Recognising images ( rhetorical theme ) - page 13
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
No. There are no filters. Recognition is done directly from the noisy stream. Where did you read about filters? The best part is the hearing mechanism, read about it. There the recognition begins immediately, first at a low "hardware" level, the sound is encoded in a certain way and then converted into this signal-code is recognized at a higher level. The analogy is incomplete but captures the essence. The principle of separation of useful information is not filtration (chunking) of the stream, but recognition in the stream, PIC recognition loops responding to the most appropriate images, that is, the selection of the most appropriate images from the stream.
Don't "poke" at people you don't know. That's one.
Filters - not filters, it's up to you. I described my approach to the problem. That's two.
No. There are no filters. Recognition is done directly from a noisy stream. Where did you read about filters? It is the mechanism of hearing that is best explained, read about it. There the recognition starts immediately, first at a low "hardware" level, the sound is encoded in a certain way and then converted into this signal-code is recognized at a higher level. The analogy is incomplete but captures the essence. The principle of separation of useful information is not filtration (chunking) of the stream, but recognition in the stream, PIC recognition loops responding to the most appropriate images, that is, the selection of the most appropriate images from the stream.
By the way, in the case of pattern recognition, the convolution operation is actually a filter. And the extraction of individual features can be called filtering, but that is a very narrow definition.
K= (360, 0.128, 0.746); - Here Open < Close as 0.128<0.746
K= (360, 0.746, 0.128); - here Open > Close because 0.746>0.128
it is quite enough to describe the size and shape of a bar (or candlestick).
Sergiy, a question. And from what considerations do you use LO/HL and LC/HL ratios? And not say K= (HL, HO/HL, HC/HL); or say K= (HL, LO/HL, LC/HL, HO/HL, HC/HL);
where HO and HC, are calculated similarly to LO and LC, but only relative to High.
Now that the topic is up, I'm going to post the final results of the pattern suggested by Denis(denis_orlov). However long trades use stops not calculated according to the Fibo pattern, because in this case the buying performance worsens. I am not sending the Expert Advisor, everybody can do it themselves. The indicator used for calculations https://www.mql5.com/ru/code/9767
Strategy Tester: Entry on wave C on trend_01Decent results when you consider that shorts were optimised on 09-10 and longs on 08.
An initial deposit of 50,000,000 is the way to go!...)
Net profit 634344.12
Maximum drawdown 83309.14
634344/83309=7.6
The net profit exceeded the maximum drawdown by 7 times. This is certainly not enough for 10 years. The report was made to show that it is possible to squeeze something out of this pattern.
Net profit 634344.12
Maximum drawdown 83309.14
634344/83309=7.6
The net profit exceeded the maximum drawdown by 7 times. This is certainly not enough for 10 years. The report is just posted to show that one can extract something from this pattern.
What figures are obtained at a fixed lot? The working out of the pattern should be estimated at a fixed lot. Otherwise you are estimating the system in general, mainly its fitting part. In general, the tester assesses the quality of the fitting. And you need to pay attention to the entry system, not to the martin as a whole and not to the fit.
In practical terms the results are not very good. The profit is obtained due to bounces, there are long periods of non-earning, or even a loss. You miss a jump and that's it, you get nothing.
What are the numbers for a constant lot? The performance of the pattern should be evaluated with a constant lot. Otherwise you are evaluating the system as a whole, mainly its fitting part. In general, the tester assesses the quality of the fitting. You should pay attention to the entry system, not to the martin in general and not to the fit.
With constant lot the balance graph is the same. MO, at lot 0.1 >36.
I do not use Martin at all.
My profit is obtained due to bounces, there are long periods of no profit or even a loss.
I have long periods of losing profits or even losing them. Show me an Expert Advisor without drawdowns.
You miss a spike and that's it.
:) No comment