Scalping, how stable is such a system?

Forums › ProRealTime English forum › General trading discussions › Scalping, how stable is such a system?

Viewing 15 posts - 1 through 15 (of 26 total)
1 2
  • 07/18/2021 at 8:40 PM #173776
    phoentzs

    As a general question to the professionals here … at the moment I am trying harder to use scalping systems in M1 and M5 on DAX and CAC40 … with results around 90% and an average stop loss of less than 1%. It’s basically a simple trend following with fairly simple input signals. Without many filters, but also intraday StopandRevers. With quite good success also in the demo account. Has anyone experience with systems that have a small TP and relatively large SL in terms of stability? How often should such systems be updated? I always create them WF70 / 30, which is about 6 weeks OOS with an M1 system. Seems to work fine. Maybe once a month?

    07/24/2021 at 5:46 PM #174096
    robertogozzi

    I personally don’t like SL’s greater than TP’s. My worst case allowance is a 1:1 ratio, as I have experienced bad results.

    A strategy with an 800-pip SL and a 400-pip TP may be backtested with a winning % of nearly 100%, but in the future it will happen, maybe 3-4 times in a row, that your SL is hit (which backtest could not predict) and you will lose ALL you had gained so far!

    I have experienced that lowering that SL from 800 to 200 (half the TP, thus with a 2:1 reward ratio, instead of 1:2), lowered the winning rate to slightly below 80%, still granting enough gains (though smaller than before).

     

     

    07/24/2021 at 7:39 PM #174098
    phoentzs

    @robertogozzi

    Thanks for your answer Roberto. Which system with a reward ratio of 1: 2 has an 80% hit rate? 😉 I am currently running various attempts with my system in the demo and come to 90% as a scalper and with a reward ratio of around 1: 2 to around 65% hit rate. Do you have any advice on how to improve this? I have to say that it only trades in the DAX and does not hold any positions overnight. So the open positions are closed at the latest in the evening at the close of trading. Everything M5 or M1.

    07/25/2021 at 10:22 PM #174143
    XORANDNOT

    Small tp and large sl enable you to optimize away perfectly most of the losing positions in a backtest, just by adapting some major parameters, walk forward or not. Because tp occurs rather often and sl pretty seldom. So it’s easy to optimize away the few remaining large stop losses.

    In reality, which is not optimized, the large stop losses will reappear much more frequently and destroy the performance of your great optimized system. This is my whole experience.

    So – for scalping systems as automated trading I’d say : better keep your fingers away from them.

    07/26/2021 at 10:57 AM #174175
    XORANDNOT

    Addendum : scalping systems may work well for a while, when volatility is low, because then a large stop loss is seldom or never reached. But as soon as volatility comes back, they stop working, because the large stop losses are triggered more often and pull down the system’s profit. You will always have some positions that never run into profit, but turn in the wrong direction right after opening, and when volatility is high, they run into stop loss. “Re-optimization” will not help, except for making backtests look better.

    07/26/2021 at 11:24 AM #174178
    phoentzs

    Mmh, I noticed that last week. When the DAX made larger, abrupt movements, my scalpers caught one or two SLs. While the systems with a larger TP to smaller SL ratio were winners. Although the systems were the same, the only difference is the TP / SL ratios on them.

    07/27/2021 at 5:12 PM #174247
    phoentzs

    I still have a little question: is there anyone who has successfully run M1 systems in the DAX? I keep seeing successful M5 systems, sometimes M2 systems, but never an M1 system. Is there a reason for that? I have M5 systems that work well and now set my M1 systems to SL larger TP. I don’t use the scalpers. The trials are now running in demo. Is it worth trying M1 systems further? The back test is impressive.

    07/28/2021 at 10:43 AM #174283
    XORANDNOT

    Backtests on very short time scales are always impressive, because you have more data points to do the curve fit, even more so on the 10 second time scale.

    I have had a few 1 minute systems that worked quite well in reality, although of course never as good as in the backtest. On the 10 second time scale, break even was a good long-term result in reality.

    07/28/2021 at 12:26 PM #174290
    phoentzs

    In the 10-second chart? With trailing code or SL / TP? Probably only in the morning 09: 00-11: 00?

    07/28/2021 at 1:03 PM #174291
    XORANDNOT

    Well, trading on such short sime scales is only intraday. Entry in the morning from 9:00-9:15 to about 11:00-12:00 or maybe until 14:00. Exit until 18:00. In the afternoon, profits are usually smaller or non-existent. In the DAX, I have stop losses of about 80 points (because the DAX often reverses completely after an excursion in the wrong direction and often halts at full 100s). Tp is about the same size, plus some mechanisms to take profits below tp when the end of a trend can be detected. Very small profits or losses other than from sl are not allowed, except at the end of the day. So, basically, you wait until a position runs into profit by chance and take it when either tp is reached or the end of a trend is detected. A trailing stop for higher profits is not bad either, something like : when the highest high was at least 50 points above entry, set the trailing stop 25 points below all following highs.

    Optimization in the 10 second chart takes a long time because of the many data points, but usually only leads to highly curve-fitted systems. Be careful not to overfit and use as few parameters as possible. It is easy to write sensationally-looking backtests on these time scales !

    07/28/2021 at 1:17 PM #174292
    GraHal
    XORANDNOT wrote:

    because of the many data points

    Are you referring to bars as data points or something else?

    Re data points … how / why is 100k bars on S10 TF different from 100k bars on for example M5 TF?

    Just interested in case I am missing something!? 🙂

    07/28/2021 at 1:22 PM #174293
    phoentzs

    Sounds a bit like a casino. Sensational back test, ok. And does it work halfway in reality?

    07/28/2021 at 1:46 PM #174294
    XORANDNOT

    Well, trading (especially daytrading) is a casino. You can’t predict anything, even if many thick books try to suggest so. It works during some periods, and it does not in others, for example with very high or very low volatility.

    You can always tell exactly what has happened to an open position, but you can never predict the future.

    07/28/2021 at 1:48 PM #174295
    XORANDNOT
    GraHal wrote:
    XORANDNOT wrote:

    because of the many data points

    Are you referring to bars as data points or something else?

    Re data points … how / why is 100k bars on S10 TF different from 100k bars on for example M5 TF?

    Just interested in case I am missing something!? 🙂

    On the 10 second time scale, you have 30 times more bars (= data points) compared to the same period on the 5 minute scale. And therefore, you can curve fit your backtest much better.

    07/28/2021 at 2:17 PM #174296
    phoentzs

    This of course explains why my M5 systems have very similar performance. The basic system without any trend filter is almost as good as my variant with a trend. The only thing that stands out is my variant with a momentum filter. But it’s somehow different in the big books.

  • Author
    Posts
Viewing 15 posts - 1 through 15 (of 26 total)
1 2

Find exclusive trading pro-tools on ProRealCode Marketplace

Similar topics:
3-10 oscillator swing trading system
a completely simple nas macd system
auto tradeing system
cascade ordering system
conversion scalping line tradingview
different instructions for an intraday trading system
expiration date for my system
i cant close system/position in real
is there such a strategy existing in autotrading?
long time to run system code causes rejection?

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