Newbie would appreciate some direction

Discussion in 'Automated Trading' started by anonymous, Jan 6, 2009.

  1. Hi, thanks for reading. First of all, here's a little bit about me. I live in Canada and I'm in my mid '30s. I have an engineering degree and am a self-employed programmer and database developer. I'm not independently wealthy and don't have money to throw away, but I do have a decent job and earning potential.

    I have a couple friends who have been day trading forex and futures for about 10 years now with, let's say, "mixed success". About 8 months ago I started to look more into what they were doing, and as a programmer it immediately appeared to me that that they shouldn't be sitting there staring at charts trading their algorithms, but should be programming them and shifting that work off to a computer, allowing them to be more productive.

    Since then I've become increasingly interested (obsessed?) in working with these individuals to produce an automated trading strategy that makes money. So far everything we've come up with hasn't been stable enough to run with real money.

    I have started off with a MetaTrader broker as that's all I know. I've written some programmes in Metatrader's MQL which is basically C for those of you who aren't familiar with it. I've quickly run into the limitations of this platform and have been in the planning phases of a system that looks eerily similar to TickZoom ( ) discussed extensively here.

    The question I can't really answer yet though is, is all of this worth it? Am I just "chasing the wind?" Is it reasonable to expect that I can see a return of the hundreds and hundreds of hours I've put into this, as well as the thousands more I'll have to put in? The farther in I get, the more I realize that this is a serious investment of time, and if I'm going to be successful at it I need to take it seriously.

    I have no problem taking this seriously. For example, I spent 5 years to get my engineering degree, so why should I expect to make thousands a month in the career after only a few months of effort? If I listen to anonymous internet experts (and here I am asking you), some will say it's possible to make money, and other's will say it's not.

    So my question is, is it possible, or reasonable for a "normal" person, to make enough pips per month in this market to be worth the effort? I understand the concepts of risk management and all that pretty well, but I can't answer the question of whether or not I'm able to compete against companies with thousands of programmers and hundreds of millions of dollars of assets ... companies that are bleeding money right now.

    Anyway, I've made this post long enough for now. I'd just really appreciate any directions or constructive criticism anyone might have. I have no problems spending months or years working on something if I get a payoff at the end that's significantly better than I'll make just working as a consultant with over a decade of experience in my field.

    Thanks a lot!
  2. janus007


    Don't tell anyone!!

    Did they waste their time? , but I guess you shouldn't blame it on the software

    If I tell/ ask you a secret....
    The "others" :
    Could be those who only trade at a discretionary basis
    Could be amongst those who tends to sound more clever the less they know
    Could be some amateur wanna-be programmers that have failed

    I know they handle a lot more money than we can't imagine, so.... for them, the keyword is invisibility, for you and I it's leeching :D
  3. Tums


    Let's dissect this event:

    1. an observation is made (monitoring the market)
    2. an analysis is undertaken,
    3. a decision is the result
    4. an action is triggered

    We call this MADA.
    if you can quantify the above, you can automate the event.

    what do we need to observe?
    how do we quantify the analysis?
    what decision... at what time? ... under what condition?

    for some people... it has been exactly that.

    but despair not, within this forum is the wisdom you seek.
  4. Thanks to both of you for your responses.

    I'm not sure I follow you here...

    In case I wasn't clear about what I meant, I was saying that if they weren't spending all day trying to capitalise on their existing trading system, staring at the screen waiting for line a to cross line b, they could be researching other trading opportunities, or possibly just spending more time with their wives and kids.

    Again, I'm not quite sure what you're saying. I should point out that I'm not knocking Metatrader. I'm actually quite happy with my broker, FxPro. They have filled and exited all of my orders without slippage. That's not saying much though as I haven't done much trading yet! My issues with Metatrader are around the limited backtesting opportunities. I'd like to do more of what I found out today is called walk-forward testing, and basically ramp up my testing and optimizations on a massive level. I can't automate this from within Metatrader so although I'm happy to continue to place my orders through Metatrader, I need better tools to help me develop and test my strategies.

    Aha! Finally, something you've written that I understand! :D

    Now you've lost me again :( Invisibility? Leeching? :confused:

    Thanks again,
    - Andrew. (uh oh, now I'm not completely anonymous)
  5. Thanks for your response too. We've been through steps 1-4 several times, but as I mentioned (and no offense meant to my colleagues, in case you read this) their success, for various reasons, has been "mixed" over the years. I think one reason is that trading psychology has run against them in several cases. My hope in automating the system is to remove this impediment to their successful trading.

    I've built 3 simple systems that all work sometimes, in some markets. However, the trick (duh) is to find systems that work in the future, and consistently so. Without further testing, I don't feel comfortable enough to risk further of my own real money.

    So far I'd divide the systems I've been writing into two broad groups:

    1. Systems that place a trade at a certain time in the day, based upon some criteria, and
    2. Systems that place trades at any point when the correct criteria have been met.

    If I backtest, say, 2 months with a system, I can almost always find a set of numbers that in retrospect would have worked for that time period with relatively little drawdown. They are beautiful graphs. If only they showed my actual account balance :) The fact also is that that means that if I'd discovered that input set after 1 month, I could have used it to make money for at least the next month. So theoretically, a curve fitting strategy has the POTENTIAL to be successful. The problem is, how does one choose the one successful input set from backtesting that is MOST likely to continue to be successful in the future? Or indeed, possibly to make the decision that none of the successful backtests are likely to be successful in the future for whatever reason.

    My hypothesis, based on my observations, is that there might be a way to make money. I also base this on the hope that the market has "momentum". I know the term is used in other ways, but I mean that if a market has been in a trend, or volatile, or sideways, it will either become more or less like that over time, or will suddenly change if external forces act upon it. Oil, for instance, went up and up, and then in July got squirrely, then went down until November, and now is lost and confused again. Backtesting strategies in oil shows that an input set for a strategy that works in a highly trending market may be flat, or slightly positive or negative, in a sideways market. Other numbers work in those markets. I know this is like kindergarten to most people on here but this is where I am.

    Therefore, knowing when to come up with new strategy inputs (I mean, how to calculate my entry and exit points, and order size) can be based upon when the market is changing. This can either be done constantly based on a fixed number of days/weeks/months in the past, or it can happen when a monitor (could be a human) detects that the markets have shifted. For now for simplicity, I'm considering the former.

    This would mean that I get a Really Big Computer (my 8 core Mac Pro is a good start but not enough) and start running optimizations. I plan to take a system, and run millions of backtests with millions of different input sets. Those tests will show me a whole bunch of different p/l graphs over several years.

    Then I plan to crunch all those p/l sets to find out whether any of them had significant periods of profit (could be 3 days, or 2 years). I want to provide the system with some indicators of what to look for. For instance, do those systems have low drawdown? More/fewer trades? Are they likely to be the higher overall profit trades? Ones with lower risk due to lower stop losses? I hope to find that there is a pattern in there, where when it's all done I can say something like the following:

    If I optimize my system on the last 8 weeks of data once a week, if it exists I will choose the set that has at least 300 pips of profit a month and a drawdown (calculate how you want) of less than 24%. If there are multiple input sets that meet those criteria, I will choose the best one by weighting the smallest stop loss at 40%, the highest profit at 20%, and the lowest drawdown at 40%. I will then run those numbers for the next week.

    Then down the road I might add the fact that using 8 weeks of data isn't always best. If the market trends change suddenly, then I need to reevaluate my inputs.

    As you can tell, this is quite a lot of work. If it makes me good money, then it's 100% worth it. I just don't want to spin my wheels on it if my plans are based upon fundamentally flawed presuppositions.

    Thanks again,
    - Andrew.
  6. I meant to say a bit more here.

    Regarding this championship, I am happy for those who did well in it. I didn't have a system that performed like that in that period. However, the cynical side of me sees that this could be an example of the "million monkeys at a million keyboards" syndrome. IIRC there were some early favourites who didn't perform as well later in the test as earlier. If you even take enough random strategies, eventually you'll find some that make money over a period of time. In reading some of the strategies there, it seems that at least one or two fell into this category. Every one of the winners there might be very successful traders and my comments might be totally off base, but it unfortunately doesn't prove to me that it's possible to make money using automated technical trading. It proves to me that out of the many people who entered that contest, some of them made significant money during the contest period.

    Back to Metatrader, the two other issues I am having with it are:

    1. I'm a Mac user. All my work with Windows programmes has to be from within a VM. This is OK, but it's not preferable. VMWare only lets me assign 2 CPUs to a VM, and I have an 8 core Mac Pro so I have to run a lot of VMs to take advantage of my system.

    2. Metatrader is not very intelligently multithreaded. Their backtesting will only use 1 core. So to utilize my computer I basically have to run up to 8 different instances of MetaTrader. I have 14GB of RAM and even that much won't let me run more than 3 VMs and 6 Metatrader instances once I start loading them up with years of M1 data. It's doable but it requires a lot of babysitting.

    - Andrew.
  7. First, lets not automatically assume that you're entitled to make a fortune just because you can write a code or two. You can write the codes until you drop dead, but if what you write is a bad system to begin with, you ain't going to make money.

    Second, while your IQ might matter in trading, your diploma certainly doesn't. Those with many years of higher education under their belt tend to entertain this arrogant notion that their their succes as traders is guaranteed. Just witness the implosion of LTCM, which was supposedly ran by a bunch of Ph.D thugs, and other hedge funds that blew up recently and you might think otherwise.
  8. lindq


    IMO, your time would be better spent in index futures, not Forex.

    And your programming skills are about 20% of what it will take to be successful.

    How many mechanics have become world driving champions?
  9. I'm certainly clear on the fact that I'm not "entitled to make a fortune". That's sort of the central point of why I'm here posting. I don't know if I'm basing my plans on good or faulty premises.

    Since I'm a newbie, I had to look up LTCM on Wikipedia. First of all, they were dealing primarily with bond arbitrage, which I am not. Secondly, they had over $4B in capital, leveraged to over $1T, before they started running into issues being too big to continue to provide 40% returns. They were over-leveraged so when a few trades went bad they were forced to close out a larger number of trades at unfavourable positions. It sounds more like a money management and a scalability problem than a fundamental problem with their mathematical models. But that's just from 5 minutes of reading. I'm not particularly interested in discussing LTCM unless somehow it relates to the questions I'm posting.
  10. Points taken. One of the guys I'm working with has been spending quite a bit of time in the last few months with S&P500 and that might be a good direction.

    I am working with two traders. My hope is that I can learn from their successes and failures and use my experience along with theirs to help make all of us more successful.

    To use your example of a world champion racer, how many drivers have become world champions without a good team behind them? How many teams have won Formula 1 races without mechanics?

    Additionally, for the record, there's a difference between a mechanical engineer and a mechanic. Mechanical engineers are designing cars, and mechanics are fixing them. Neither one of course is necessarily qualified to become a world champion car racer. I also don't need to be world champion; I just want to win a few races ;)
    #10     Jan 7, 2009