MT4 List or Array? | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Comprehensive list of results list of MT4 EAs and indicators 0 replies

Two Dimensional Array vs One Dimensional Array 13 replies

MT4 - Indicator Array Problem. 0 replies

MT4 array funkiness... wtf? 4 replies

Calculate Stochastic Formula on a array (non indicator buffer array) 2 replies

Attachments: MT4 List or Array?
Exit Attachments

MT4 List or Array?

  • #1
  • First Post: Edited 8:53pm Jun 10, 2014 8:11pm | Edited 8:53pm
Which is better / easier.... list or array?

Ex: I have 4 separate entry conditions and want to test if any is better than the others or if a combination is better.

Buy if A
Buy if B
Buy if C
Buy if D

and additional combinations, for ex:
Buy if ABCD
Buy if AC
Buy if BC
etc.

Since there are 4 bool conditions, there is a total of 16 combinations and15 possible buy conditions from FFFF to TTTT (16 minus FFFF).

Which is more effective a list or an array? Is one more effective than the other? Is one easier to implement than the other? In another EA I used a list and had over 30 if statements per side for each possible buy or sell.

What's the most logical approach for an array in this case?
  • #2
  • Jun 10, 2014 9:13pm Jun 10, 2014 9:13pm
  •  FerruFx
  • Joined May 2007 | Status: MT4/MT5 EAs/Indicators/Alerts coder | 6,542 Posts | Online Now
You don't need to check all possibilities. Just invalid your trade as soon as 1 of the conditions is wrong.

Inserted Code
BUY = true;
if(Use_A && !A) BUY = false;
if(Use_B && !B) BUY = false;
if(Use_C && !C) BUY = false;
if(Use_D && !D) BUY = false;
 
if(BUY) send your buy order
MT4/MT5 EAs/Indicators/Alerts coder
  • #3
  • Jun 10, 2014 10:17pm Jun 10, 2014 10:17pm
Quoting FerruFx
You don't need to check all possibilities. Just invalid your trade as soon as 1 of the conditions is wrong. BUY = true; if(Use_A && !A) BUY = false; if(Use_B && !B) BUY = false; if(Use_C && !C) BUY = false; if(Use_D && !D) BUY = false; if(BUY) send your buy order
Thanks for the input. I have to wrap my head around that one and work through the logic.

Your coding input is always spot on and I appreciate it. I just need to think this one through. It's not something I thought of or seen before.

You say I don't need to check all possibilities, but from a testing viewpoint I was going to set up 4 bools and 15 separate if statements containing all possibilities of combinations in a list:

Inserted Code
extern bool UseCondition1 = true; // test T and F
extern bool UseCondition2 = true; // test T and F
extern bool UseCondition3 = true; // test T and F
extern bool UseCondition4 = true; // test T and F

Testing all 4 bools simultaneously gives 16 combinations of course F F F F would not generate anything so no corresponding if statement for that one;
  • #4
  • Jun 10, 2014 10:53pm Jun 10, 2014 10:53pm
  •  FerruFx
  • Joined May 2007 | Status: MT4/MT5 EAs/Indicators/Alerts coder | 6,542 Posts | Online Now
Not sure to see what you are trying to do. Not so clear explanation (at least for me !) ...
MT4/MT5 EAs/Indicators/Alerts coder
  • #5
  • Jun 11, 2014 12:57am Jun 11, 2014 12:57am
Quoting Pipologist
Which is better / easier.... list or array? Ex: I have 4 separate entry conditions and want to test if any is better than the others or if a combination is better. Buy if A Buy if B Buy if C Buy if D and additional combinations, for ex: Buy if ABCD Buy if AC Buy if BC etc. Since there are 4 bool conditions, there is a total of 16 combinations and15 possible buy conditions from FFFF to TTTT (16 minus FFFF). Which is more effective a list or an array? Is one more effective than the other? Is one easier to implement than the other? In another EA I...
think binary....
  • #6
  • Jun 11, 2014 2:33am Jun 11, 2014 2:33am
Quoting Pipologist
Which is more effective a list or an array?
If you know the maximum possible outcomes, and you do not need to insert elements at arbitrary positions, but only for example at the end, then an array would be more efficient.
Return To Disorder
  • #7
  • Jun 11, 2014 2:48am Jun 11, 2014 2:48am
Quoting FerruFx
Not sure to see what you are trying to do. Not so clear explanation (at least for me !) ...
I have four separate simple buy signals. They are set as boolean.

Each signal on it's own can generate a signal which equals 4 possibilities for testing.

When all possible signal combinations are combined there are 16 possibilities (4 to the power of 2). 4 bools combined equals 16 signal possibilities.

Since the first one is 0 0 0 0 or F F F F, that would not generate anything. There are 15 possible buy signal combinations.

For example,

T F F F // Signal A only
F T F F // SignalB only
F F T F // Signal C only
F F F T // Signal D only
T T F F // Signal A and B
T F T F // Signal A and C
T F F T // Signal A and D
T T T F // Signal A and B and C
T T F T // Signal A and B and D
etc.
etc.

I want to run a backtest to see which signal combination works best over a certain time period for a certain symbol.

So I have UseSigA, UseSigB, UseSigC, UseSigD as testing switches each boolean.
and I have

if (UseSigA)
SigA = SigAformula;

repeated for each signal

then

if (UseSigA && SigA)
buy;

if (UseSigB && SigB)
buy;

etc.

and

if (UseSigA && UseSigB && SigA && SigB)
buy;

etc.

this is done for all possible 15 combinations.

I'm pretty sure I could do it in a list as above and am working on it now.

I was wondering if an Array would be better and how so, and how to implement it or at least where to start if using an array.

Most of the time when faced with a challenge like this, I just do binary solutions with a massive list of combinations.

I was thinking that an array might work?

I've never worked with an array so I'm totally ignorant when it comes to an array, but was thinking they could be more powerful and save me time after learning them.

The more combinations of binary possibilities, the longer the list and the greater potential for coding errors, like repeating a combination or not including one.
  • #8
  • Jun 11, 2014 3:03am Jun 11, 2014 3:03am
Quoting rockit
{quote} If you know the maximum possible outcomes, and you do not need to insert elements at arbitrary positions, but only for example at the end, then an array would be more efficient.
Ahhh...ok;

I know the maximum possible outcomes.

I'm not sure if I need to insert elements at arbitrary positions. For now I think I don't.

Just want to check 16 possibilities and test and see which if any produces a significantly better outcome.
  • #9
  • Jun 11, 2014 3:15am Jun 11, 2014 3:15am
  •  FerruFx
  • Joined May 2007 | Status: MT4/MT5 EAs/Indicators/Alerts coder | 6,542 Posts | Online Now
Quoting Pipologist
{quote} I have four separate simple buy signals. They are set as boolean. Each signal on it's own can generate a signal which equals 4 possibilities for testing. When all possible signal combinations are combined there are 16 possibilities (4 to the power of 2). 4 bools combined equals 16 signal possibilities. Since the first one is 0 0 0 0 or F F F F, that would not generate anything. There are 15 possible buy signal combinations. For example, T F F F // Signal A only F T F F // SignalB only F F T F // Signal C only F F F T // Signal D only T T...
After A,B,C,D calculated, the 4 lines of code in my post #2 will give you the same result than what you are trying to do with multiple checking ...
MT4/MT5 EAs/Indicators/Alerts coder
  • #10
  • Edited 3:43am Jun 11, 2014 3:21am | Edited 3:43am
Why do you think you need a list or array?
If I understand this well, I would do:
Permute all the possibilities, and at every permutation, if the result is better than the last one, store its 4 boolean values mulitiplied by a (binary) weighting as a sum in a variable 'index' (i.e. true=1, false=0, A=8, B=4, C=2, D=1; meaning if only A is met, index would yield 8, if A&&B index = 12 etc.). At the end 'index' is a pointer to (or index into) the best combination.
Return To Disorder
  • #11
  • Jun 11, 2014 3:49am Jun 11, 2014 3:49am
Quoting rockit
Why do you think you need a list or array? If I understand this well, I would do: Permute all the possibilities, and at every permutation, if the result is better than the last one, store its 4 boolean values mulitplied by a (binary) weighting in a variable 'index' (i.e. true=1, false=0, A=8, B=4, C=2, D=1; meaning if only A is met, index would yield 8, if A&&B index = 12 etc.). At the end 'index' is a pointer to (or index into) the best combination.
Thanks for the idea. I wouldn't know how to implement that but it got me thinking...

Since there are only 15 combinations of buys, just have a extern int of "buysignalvalue" 1 through 15.

And then, show either

if (buysigvalue == 1)
buysigA formula

or

if (buysigA)
buysigvalue = 1

repeat for each combination of buys and sells

and then run test with buysignalvalue 1 to 15 and sellsignvalue = buysigvalue

I know this way may seem redundant to you rockit and ferru but it's the only way I know how to do it and get immediate results. The other ways may be quicker to code and easier to import into other eas and similar logical problems but they're a little beyond my immediate programming skills.

I will remember them and try to get them to work when I have some more time.

Thank you all three of you rockit and ferru and max. Even if I can't immediately implement your suggestions they are valuable and I'm sure there are others here on FF who read these and use them as well.
  • #12
  • Edited 8:21am Jun 11, 2014 6:53am | Edited 8:21am
I think I got it now. You have to permute all the signal combinations. Note that A..D (a.k.a. signals) have to be 0 or 1, not true or false:
Inserted Code
prev = result = 0
for a = 0 to 1
    for b = 0 to 1
        for c = 0 to 1
            for d = 0 to 1
                if !(A^a) && !(B^b) && !(C^c) && !(D^d) 
                    result = trade()
                    if result > prev
                        prev = result
                        combin = a*1 + b*2 + c*4 + d*8
 
chrarr = {'A', 'B', 'C', 'D'}
sigstr = ""
for i = 0 to 3
    if combin & pow(2, i)
        sigstr += chararr[i]
 
Print "winning combination: " + sigstr
Return To Disorder
  • #13
  • Edited 10:39pm Jun 11, 2014 7:08am | Edited 10:39pm
Quoting rockit
I think I got it now. You have to permute all the signal combinations. Note that A..D (a.k.a. signals) have to be 0 or 1, not true or false: prev = result = 0 for a = 0 to 1 for b = 0 to 1 for c = 0 to 1 for d = 0 to 1 if (A*a + B*b + C*c + D*d) > 0 result = trade() if result > prev prev = result combin = a*1 + b*2 + c*4 + d*8 chrarr = {'A', 'B', 'C', 'D'} sigstr = "" for i = 0 to 3 if combin & pow(2, i) sigstr += chararr[i] Print "winning combination: " + sigstr
I haven't coded that way before so I have to really practice and learn it.

But your comments got me thinking even more about this. There is one extra permutation of "Buy Signal" per letter combination (using just 8 in our example would be 8 + 1). Mathematically there are 8 combinations using A. However logically from a possible trade standpoint there would be 9. The last one is a simple "if A of any combination". So the last plus 1 permutation would look like this, this is in addition to the individual 8 each:

if (A1 || A2 || A3 || A4 || A5 || A6 || A8 || A8)
Buy;

If it was looking for permutations of A B, then it would be 4 plus 1.
if (AB1 || AB2 || AB3 || AB4)
Buy

So it would have to check all 8, or all 4. or all 2 permutations per letter combination plus one in each case since one could say "buy if any of the cases of A is true" or "buy if any of the cases of AB are true".

So each letter would have a total of 33 permutations, but after the first set, several of the permutations would be repeated.

The most simple way for me, and redundant, repetitive and time consuming would be to create a list of 4x33 buy scenarios and just give them a unique integer between 1 and 132 and then run the backtest. Since these are no indicators to be tested the test should not be that cpu intensive.
  • #14
  • Edited 8:26am Jun 11, 2014 7:24am | Edited 8:26am
I do think that you would need an array (and more). To store results for every combination. At the end you would look up the combination with the best result.

Sorry I do not understand the 8+1 concept concerning the results. It is 8 combinations per letter, all else is up to statistical analysis.

Edit: also I think the test (in my code above) has to be like so: if(!(A^a) && !(B^b) && !(C^c) && !(D^d)) {trade()}
Return To Disorder
  • #15
  • Jun 11, 2014 10:30am Jun 11, 2014 10:30am
Quoting rockit
I do think that you would need an array (and more). To store results for every combination. At the end you would look up the combination with the best result. Sorry I do not understand the 8+1 concept concerning the results. It is 8 combinations per letter, all else is up to statistical analysis. Edit: also I think the test (in my code above) has to be like so: if(!(A^a) && !(B^b) && !(C^c) && !(D^d)) {trade()}
As you said there are 8 possible combinations per letter mathematically. However, if you add these up into possible buy scenarios, there are more combinations.

Here's an example:

For A combinations:

Buy if

1. A O O O
2. A B O O
3. A O C O
4. A O O D
5. A B C O
6. A B O D
7. A O C D
8. A B C D

There is buy if ANY of the above not just 1~8.
9. if (A1 || A2 || A3 || A4 || A5 || A6 || A8 || A8) which would produce significantly more signals.

# 2, 5, 6, and 8 are variations of AB
but there is also buy if ANY of 2,5,6,or 8 which would produce more signals than just 2, 5, 6, 8 alone.

etc.
  • #16
  • Jun 11, 2014 10:51am Jun 11, 2014 10:51am
So I'll make a spreadsheet with EVERY possible buy combination tomorrow and then delete duplicates which should give me the real total of buy combinations.

I think it's more than 15 not including zero. Mathematically in an array there would be 15, but in actuality there are more.
  • #17
  • Jun 11, 2014 12:57pm Jun 11, 2014 12:57pm
Quoting Pipologist
As you said there are 8 possible combinations per letter mathematically. However, if you add these up into possible buy scenarios, there are more combinations.
No, there are not.
There are 16 in total (15 that is).
And they can all be addressed in 4 nested loops, as I have put above (I just got the test wrong at first; one has to use xor and and and not operators).
Return To Disorder
  • #18
  • Edited 10:26pm Jun 11, 2014 6:42pm | Edited 10:26pm
Quoting rockit
{quote} No, there are not. There are 16 in total (15 that is). And they can all be addressed in 4 nested loops, as I have put above (I just got the test wrong at first; one has to use xor and and and not operators).
At first I thought there were 15 (16 minus F F F F, or O O O O).

From a mathematical perspective that would be true, since there are 4 binary combinations... 2x2x2x2

However, there are other tradable combinations...for example for the AB combination:

You could say "buy if ABXX", or "buy if ABCX", or "buy if ABXD" those are 3 separate outcomes from the original 15 (16).

You could also include simply "buy if AB", which would mean "buy if ABXX || ABCX || ABXD" which would generate many more trades than testing one each of the possible three against each other, so in reality there are 4 possible combinations of AB not just 3

If you extrapolate this out with all of the combinations, for example "A":
test if (AXXX) (TFFF) >> Buy
test if (ABXX) >> Buy
test if (AXCX) >> Buy
test if (AXXD) >> Buy
test if (ABCX) >> Buy
test if (ABXD) >> Buy
test if (AXCD) >> Buy
test if (ABCD) >> Buy

For possible "A" combinations there would be 8 above plus an additional tradable combination that in reality should be tested as well. Because "buy if A", could mean either "buy if AXXX (AFFF, if using T/F)", or it could mean "buy if any combination of A", which would mean

"if (A || AB || AC || AD || ABC || ABD || ACD || ABCD) >> Buy" this is a very tradable outcome, but not included in the originally calculated set of 8.

So, "buy if A", has not 8, but 9 tradable situations.

"Buy if A" could mean any of the 9 or just the 9th one, whereas" buy if only AXXX" could only mean the first combination.

"Buy if AB" could mean "buy if ANY AB", or it could mean "buy if ONLY ABXX".

Extrapolating all of the possible TRADABLE outcomes A, B, C, D and derivatives and then deleting for duplicates gives me 29 possible tradable outcomes, each generating a different set of trades over time.

I'm not familiar with nested loops and how your code works but looking at the multipliers 8, 2, 1 I think the code only allows for the static "OR" outcomes and those would add up to 15.

Maybe your code accounts for the 9th condition and am just not seeing it?

If it doesn't account for it, I'm pretty sure that in testing and trading, it will yield different results.

condition 1: if (AXXX) (TFFF) >> Buy
condition 2: if (ABXX) >> Buy
condition 3: if (AXCX) >> Buy
condition 4: if (AXXD) >> Buy
condition 5: if (ABCX) >> Buy
condition 6: if (ABXD) >> Buy
condition 7: if (AXCD) >> Buy
condition 8: if (ABCD) >> Buy
condition 9: "if (AXXX || ABXX || AXCX || AXXD || ABCX || ABXD || AXCD || ABCD) >> Buy"

I posted a pdf of the spreadsheet with ALL combinations per signal letter and deleted the duplicates.
Attached File(s)
File Type: pdf 4c Outcomes.pdf   35 KB | 905 downloads
  • #19
  • Jun 11, 2014 10:44pm Jun 11, 2014 10:44pm
Anyway, I think knowing all the possible trading scenarios before testing is more important than considering using a list, nested loop or an array.

I know how to code this out list wise.

The other coding types I will study for possible future implementation.

Thanks for contributing to this thread. It really opened my eyes to my potential trading logic errors before I start coding it and testing it.
  • #20
  • Jun 12, 2014 12:54am Jun 12, 2014 12:54am
Quoting Pipologist
Anyway, I think knowing all the possible trading scenarios before testing is more important than considering using a list, nested loop or an array. I know how to code this out list wise. The other coding types I will study for possible future implementation. Thanks for contributing to this thread. It really opened my eyes to my potential trading logic errors before I start coding it and testing it.
you are still going about it the hard way imo

test the outcomes just once and generate a unique number using binary, then simply use a case statement for the different outcomes.

in this way you only test the scenarios once and do not have to do loads of IFs with condition checking. simplest,quickest, most efficient way of doing this
0 traders viewing now