Simple Moving Average Crossover Strategy

Forums › ProRealTime English forum › ProOrder support › Simple Moving Average Crossover Strategy

  • This topic has 95 replies, 12 voices, and was last updated 1 year ago by avatarCRISRJ.
Viewing 15 posts - 61 through 75 (of 96 total)
1 2 3 4 5 6 7
  • 03/31/2022 at 4:06 PM #190870
    phoentzs

    Oh… so the last two filters are too much?

    03/31/2022 at 4:46 PM #190872
    nonetheless

    Your first version looks very similar before 2018, so the changes haven’t had much effect on the long term picture. Like i said, it’s curve-fit to that sample period.

    Most likely my version is also curve-fit to my sample, it’s just a 5 year curve instead of 2 1/2

    It doesn’t mean that it won’t work. Maybe those market patterns will persist and it will make a ton of money, who knows?

    03/31/2022 at 10:19 PM #190884
    phoentzs

    I have another basic question, maybe someone has an assessment or opinion on this. Referring to my version V3 MA-Cross… I built a breakout version (theoretically worse entry price) with the same trend filter and the same system structure… then I created a version with MACD rebound (theoretically better entry price) with the same structure… and now here with MA-Cross( theoretically the golden mean from the entry price). The results of all three systems are damn similar, performance, hit rate, drawdown, performance… everything differs only slightly. That makes me suspicious, should the type of entry be relatively irrelevant in the end? Because you’re most likely to end up in the same movement anyway, so it’s more important how the position is managed. Trailing, SL, TP, etc.

    Does anyone have an opinion on this or has anyone tried the same thing before?

    04/01/2022 at 8:46 AM #190901
    VinzentVega
    nonetheless wrote:

    I expect you’ve decided that you don’t really care what happened before your test period, but just in case you’re curious:

    This is the basic problem when backtesting over the full period. It is all optimized until the last trade. The coming realtrades fail then.
    if you still remember, all the vector setups from 2020 had a huge performance, but immediately after all failed and delivered horror results. i still have some of the setups saved and run them again and again in the current backtest.

    04/01/2022 at 9:19 AM #190903
    GraHal

    I agree VinzentVega

    Optimise over 1,000,ooo bars or even 100k bars and then drop bars down to 10K or even 1K bars (depending on TF). Then ask yourself … would I be happy if, when I put the System Live, results are as per the 10k bars results or 1k bars results?

    In my opinion optimising over 1m bars, or even 100K bars (depending on TF) produces variable values that are not focused on price action in the very recent past and therefore not likely to give good results in the immediate future.

    In theory, Walk Forward optimising should solve the above conundrum, but, in my experience, WF is not the key to success!?

    WF might help if only we could make a choice from more than 1 result for each WF period?

    04/01/2022 at 9:42 AM #190909
    phoentzs

    You’ll laugh at what I’ve been testing for hours… I took exactly my system here as an example, because sometimes I’m undecided. I take the weekly trend as a rough filter and simultaneous exit as a constant. Likewise trading hours and time exit. The time exit in particular interrupts even rare, super good trades at some point and the system has to look for a new trade afterwards…
    Now I took time blocks of 3-6 months from the past quite randomly and optimized them. So the smaller momentum trend and money management. Then take these values and just look at the complete 200,000 candles. Lo and behold, the system always remains positive, in the past before the optimized period and also in the future. It was best if the optimized period included the extreme values ​​of Corona. And, still important to me, wins outweigh losses. So hit rate around 50%.
    Is that now a sign that the system is robust?

    04/01/2022 at 9:42 AM #190910
    VinzentVega
    GraHal wrote:

    In theory, Walk Forward optimising should solve the above conundrum, but, in my experience, WF is not the key to success!?

    That´s my experience too. WF is not the solution of the problem. Myabe only one step on the way. I think it´s better to keep the setup as simple as possible. SL not too big, TP, similar to the SL,  possibly ignore trailing completely.

    04/01/2022 at 10:05 AM #190914
    deletedaccount051022
    phoentzs wrote:

    It was best if the optimized period included the extreme values ​​of Corona.

    Completely agree.  I used to run risk models in a previous role that included what are termed as ‘stress tests’, to calculate the effect on a given portfolio that a number of scenarios would have (Scenario Margining).  For example, a dot.com crash, Brexit, Oil/war, drop in global indices, interest rate hikes etc.  What you are doing replicates that nicely.  You’re not looking for the strategy to generate a return, but more as   GraHal says, to see if you can live with the drawdown.

    The simpler the strategy, the less likelihood of curve fitting, and the more robust it will be.  However, it will likely only work in certain market conditions, so identifying these is key.  Knowing when not to trade is as important as knowing when to trade.   vschmitt  / Artificall has a great e-book which I recently purchased that covers this in further detail.  Highly recommend it, link below – I don’t get any commission by the way!  🙂 ;

    https://artificall.com/product/automated-trading-thanks-to-prorealtime-ebook/

     

    04/01/2022 at 10:15 AM #190920
    phoentzs

    So, I expect applause… 😉
    I took my strategy variant, exactly the same code as I said, the weekly trend and trading hours are constant. I optimized 12 months from September 2019 to September 2020, i.e. with Corona. Long and short, no trailing and always looking for the lowest drawdown. Never the highest performance. That would be image T1.
    Then exactly these values over 200000 bars until today. This is picture T2. Now I would like an opinion on this.

    04/01/2022 at 10:28 AM #190923
    phoentzs

    For me that means you take a system and build 2 variants from it. Optimize one over a period of time, then optimize the second system over a different period of time and use both as a portfolio. So you could be relatively sure that both together generate quite robust good profits.

    04/01/2022 at 11:02 AM #190936
    deletedaccount051022

    Very impressive.  I always try to track the correlation of the performance of a strategy, especially long only, to the performance of the underlying index it trades.  The market dips to begin with, reflected in both your strategies to a small degree, then moves upwards.  The author Kevin Davey argues that position management is almost more important, because if you are going to go long in an uptrend then the entry to buy simply gets you in, and how you then manage the position is key.  Too tight a SL then you risk being kicked out prematurely.

    04/01/2022 at 11:09 AM #190939
    deletedaccount051022

    To provide an example, here are two strategies. The top equity curve is for the MA Cross that I’ve posted before, the second is the revised version of your MACD strategy that I posted the other day.  I like that the MA Cross has stayed out of the market since mid Jan, as the filters kicked following a more extreme move downturn in the market.

    However, your MACD strategy has continued and made decent returns since then.  A long only strategy that is generating a return in a downward moving market.  Low correlation and impressive, nice work.

    04/01/2022 at 11:22 AM #190946
    phoentzs

    Yet the strategy is so simple. Completely simple and works.

    04/01/2022 at 11:35 AM #190951
    PeterSt

    You should give it to one of your Market Place friends, so they can sell it to you.

    04/01/2022 at 11:42 AM #190952
    PeterSt
    phoentzs wrote:

    I have another basic question, maybe someone has an assessment or opinion on this. Referring to my version V3 MA-Cross… I built a breakout version (theoretically worse entry price) with the same trend filter and the same system structure… then I created a version with MACD rebound (theoretically better entry price) with the same structure… and now here with MA-Cross( theoretically the golden mean from the entry price). The results of all three systems are damn similar, performance, hit rate, drawdown, performance… everything differs only slightly. That makes me suspicious, should the type of entry be relatively irrelevant in the end? Because you’re most likely to end up in the same movement anyway, so it’s more important how the position is managed. Trailing, SL, TP, etc.

    Does anyone have an opinion on this or has anyone tried the same thing before?

    FWIW : I noticed that the number of entries is relatively low. On a side note, I also noticed that in one of the later trades a huge runup occurs, representative for a lot of the gain (I recall a couple of 1000 for that one trade which also ran for ages (looked like a kind of bug to me)). Maybe this latter is less important. But :

    With such few entries, the chance exists that the entries which do occur are so strong (for any of your entry-trigger means) that they just always occur / are unavoidable.
    N.b.: This is just a response to your question above. Not a complaint of any sort (the contrary).

    1 user thanked author for this post.
  • Author
    Posts
Viewing 15 posts - 61 through 75 (of 96 total)
1 2 3 4 5 6 7

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
% distance from moving average condition
alpha-decreasing exponential moving average ( adema) from tradingview
any way to plot a time based moving average over a tick chart ?
checking for flatlining moving average
conversion of adx moving average from tradingview
dominant moving average
moving average displayed at certain time/points
moving average values are different between screener and charts
simple bb strategy that needs help
simple strategy code conversion

Create your free account now and post your request to benefit from the help of the community
Register or Login