Welcome, Guest. Please login or register.

August 06, 2020, 09:39:52 AM

Login with username, password and session length

Shoutbox

Recent

Members
Stats
  • Total Posts: 779
  • Total Topics: 70
  • Online Today: 17
  • Online Ever: 144
  • (August 14, 2019, 12:20:12 PM)
Users Online
Users: 0
Guests: 9
Total: 9

9 Guests, 0 Users

Author Topic: SFE General improvements after the extreme situations of the last days  (Read 1112 times)

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 301
    • View Profile
    • SmartForexExpert
Hello,

I will develop the next concept in more detail with extended explanations, but for the moment, I right now and because we are all 'excited' for good or bad with the movements of the last days, I want to ennumerate the main lines I'm working last days, which I'm sure will lead to a very good improvement of all the SFE EAs. The extreme situations are sometimes a catalyst for find good improvements, because reveals weakness points of the EAs.
 
*A* the lot size based on % risk loss are being implemented in all the EAs, including 'price action' EAs. This is changing a lot the dynamic of the EA, but in the good sense under my point of view, this approach has a lot of good properties.

*B* the long term volatility has being taking into consideration for adjust some parameters of the EAs with very good results. I begun with this on the 'reversal' EAs but at this moment I'm also finding applitacion on 'price action' EAs. The volatility has been always the main theme of any EA, but not in this long term view.

*C* the importance of the hours to open positions has been always in play, but something happens with two days of the week, Monday and Friday, and this reveal a lot of problems and possibility to fix and increase the performance of all the EAs in general.
« Last Edit: March 09, 2020, 11:11:23 PM by Attractor »

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 301
    • View Profile
    • SmartForexExpert
*A* new autolot mode: percent risk / trade.

I'm implementing this autolot mode to all the EAs in the upcoming versions. Fixed size and the ratio in relation to the balance will still exist, but probably I will use this mode in the official signals and the my personal accounts.

For example, if you use a value of 2 in this parameter, mean that the size of the trade will be the required for have a loss of 2% of the balance if the trade hits the 'initial' SL (some EAs will move the SL during the order live, so a hit of the SL can be less loss than the initial value) . Remember that in some EAs, like NS or STE, the real SL is more near than the order SL, this is for hide the real SL position, so if someone do a calculation on these EAs and the % of loss is not the correct, is for this reason.

What are main the implications of this mode:

The size of the movement is not related with the potential profit or loss. This mean: big days will not be as big, because the lot size will be very reduced, and common days will be more interesting. The profitability of the EA depends less of what happens the big days, because now, all the trades have the same potential loss and profit.

But not only this, with this mode, the different symbols are also balanced, for example, a trade in XAUUSD will not have more loss/profit than a trade in AUDUSD, while if we use the same lot size, the difference is a lot, because every pip in XAUUSD is a lot much $ than in AUDUSD.

As we know the potential loss in percentage, and the number of trades the EA can open at max or the trades opened in any moment, we can have an idea of the potential loss if all hits the SL.

Another implication, years with big big movements like 2016 vs years very flat like 2018,2019, the difference will be less. This is, you will make less profit in 2016, but more profit in 2018,2019. 2016 will be still a better year because was a lot of setups and success setups, but with this mode, the periods with low movements will have more potential profit.

As we can see, for the common trades, who search for a safety approach, this mode is more calm, because you will not have big sessions where +-25 % can happen in both directions, and you are protected to very big movements, because the trade will be tiny.

At develop and backtesting level, this mode, with this 'escalation' between trades and symbols,  is very interesting because you have a better perception for analyse the EA and his parameters, because you are not under the effect of what happend in few great sessions in the past, all the trades counts the same.

Well, I worked and tested with this mode not only in NS and STE, also in BREAKOUT and ATTR for the moment, and also for 'price action' ea I like this mode. The long term profit is also very good, although obviously the punctual results can vary a lot, but what is sure is that is less hearthbreaking and probably more safe follow the activity with this mode.

« Last Edit: May 26, 2020, 11:23:18 AM by Attractor »

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 301
    • View Profile
    • SmartForexExpert
*B* long term volatility

In relation to the problems with SFE Stealth, who have been a safe heaven until last weeks, I searched a way to filter the volatility but more in the long term view. This remove a lot of trades, but is a must in the long term, for survive to the more moved periods.

If an EA can survive without or low losses when the market is not favourable, then, there is not any problem to keep the EA until the conditions turns good again.

Just Stealth had very bad luck because the last 3 trades has been 3 full SL, bad luck because the last two, were at 1 step of the break even few minutes before the SL, but the backtest show that Stealth has a good future and can survive in more moved markets. So lets wait a bit.

This filter, also has been implemented in the upcoming version of NS, then, for first time, NS can do a bactkest since 2008 with a good look, because before this filter, the period 2008-2013 was very ugly.

Then, is the year is a lot calm, this filter can remove profits, but if our objective is the long term, is a very good addition.

The interesting, is that I'm finding also useful this filter for 'price action' EAs. price action don't have fear of movements and also want to trade when the movements are low, because is also a chance of profit, then, for example in ATTR, I used this filter not for exclude setups, but for reduce TP distance when the long term profile of the market is very calm. This way, I can hit the TP more times than the usual in years like 2018,2019, where a lot of times we see the movement go and return before hit the TP. Then, setting the TP more near the profit is bigger in this kind of market.
« Last Edit: March 10, 2020, 08:40:34 PM by Attractor »

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 301
    • View Profile
    • SmartForexExpert
*C* Week Days

select the session / hours to open trades is very important in any system. The need of do a own selection of hours has been demanded, but in some EA is very complicated because if the system have several subsystems, symbols, timeframes, the way of backtest or define all this is not simple. I will have this in consideration and if is possible, will be implemented in future upgrades, but for the moment, the hours will be settled internally in the upcoming upgrades in the next days.
Last days a checked again in different EAs the hours, and all have sense, so I'm quite confidence in the way the internal hours are settled.
Some members has noted that for example in the NY session SFE Attr works bad. NY is the main session for any 'price action' EA, so, we cannot get rid of this session. For 'price action' EAs, each symbols will trend to be traded in his own sessions, then, for major part of symbols, NY session is  a must.
But reviewing this, I noticed that the problem is related with the weekdays:

MONDAY: is the weakest day in movements of the week, and almost always without new events. Is the day with less signals settled, and usually the movements are not enough strong to make the necessary profit. Then, in a good number of EAs, I'm seeing that use Monday is useless, they are  not making profit on the monday trades.

FRIDAY: is not weak, but is the most mad day of the week. The most zizgag movements happens on Friday. Is the close of the week with the close of the EEUU stock market, and last hours forces balance and makeup are common. In SFE Breakout EA, Friday is disabled for major part of symbols, and recent test on Attractor shows that Friday is a bad day for the 'straight' subsystems, while 'counter ' sysbystems can be still profitable for this kind of failing movements.
 
As the oldest SFE 'price action' systems did born in the last very good years of price action, almost every day of the week including Monday and Friday were profitable, but in the long term, and understanding the kind of market these two days, can be a smart choice don't use them or use them very selective, and in some systems they aren't profitable, which make sense at final.

The introduction of the 'scaled' trades (see point *A*)  once I'm using now 'risk % based' for the lot size of the trades, give a better idea of what is profitable and what not, because the stathisics aren't deformed for some singular days. I mean for example, you can get that GBPUSD is profitable at 0 hours on monday, because without scale the trades, 1 time happened a very big movement after an election process, and in 10 years, this only movements back the profitability of this symbol / hour / day. But if you scale the trades, this symbol / hour / day is not profitable. And this last approach is the correct, because the event of the example is a vey exceptional event, and maybe will not repear again in 50 years just at this day and hour.

At the same time, use less Friday, will traduce in very reduced number of trades over the weekend, which will be very healthy for our rest the weekend.

SFE Attractor 2.0: Monday has been disabled in general. And Friday has been disabled for 'straight' subsystems. Either long term backtest and current live performance back this decision.

Attractor

  • Administrator
  • Sr. Member
  • *****
  • Posts: 301
    • View Profile
    • SmartForexExpert
The recent implemented autosize mode risk/trade, worked until now for symbols defined as 'forex' kind in the properties of the symbol in the mt platform. That was because the platform can have symbols defined with other types, like cfd, future.... and for some reason, is risky try to get the size of the trade in some kinds of types, because this calculation based on the parameters that are given by mt can be wrong if the parameters of the symbol have some error. The parameters of the symbol are settled by each broker, and sometimes errors has been detected. THis has been detected in symbols non 'forex' kind, for this reason until now, was disabled return autolot risk/trade for non forex kind.

At same time happened that some brokers has the XAUUSD symbol defined as 'cfd' instead of forex, and the EAs were returning now the minimal size, because autolot risk/trade was disabled for this kind as I just explained.

Then, I have thinking in a workaround which is the next:  now autolot risk/trade, returns always a size based on the parameters of the symbol, but for symbols not defined as forex kind, where exist a risk that the calculation is wrong because the symbol can be defined wrong, the EA does the next check on the openend orders:

- when a order is opened, and the trade move some pips, which is computable is the current profit or loss. There don't exist errors in the mql returned values. Then I do the next: I compare the theorical loss at the beginning of a new bar of the trade with the real loss, and if this value is very different, means the autolot function has failed to calcuate the value (for the problem of the broker defining the parameters of this symbol)

We can see this problem in the next image ( a case failure, in XAUUSD on Alpari) :
1. the max. theorical loss.
2. the real loss reported by the platform.
3. the theoric loss with the current movement, the loss that the EA has worked with.
Real loss (2) is much high than (3), therefore the size is not correct.
Then proceeds to launch an alert message (4) and close the trade (5).



The next updates of the EAs will implement this new kind of procedure in relation to the size (risk/trade).

Under the point of view of the user:   use the EA  normally, if for some reason, in your broker, xauusd or other defined symbol different than forex gives an error in the calculation, the trade will be closed and you will see a message in the alert panel.
Then, you can't use this symbol with this autolot mode unless the broker fixes to correct values in definition of the symbol, but at least the trade  has been closed and you are notified about this risk.

The brokers I use, ICMarkets, FPMarkets, Tickmill, Darwinex, manages XAUUSD correctly and will be very uncommon find a broker with this problem, but for not have the possibility of a big loss if this problem happens in your broker, this mechanism has been implemented.

If someone want to know this before go to live, this also can be seen in the backtest, is not a live trading issue, happens in the backtest already. You only have to launch the backtest, and the inspect the log for find messages containing the words of the image above. If no lines are returned by the log with this kind of message, means all is correct.
« Last Edit: May 18, 2020, 11:56:14 PM by Attractor »