if you want to make a wish list of conditions to take the trade... what would that be.

    5) Go thru your wish list and see if the various filters actually help your system. I usually find one or two things that stand out...

    Then again I remember in Market Wizard interview, Ed Seykota said he never really understood why his Donchian setups worked...

    The problem with many inputs (degree of freedom) is just that the chance of curve fitting increases. For example, if you only have 2 or 3 data points... you can easily come up with a formula with enough conditions to fit the curve perfectly. You can address this problem somewhat by testing your system with large amount of data (trade sequence). If your system does not generate enough trades for YM, you can test it with ES historical data since they are highly correlated.

    In summary, if you can capture (approximate as best as you can) the psychological core... you probably got something I like. If you are just throwing a bunch of TA, numbers and indicators and soy sauce into the wok... It might still work but you better diversify in case the gravy dries up because the next takeout is serving the same food.

    If Yan can cook, so can everybody.
     

  7. In software dev Design is everything...

    What about making a rock solid Flow Chart of your idea(s) and then having a couple of trusted fellow software architects check out the flow chart for design flaws...

    Check out Steve McConnells book Software Project Survival Guid - talks about how 'outragesouly' important up front design is to saving one hundreds of hours of wasted work...

    We do dev wrk for large enterprises and i like his methods... should work for trading systems too... also Van Tharps - Trade Your Way to Financial Freedom considers many system parameters that some miss...

    anyway... hthelps a little... :) cj


    __________________
    If You Have The Vision - We Have The Code...
     
  8. travis

    travis

    One extra problem I have is that as soon as I get on tradestation and into the code I lose the perspective on the market and get lost into the code. Just like someone who wanted to say something in a foreign language but doesn't speak it well enough.

    I think I might be better off still trading discretionary until what I do is perfectly clear to me, so I can at least write it down in words. And only later, I will translate my words into easylanguage. I think my problem is that I don't even know 1) if my method works and 2) what that method exactly is, because it's changing quite rapidly.

    All I know for now is the two timeframes I look at: 5 minutes and 30 minutes. The importance of what time it is, support/resistance and I also look at the book. I look at about 20-30 things at the same time, I don't know...the only thing that bothers me is that if I can't code it, as soon as I am not in the right mindset, I will lose money simply because I will weigh differently the ingredients. Basically for me the right mindset means that you can wait over 30 minutes before making an intraday trade. But sometimes I don't have that patience, and that's when I lose.
     
  9. nitro

    nitro

    That is a myth that is long overdue to be contradicted - it doesn't work.

    nitro
     
  10. nitro

    nitro

    Travis,

    I am going to be honest with you. Stop right now with the system stuff and work on tape reading and old fashioned trading. I am telling you this because you have close to zero chance of coming up with a system that works over any period of time of say more than one week and maybe by some miracle a month.

    I will tell you this and I will save you alot of time: find your niche in equities and leave futures for others. In addition, you can have a set of indicators that you can master, be it stochastic, or RSI or MACD or S/R or some combination thereof, that along with you mix with tape reading and discretion, and add to that the emotional control of a fighter pilot and the discipline of a marine, and if you work hard enough you can compete. Stay away from equities that are dominated by computers for the same reason you want to stay away from futures.

    I can tell you flat out that if you try to compete with or against a computer you will lose. The people doing that are in a league ten thousand times more sophisticated than you can imagine.

    nitro
     
    #10     May 13, 2005