How do you achieve a qualitative leap in market analysis? There is an option: - page 3

 
If a strategy opens and closes a single position within a minute candle, the tester will not show real results. I completely agree with you. But if there is a reversal, i.e. the current position is closed and a position is opened in the other direction at the same price as the previous one. Then such a one-step action can be estimated by the tester to be close to true. If someone decides to close their position, it means that they believe that the price will go in the opposite direction. Finding the conditions for flips is an interesting thing. What was presented is a flip system, i.e. it is in the market all the time.
 
Well, such a rollover system can be tested reliably enough on 15-minute data as well. The system, if it wants to be robust, has to be rough to minute data, and even more so to random ticks. What is the history of ticks then?
 
So in order. 1. There are 1-minutes, then why 15-minutes? 2. The higher the timeframe, the more information we lose about the real "price" behaviour. In general, I analyzed and developed an algorithm on a 10-second history and what is available in the History Center. I did not use the tick history, it is too individual for each broker. Although I may be wrong.
 
Mathemat, what test do you think you need to do to make a correct picture of any system. It's interesting, really, to hear your point of view. Let's be more specific, "trying is not torture": I want to find a bug in the system myself.
 
I have expressed the wrong idea. What I meant was that all of them should be available, but it would be better to test the system on TFs less than 15 minutes (previously converted minutes according to instructions in the file). Then the system will have at least some roughness to random ticks, that is stability and some confidence in the results. For this reason I do not think it is reasonable to have the history lower than minutes.

1. There are 1-minutes, then why 15-minutes?


I do not know exactly, but I can suggest that it is intended just to speed up calculations when testing on TFs larger than 1 minute.
 
The system has sense to test it on some timeframe, if it uses indicators, which are drawn just by bars. If the system has no indicators, but the situation is estimated only by the Bid change character, then there is no difference, on what timeframe to test. The Bid sequence will anyway coincide 99% (1% for recalculation). Frankly speaking, I still don't understand why the vast majority of traders use indicators in their systems (which, moreover, lag a lot of information about the price behavior). I now apply the theory of information compression to market estimation. Both lossless and lossy compression. It is not present in this system. And in general it would be interesting to know how someone approaches market estimation, what researches are conducted and their results. This is just a drop in the ocean of market information.
 
I don't get it, are you here for a pissing match or for help?
 

Getch, Mathemat: It's good that we can write, but we can't read very well. Believe me you are not the first, several sections on this site are waiting for you, all the answers to your questions are there. Good luck, and good trends.

 
Figar0, thank you for wanting to help. Only you misidentified such help as being needed. No offence, read the first post more carefully. If there are any real suggestions, rather than recommendations to read better and harder, it would be very interesting to hear. If, on the other hand, you start seeing the rake everyone is stepping on again and criticising the systems, just without even looking closely at the report, then .... is irrelevant.
 
Mathemat:

Bars in history 774050 Ticks simulated 3824199 Simulation quality 25.00%

Yes, expectation is 30 pips, not pips. But the modelling quality is not close to 90%. What objectivity in evaluating the strategy can we talk about?

Expectation = 30.38, not pips. So, it is a pure pipsing with the price of about 3 points on the pound.