Posts made by seb 0

@gozey
The best way to exit a trade when the direction of the market direction is about to change is at the local high for a buy trade and vice-versa.
The challenge rather is knowing when the market direction is about to change. Some patterns that can predict such things are : clear trend for a certain amount of time ; proportion of bull candles significantly higher than the bear ones ; or even the level of auto-correlation increasing sharply. These patterns will be detected by little more complex systems than a single common indicator though.

@joseph-3

By length I meant total duration(11 days), I personnaly test my strategies on few years, witha a few thousand trades.
The DD doesn't tell you about the performance of the strategy though since it is relative to your deposit, you can lower it by increasing you deposit or by decreasing lots size.
The revate/DD seems good indeed, but the sharpe ratio doesn't.

@joseph-3
You can't say if a strategy is significantly profitable or not based on a 11 days sample test only, here are some indicators of a strategy performance to take in count from a statical point of view :

• The profit/risk ratio
• The number of trades
• The length of the priod
• The regularity of the profit(slope)

0.05 is a pretty bad Sharp Ratio, to me none of these criteria are verified.

@mohamed80
The order doesn't really matter here cause the candle 1 means the previous candle(closed).

@mohamed80
Sure you can do it with with boolean variables as with double variables, this way :
https://fxdreema.com/shared/37zJjRjbb

@mohamed80
The cross up of an indicator is always preceded by a cross down, that is also always preceeded by a cross up, that is also always preceeded by a cross down... and this is right going forward as backward.
So the question you should ask rather is : How can I set a signal whenever a cross up is preceded by another cross up within a certain amount of time ?
The answer is : how many time ?

@alphaomega
Yes it isn't the first topic from roar that became nowhere to be found.. I understood he removed some, because of the misuse(from users) of some EAs he has shared, but I don't know the reason about this one.

@miki
To be honest I have no idea what the colors mean, the second line is the same as the first one until the pow brackets, but top pows are black while the others are green so..

@miki Oh really ? Mine doesn't detect any, are you on mq4 ?

@miki
No problem : )
I explained in my answer to @l-andorrà above why I did that, I just don't want negative input values to give an output number, to exclude the possibility of a signal in this situation.

@alphaomega
That is it you are on the way!
The middle brackets of the first formula are unnecessary actually.
I tried to explain how the formula is playing its role above, but it isn't easy to understand at first I know

@l-andorrà
No I didn't I just copy pasted the blocks to extract the project as ex5, the only difference..
Sure I am on FxPro

@l-andorrà
I don't know it's working fine on my platform any time

Try with "trend line", with same price2 and ray right maybe

@l-andorrà
Yes Ill explain
In the custom blocks I am actually using the RSI formula. Because it converge to 100 that is the scale on which the score is expressed.

• The first line is 100-100/pow(pow((1+rsi-Rlvl),0.5),2), lets begin with 100-100/(1+rsi-Rlvl).
  When the "rsi" value is getting important the value of the formula score is increasing also, the score so. Basically I applied the RSI formula to the RSI, I could just use the value of the RSI and interpret it as a score, but I needed it to be more sensitive in the extreme values.
  The score is beginning to be positive when "rsi" is above Rlvl-1(lets say 55) only. And the RSI formula allows a non linear value variation(increases fast at the beginning then slower and slower).
  Then I added the exponents : 100-100/pow(pow((1+rsi-Rlvl),0.5),2). Because when "1+rsi-Rlvl" is below 0 the score value is positive again and I don't want that. So with the exponents I achieved : (x^(1/2))^2 = x if x > 0, (x^(1/2))^2 = nan(ind) if x < 0 You can't square root a negative value

• In the second line : 100-100/pow(pow((1+k2*(-dif+atr)/atr),0.5),2 It's the same thing with a k2 coefficient parameter.
  With (-dif+atr)/atr, when "dif" increases(ma100 is more above ma200), the score increases also. I just expressed the "dif" relatively to the "atr".
  I also added "+atr" cause I want the score to begin being positive when ma100 slightly < ma200 because, when its the case on these periods, there still can be the MAs(ma50>ma500) reversed on a higher or lower period. And the more ma100>ma200, the more configurations of MAs will be in the same order, the bigger the score.

• In the third line : Same thing, with here, the [bottomWickSize relative to the bodySize] as "x" variable.

Then I am averaging it dividing by 3.

The thinking is the same for the opposite signals.

Finally I open a trade when the averaged score is above a certain threshold.

Here Im showing you a flexible way of applying a pretty simple RSI strategy.
https://fxdreema.com/shared/UDvtPF86
The left side of the EA is about a classic/rigid triple conditions built. The right side is the flexible version of it.
I attached below the results on the EURUSD H1 for each version for comparison.
The first version of the strategy :

With 11000 trades and 1.08 PF overtime

The second version of the strategy :

With 28000 trades and a 2.05 FP overtime
It seems the flexible version of the EA gives way more trades and a better stability overtime.

Note : The formulas are not elegant at all and a bit ambiguous so tell me if you have some questions

@roar
Good to know, then did you found that your EA was symmetric or not ?

@roar
Yes I was thinking about that, the proportion of retail traders being higher the market would be less competitive, and it would be easier to take profit from that.
Unfortunately my current broker doesn't allow to access to the cryptos

@roar
The digits pow trick is pretty smart : )
Do you think that patterns on the bitcoin are as much respected as on the forex ?

@roar
Then it would be a kinda ATR indicator, but plotted on a chart, and anticipating a potential move rather than evaluating a preexisting one. I didn't fully study the project yet but I think I get the logic.
Do you use these lines to place the exits proportionally to the size of the anticipated swing ?
Closing 1/3 of the position for each

@nestrader
Spotting candle stick patterns is still a pretty subjective approach. It's not black and white, no pattern configuration exactly look like another.
Then working with a list of rigid conditions, you have this dilemma : a) if the conditions are strict, youll have little or no signal b) if the conditions are not too strict youll have a lot of false signals.
For me the better way of finding these patterns is by working with scores. Then the only condition will be : is the score above a certain threshold ? But for that you need to be able to convert the patterns configurations into a representative score number through mathematical formulas.