optimising gives different results each time I do it - page 2

New comment
 
No. I can pick any dates. I don't use today. I am puzzled.
 

What build are you using and by which broker is the software provided?

Do you only have one installation of MT4 on just one PC?

I ask this, because you could install MT4 again to a different folder (or another PC) and see if the new installation is also having the same problem or not.

Preferably use the software provided by MetaQuotes directly or from a different broker, just in case it is the broker's version of the software which is at fault.

If the new installation works properly, then you would know that something is wrong with the original installation.

If however, you continue to have the same problem, then consider that maybe you might be doing something wrong.

 

I know that this thread is probably cold, but I just remembered one point that I constantly run into but had forgotten.

I remembered today because I ran a back-test for someone else's code and had a similar issue. In my own code I simulate the value of MODE_TICKVALUE in back-testing, but in this case since it was another's code, it used the value directly.

Unfortunately, in back-testing the value it provides is the current value and not the one that should based on the prices being simulated. What happens is that every time you run the tester and in the case that the code does proper money management based on a percentage of the balance and it uses the Market-Info for MODE_TICKVALUE to do it, the value returned will differ if current prices have changed.

In my own code, since I simulate this value using the BID to calculated the inferred MODE_TICKVALUE, I do not usually see this problem.

Maybe, this is the case for you "cciguy"!

PS! I was wondering how other people here resolve the MODE_TICKVALUE issue in their cases!
Does anyone here have a general purpose solution or do you also use the BID on case to case solution?

 
FMIC:

I know that this thread is probably cold, but I just remembered one point that I constantly run into but had forgotten.

I remembered today because I ran a back-test for someone else's code and had a similar issue. In my own code I simulate the value of MODE_TICKVALUE in back-testing, but in this case since it was another's code, it used the value directly.

Unfortunately, in back-testing the value it provides is the current value and not the one that should based on the prices being simulated. What happens is that every time you run the tester and in the case that the code does proper money management based on a percentage of the balance and it uses the Market-Info for MODE_TICKVALUE to do it, the value returned will differ if current prices have changed.

In my own code, since I simulate this value using the BID to calculated the inferred MODE_TICKVALUE, I do not usually see this problem.

Maybe, this is the case for you "cciguy"!

PS! I was wondering how other people here resolve the MODE_TICKVALUE issue in their cases!
Does anyone here have a general purpose solution or do you also use the BID on case to case solution?

Are you aware that we generally tell people to dis-connect their testing environment so that their Tick_Value does-not keep changing?

If yes, why aren't you doing so?

 

Because I want to simulate the "real" TICKVALUE and not some fixed value based on current prices!! That is why, I simulate my own TICKVALUE based on the BID prices.

For example, if I am testing the EURUSD and the Account is in EUR (on a 5 digit broker) then the TICKVALUE is 1/BID!

Another example, if I am testing the USDJPY and the Account is in USD (on a 3 digit broker) then the TICKVALUE is 100/BID!

By BID, I mean the simulated BID price of the currency pair being tested. This way the testing is much more accurate than having a fixed TICKVALUE for the entire test!!!

Besides, I am not the one with the problem, "cciguy" is the one having the problems, not me. I am just letting him know that this may be one of the reasons for his issues.

I just wondered how many others use a solution like mine or maybe even something better and more general. "Ubzen", from your answer I can infer that you use the "fixed" solution which is less accurate.

 
FMIC:

1) Because I want to simulate the "real" TICKVALUE and not some fixed value based on current prices!! That is why, I simulate my own TICKVALUE based on the BID prices.

2) For example, if I am testing the EURUSD and the Account is in EUR (on a 5 digit broker) then the TICKVALUE is 1/BID! This way the testing is much more accurate than having a fixed TICKVALUE for the entire test!!!

3) Besides, I am not the one with the problem, "cciguy" is the one having the problems, not me. I am just letting him know that this may be one of the reasons for his issues.

4) I just wondered how many others use a solution like mine or maybe even something better and more general.

5) "Ubzen", from your answer I can infer that you use the "fixed" solution which is less accurate.

1) Why stop there ... do you also simulate every other SymbolInfo() variables like Margin_Required || Cross-Currency Rates... etc? Example Here.

2) So you consider the Bid and dis-regard the Ask? Do you really believe thats more Accurate?

3) cciguy is having issues with in-consistent results from test-to-test, he isn't having issues with accuracy of tick_value.

4) Did you search the forum to find out how many other people tried that solution?

5) I've used both the fixed and simulating every Symbol_Info Variable. If its accuracy of simulation I'm looking for from my back-tests, I'd built my own strategy tester which includes real-tick data (ask and bid), create every Symbol information, simulate slippage and latency. etc.

What I do-not do is encourage someone who's having in-consistent results to keep changing the environment. If that person say I understand why my results are changing, and I want it that way ... then thats fine. Just don't ask me why "my back-tests keep changing" ... I think thats a fair proposition.

 

"Ubzen", now you are being paranoid! Did I do something to upset you in order to get you all worked up?

All I suggested to "cciguy" is that in case he was using the TICKVALUE in his code and since it does change from test to test, that it might be the reason for the different outcomes.

So please, "ubzen", stopping being difficult and if you have some beef with me then just say so and do it in private.

Otherwise, this conversation is over from my part!

Goodbye!

 
FMIC:

"Ubzen", now you are being paranoid! Did I do something to upset you in order to get you all worked up?

All I suggested to "cciguy" is that in case he was using the TICKVALUE in his code and since it does change from test to test, that it might be the reason for the different outcomes.

So please, "ubzen", stopping being difficult and if you have some beef with me then just say so and do it in private.

Otherwise, this conversation is over from my part!

Goodbye!

I didn't call you names, I asked you questions, and I tried to answer your questions referring to myself.

If I had beef with you, I would say I have beef with you ... trust me.

Here's another question.

I wonder what give you the idea that I'm Paranoid, or have Beef, or being Difficult, or Worked-Up?

12
New comment