Not logged in

Forums

Viewing 15 posts - 1 through 15 (of 16 total)
  • #112487 |
    Participant
    4 Posts

    Hi,

    as a longterm Donnaforex.com trading community member, i posted my feedback here http://www.donnaforex.com/forum/index.php?topic=6939.msg318568#msg318568

    OK so i played around with StrategyQuant a whole week, spent around 100h with it. I read the manual. The big gap is there is no real “how-to” documentation. With all the different settings, a newbie can do no more than try&error which is frustrating and eats time.

    For example it is unclear:
    – Which TF is best? I tested M5, M15, H1, H4, D1. The bigger the TF, the quicker it produces random strategies.
    – How long should i choose the history period for In-sample, In-sample-validatio, Out-of-sample?
    – which ranking filter should i choose in which step of the process?
    – which initial filter for random generation should i choose? At beginning i applied too many filters, so strategies were nearly all dismissed. At the end i only worked with filter “trades <= 300”, no matter if profitable or not, let genetic evolution find what is profitable. worked.
    – how many genetic evolution cycles are good?
    – which filters should i apply for genetic evolution? I only used In-sample-validation stagnation for 10 periods.

    Results
    – you really need a fast computer for this. My 8 vCPU i7 running at 4GHZ, 8GB RAM, SSD and Win8.1×64 was nearly at full throttle all the time. Multithreading works great. I also tested it on AWS cloud with 32 vCPU XENON, but it was only a tiny bit faster than my i7. Think the virtualization layer eats a lot. The more trades you have on your strategies, the more RAM you need as it stores all trades within RAM and then crushes this data with genetic algorithms. My tests used between 1000 – 10.000 trades. 1000 was quite rapid to process. 10.000 was taking days. This is the fastes method. If RAM is full, data goes to my SSD. With no SSD it’s slower.

    – I was finally able to generate some strategies, using nearly all available building blocks on M5, M15, H1 with some acceptable equity curve, PF1,2 – 2, Avg Return/DD 3-7, stagnation period <30%.

    – My avg. yearly profit only was 20-40% using 5% risk/trade on M5. As this was the BEST coming out from thousands of random strategies together with genetic evolution, i was quite dissapointed.

    – Best part is the robustness test using Monte-Carlo. However there seems to be a bug inside. You can’t use it together with genetic evolution.

    For now i would not buy it, while i surely still have a lack of education on it.

    #125918
    Customer
    197 Posts

    Hi Cyberryder,

     

    You can find excellent howtos in articles section.

     

    Regards

     

    m

    #125921
    Customer
    434 Posts

    Hi matka,

     

    i am not sure if excellent…which article do you mean?

     

    Thank you

    Patrick

    #125924
    Customer
    434 Posts

    Thank you matka, already read those. it was good at beginning when i bought SQ…now i need to know more 🙂 

     

    Best regards

    Patrick

    #125934
    Customer
    505 Posts

    Hi cyberryder,

     

    the SQ is perfect.

    It is the best product on the market at the moment.

     

    You will never find this information “For example it is unclear:”…..

    in any documentation.

     

    It is not so easy to find a good way to generate profitable strategies. The SQ is only a tool.

    It is a very long way to find good strategies.

     

    The robustnesstest is good. But to find profitable strategies we need more curvefitting tests.

     

    The way is realy hard. I worked two years with the SQ. I don´t can say I have a good way to build profitable Strategies.

    But I know more as in the beginning.

     

    At the moment wie work together with 7 people to find a good way.

     

    I have written a tool to install and monitor the SQ strategies. But most of the strategies are curvefitted.

     

    You can use the building process that matka suggest. But this is not the holly grail.

    thomas

    #125936
    Customer
    434 Posts

    so you did not find good strategies yet? i thought if you make portfolio its not a big deal if one strategy has DD and you turn it off.

     

    i though a find for example some advice how to know if the strategy is good before i make WFM test…because it takes more than 4 hours

    or if you make RT test how change do you accept…because i am not sure what change to accept.

     

    no debt SQ is good tool. but i need better know-how 🙂 if i wait few years forex can close before i start trade .. 🙂

    #125971
    Customer
    10 Posts

    I purchased the pro version a couple of weeks ago, and my experience has been similar. My computer is even slower, so it takes a few days. The average return for all the strategies I’ve tested is around 30% per year OOS, and who knows how well these strategies perform live?

    30% per year is great for large traders, but I’m not a hedge fund or an institutional trader. I have a small five figure account, so 30% per annum is a bit disappointing. I can get the same return buying value stocks (which is a lot more fun BTW) or from 2-3 months of online poker.

    The lack of good documentation is also disappointing, and there doesn’t seem to be a lot of activity in the private areas of the forum. I regret purchasing, but maybe the next version will be better?

    #125976
    Participant
    4 Posts

    Hi cyberryder,

     

    the SQ is perfect.

    It is the best product on the market at the moment.

     

    You will never find this information “For example it is unclear:”…..

    in any documentation.

     

    It is not so easy to find a good way to generate profitable strategies. The SQ is only a tool.

    It is a very long way to find good strategies.

     

    The robustnesstest is good. But to find profitable strategies we need more curvefitting tests.

     

    The way is realy hard. I worked two years with the SQ. I don´t can say I have a good way to build profitable Strategies.

    But I know more as in the beginning.

     

    At the moment wie work together with 7 people to find a good way.

     

    I have written a tool to install and monitor the SQ strategies. But most of the strategies are curvefitted.

     

    You can use the building process that matka suggest. But this is not the holly grail.

    thomas

    Servus Thomas 😉

    wow – 2 years and still struggeling to find good strategies. Pfff.. no way anybody will buy this product then with such poor value prop. Yes, i think the product is top notch within it’s class, but at the end only results count. I think nobody expects a holy grail, but i had much higher expectations. As Aegis stated before.. 30%/y is not sufficient for a retail trader, it’s OK for hedgefunds, they are telling us even stable 1-2%/month is top notch – yeah great if you have some hundred millions on the line. Retail traders have to use much higher risk to make at least 100% a year.

     

    BTW: How-to’s are not specific enough. I already read that. What we need is a very deep walk-trough, not a quick article.

    http://www.strategyq…h StrategyQuant

    http://www.strategyq…uilding process

    http://www.strategyq…g process emini

     

    Anyway i hope the product will further develop and maybe i will then purchase it one day. Will watch from the sideline.

     

    Thanks for feedback all! Feedback from Donnaforex.com community is very similar – all people are struggeling.

    #125990
    Customer
    197 Posts

    This is a myfxbook screenshot from one of my recent strategies forward test. It gained 1.88% since July 8 with 0.16% drawdown. Retesting it in SQ with risk resulting 10.21% drawdown gives 11140.91% annual percentage return. Of course this is only a hypothetical example, it is easy to adjust APR hence it is not very important factor, at least for me.

     

    Forgive me but I am struggling to understand how single person with limited budget can think about beating hedge fund with almost unlimited resources like skilled professionals, computer clusters, excellent quality data, time etc. For me playing with SQ is just a very good intellectual fun, I am not looking for rapid financial results. I think it is much easier to find something with this kind of approach, but I can understand that most people are looking for something else. Good luck.

    Attachments in this forum are visible only for registered users.
    #125991
    Customer
    918 Posts

    I think the tool is only as good as the data it is fed.  I think it is unrealistic that a single strategy, on say EURUSD hourly timeframe, will ever be reliably profitable using only hourly data and no exogenous variables fed into the model that underpins the strategy.

     

    If creating strategies were this simple, there would be no human traders required, and no need for any external data feeds beyond the price itself.

     

    Just looking back at past values (indicators) cannot make reliably profitable strategies. The only ones that appear profitable are curve fitted to particular market phases, which come and go.

     

    So what is required to make money reliably?

     

    1. Analyse data in multiple timeframes
    2. Include other exogenous variables (other currency pairs, bond yields, futures spreads, economic data, sentiment etc)
    3. Indicators on the above that cover trend detection, consolidation, ranging, volatility, overbought/oversold, price extrapolation, support and resistance areas, etc
    4. Create portfolios of strategies that cover different longer term market phases, bull, bear, consolidation, ranging, volatile, etc

    From what I hear from Mark, the next version of SQ (V4) will make creating the above much easier.  I am looking forward to this major expansion of what SQ can deliver.

    #125993
    Customer
    505 Posts

    @matka

    I have many strategies on demoaccount with this equitycurve

    The problem is. If I install 1000 Strategies some strategies are good of this 1000. This is a random process.

    I have tested this so many times.

     

    If I switch this good strategy to real account the equity curve goes down.

     

    Look this strategy looks perfect. This equity is from demoaccount from ActiveTrades.

    The history on demoaccount is two years. The longtime Backtest looks simmilar.

    Attachments in this forum are visible only for registered users.

     

    I think If I switch this strategy to realaccount it will go down. (I have tested this with 10-15 Strategies in the past)

     

    I think SQ 4.0 don´t need more indicatores, or some special features.

    The importend think is the curvefittingtest.

     

    It is not clear if this problem can be solved in the future.

    It is possible that a combination of WF, robustnesstest, bootstrap-test can hit the curvefitting.

     

    At the moment it is not clear how I can beat the curvefitting?

     

    Some Strategies are curvefitted and some not. But If I build a portfolio of strategies the amount of curvefitted strategies outweighs.

     

    It is possible that a special filtering with metrics can beat the curvefitting?? But I don´t found this metric combination at the moment.

    thomas

     

    PS:

    This strategy works good too.

    Backtest 3 Years and demoaccount 2 years.

    Same equitycurve on backtest and demoaccount.

    Same trades on realaccount.

    Generated with the old GeneticBuidler 2.2.X

    Attachments in this forum are visible only for registered users.

     

    #125994
    Customer
    10 Posts

    Forgive me but I am struggling to understand how single person with limited budget can think about beating hedge fund with almost unlimited resources like skilled professionals, computer clusters, excellent quality data, time etc.

     

    Generally, hedge funds and large traders can expect smaller returns because they’re moving larger amounts of money. I would be perfectly happy with 50%/year and a maximum DD of around 15%, but I haven’t come across that system yet. Most of the strategies I’ve found generate 30%/year with a 15% DD.

    #125997
    Participant
    4 Posts

    The history on demoaccount is two years.

    2 years of demo account testing? How long do you expect to live? 500 years?  😉

    Seriously i only use demo for some days/weeks max and then switch to live with 0.01lot. So my testing environment is always the live server with 0.01lot.

     

    And i think no one can expect anymore any strategy to be working forever, no matter how good. After 2y of testing, markets may have changed and you can restart again. What we need is an approach to find one strategy and then to re-tweak it iteratively to adapt to current market. From my understanding this is exactly what i can do with StrategyQuant and WWF, therefore i was very interested in this product.

    #125998
    Participant
    4 Posts

    @matka + aegis:

    You can easily make constant 5-30%/month using grid trading strategies for example. You just need a good strategy how to come out of DD phases. Most systems however blow accounts over time. The best EA out there is Envy http://www.donnaforex.com/forum/index.php?topic=4935.0. There are also some skilled traders on forexsignals.com + connectforex.com, who have showed they can do it. Grid and especially martigrids are more or less russian roulette and you carry an extreme risk, nothing any hedgefund will do, so this is the big difference. There are a lot of people not seeing this as trading, but gambling, i would say it’s a mix of both.

     

    FE_Combo system for example is a “normal” trading system using 5 strategies, running since long time and making 24%/y with 8% DD. You can always up the risk/trade and you’ll have more gains and more DD. It’s always the question of the risk you take. http://www.myfxbook.com/members/autotrade/fe-combo/544319

Viewing 15 posts - 1 through 15 (of 16 total)

You must be logged in to reply to this topic.