Discussion of article "Resolving entries into indicators" - page 3

New comment
 
Andrey Khatimlianskii:

I don't get it.

If we are talking about selecting indicators for a strategy, what does the quantity have to do with it? Does it mean the percentage of hits?

From mathematics, a percentage is one hundredth of something. Speaking about the number of trades, I meant that when reverse-engineering other people's systems, it is not necessary to sum up the profit on trades, but just count their number. Where there is a maximum number of trades (ideally, a full hit), those signals are close to the strategy used. Of course, the maximum number of trades will correspond to the maximum percentage of hits. Simply, if it is more convenient for you to analyse percentages, you will have to divide the received number of deals on signals by the total number of deals and multiply it by 100 when building charts.
 
Dmitriy Gizlyk:

When analysing, not a specific deal is shown, but the sum of profit on deals at comparable moments of indicators. It is obvious that not all deals in the specific sum of values are profitable. But on the analysed time period the coincidence of EA signals and filter readings gives profit.

You yourself write that only the coincidence of trades and filter readings are analysed.
And if you remove trades and leave only the filter, there will be other inputs. They will not be profitable.

I was talking about the situation when trades of a strategy can overlap.
For example, there was a buy signal at 12:00 and the position was closed only at 20:00. The next one opened at 23:00.
But at 14:00 and 16:00 there could be 2 more buy signals (which were not executed because the position was already open). So, if the filter cancels the entry at 12:00, but does not cancel one of the next entries (14:00 or 16:00), there will be another trade, which was not analysed for profitability and combination with the filter at all.

That is why the results will be different from the variant with the filter embedding and subsequent optimisation.

 
Dmitriy Gizlyk:
In maths, a percentage is a hundredth of something. Speaking about the number of deals, I meant that when reverse-engineering other people's systems, it is not necessary to summarise the profit on deals, but just count their number. Where there is a maximum number of trades (ideally, a full hit), those signals are close to the strategy used. Of course, the maximum number of trades will correspond to the maximum percentage of hits. Simply, if it is more convenient for you to analyse the percentages, you will need to divide the received number of trades by signals by the total number of trades and multiply it by 100 when building charts.

Ahem. Quantity is not enough. It is necessary that the trades coincide with the original.

This is the percentage of hit (overlapping of simulated trades with real ones) I meant.

 
Andrey Khatimlianskii:

Ahem. Quantity isn't enough. You need the trades to match the original.

That is the percentage of hit (overlapping of simulated trades with real ones) I meant.

This will be the next step, after defining a set of indicators.
 
Andrey Khatimlianskii:

You yourself write that only coincidences of trades and filter readings are analysed.
And if you remove trades and leave only the filter, there will be other inputs. They will not be profitable.

I was talking about the situation when trades of a strategy can overlap.
For example, there was a buy signal at 12:00 and the position was closed only at 20:00. The next one opened at 23:00.
But at 14:00 and 16:00 there could be 2 more buy signals (which were not executed because the position was already open). So, if the filter cancels the entry at 12:00, but does not cancel one of the next entries (14:00 or 16:00), there will be another trade, which was not analysed for profitability and combination with the filter at all.

That's why the results will be different from the variant with filter embedding and subsequent optimisation.

Yes, I have analysed the matching of trades with the filter readings. And if the filter cancels one trade, but misses another one later, it is very likely that the new trade will bring profit. This follows from the statistical analysis and is confirmed by the post-testing conducted at the end of the article.
 
Dmitriy Gizlyk:
Yes, I analysed the coincidence of trades with the filter readings. And if the filter cancels one deal, but misses another one later, then with a high probability the new deal will bring profit. This follows from the statistical analysis and is confirmed by the post-testing conducted at the end of the article.

You still don't understand me, but that's fine.

 

I think even a good strategy doesn't always work, you have to take the situation into account as well.

 
Ivashka222:

I think even a good strategy doesn't always work, you have to take the situation into account as well.

If you are trading manually, then it is undoubtedly necessary to take into account the situation. If we are talking about trading with an Expert Advisor, it is impossible to take into account all possible situations when writing an Expert Advisor. That is why we build an algorithm that is as independent of the situation as possible. Of course, you have to pay for universality with losing trades, but the strategy should give profit on a long time interval, covering the past losses with current profit.
 

I'm trying to run it but I can't.

It gives an error in StaticMACD.mqh

2017.12.29 08:11:23.672 2017.12.14 18:16:59 array out of range in 'StaticMACD.mqh' (375,45)

123
New comment