Norm? - page 26

New comment
 
Wex:
I can help you.
It was a joke) It is not possible to restore an arbitrary system on trades, unless it is very simple, or tied to a clock/calendar reading.
 
bas:
It was a joke) You can't reconstruct an arbitrary system from transactions, unless it's very simple, or tied to a clock/calendar reading.

A nail can't have north and south. And there is no way a whole orchestra can fit on a thin brown film. It is utterly impossible.

Also, you can't have sunspots in the sun.

I totally agree with you: when someone says they can pay, it's probably a joke.

And the solution to any complex technical issue depends only on funding. (Not so long ago they built a gizmo, and it wasn't very expensive, so there's still money left over to make it generate black holes.)

 

Wex:

I totally agree with you: when someone says they can pay, it's probably a joke. And the solution to any complex technical issue depends solely on funding.

OK, give me any chart of your choice and I'll put your transaction history there. If you can reconstruct the system from them, you will easily get rich without my payment ))))
 
Wex:

Also, the sun can't have specks in it.

Then what prevents you from marking trades on a history with the ability to see both ways, and then reconstructing the logic from that marking? One would think that this is none other than the Grail. But neural network algorithms can do it, with varying success, but the results do not fit the Grail. The market constantly reflects on its own inefficiencies, with a wide range of algorithms, from the simplest to the completely unreadable for humans, arising from training in ANN.

Therefore any algorithm, no matter how it originated, in the human brain or in ANN training, will quickly cease to be effective, because there is a very small chance that no one in the world recognizes this kind of inefficiency and starts compensating for it along with you by playing against it. Whereas in the last century the Grail could live for many years, now that time has been reduced by an order of magnitude and will apparently decrease.

 
bas:
OK, give me any chart of your choice, I will mark there the history of deals. If you can reconstruct the system on them, you will easily get rich without my fee)))

Some developer was offering his free EA on the internet, but in EX4 format and with protection. Everyone was indignant as to why they could not get the source code. And the idea was visible - after running it in the tester - to the naked eye. As for someone else's deals and "stats" - they are probably of little interest to anyone. There is no point in wasting time on it. I meant that I can help you with development of your own TS. Nothing more.


EvMir:

And then what prevents you from marking deals on a history with the ability to see both ways, and then, based on the data of this markup, restore the logic? One would think that this is the only grail. But neural network algorithms can do it, with varying success, but the results do not fit the Grail. The market constantly reflects on its own inefficiencies, with a wide range of algorithms, from the simplest to the completely unreadable for humans, arising from training in ANN.

Therefore any algorithm, no matter how it originated, in the human brain or in ANN training, will quickly cease to be effective, because there is a very small chance that no one in the world recognizes this kind of inefficiency and starts compensating for it along with you by playing against it. Whereas in the last century the Grail could live for many years, that term has now been reduced by an order of magnitude and appears to be diminishing.

You are absolutely right. However, an average robot makes a decision analyzing present prices, not absent ones. Therefore such a "markup" would be illogical.

 
lordlev:
published in the marketplace. TimeMachine
Can you tell me why in the screenshots every two years start at 10,000 deposit? Was it 7 different tests (fits / forwards) or something else? If you test on the whole period of 14 years (I tested on all ticks with a lag for interest), the FS figure is in the region of 50. Thanks in advance for the reply.
 
alexeymosc:
Can you tell me why in the screenshots every two years start at 10,000 deposit? Was it 7 different tests (fits / forwards) or something else? If you test on the whole period of 14 years (I tested on all ticks with a lag for interest), the FS figure comes out to be around 50. Thank you in advance for your reply.
If you have tested over the whole period you have seen what a humpy graph you have come out with. I didn't think it would be presentable. That's why I divided it by 2 years and made these screenshots. I was testing on OHLCm1. There is no point in using ticks. They are synthetic anyway. Arbitrary delay makes no sense either, unless of course the potential customer wants to place the robot in the "kitchen". A person who decides to buy my system will most likely use FIXAPI. And there can't be any delay in execution. Using arbitrary delay in the tester will always give different test results.
 
lordlev:
If you've been testing all along, you've seen how humpy your graph came out.

I see. But, by the way, it didn't really come out hunched over. The FV is over 50.

Actually, congratulations, I guess. Good stuff came out (in the tester).

 
alexeymosc:

I see. But, by the way, it didn't really come out hunched over. The FV is over 50.

Actually, congratulations, I guess. Good stuff came out (in the tester).

well, yeah. we'll just have to wait until the real stats for the year come in and draw conclusions
 
lordlev:
I tested by OHLCm1. There's no point in testing on ticks. They are synthetic anyway.
If your trades are opened not on opening prices, but inside bars (minute bars), then both OHLC and synthetic ticks will give distorted information. This applies to entries. So you have to choose. You just chose the prettier picture, and there is an option with a worse one, but the test mode is the most accurate. I was testing on ticks on the minutes. Take and stop triggering will be modelled acceptably in either mode. On the whole, I repeat, the result is fascinating, on any test mode it plows well.
1...192021222324252627282930313233
New comment