Where is the line between fitting and actual patterns? - page 38

 
sever30:

the tester reveals no patterns, and any optimisation results that are pleasing to your eye, are a fit.

Do you have anything to nail it with? Apart from a spotlight in the pitch black?) Our TS has a hell of a mathematical apparatus capable of finding and recognizing any patterns. And our job is just to teach it how to do it. That's what we're dealing with.
 
sever30:

the tester reveals no patterns, and any optimisation results that are pleasing to your eye, are a fit.

Question for the back of your mind -- how do you tell if a pattern is a pattern?
 
Figar0:

1. Got anything to nail it with? Other than a spotlight in the pitch black?)

2. Our TS has a hell of a mathematical apparatus capable of finding and recognising any kind of pattern. And our job is just to teach it how to do it. That's what we're dealing with.

1. I can mushroom.

2. Uh... did I miss something? "our TC..." - I'm not in the loop, where is this miracle? :)

 
Figar0:

You already replied to this post of mine on the last page highlighting the same Sample and OOS, but without the "25 again!" and screaming red) Am I having deja vu? Relax)

I just meant that the acronyms bOOS and fOOS are not the most obvious thing, my brain was simmering for a few seconds over their deciphering)

And the "dough" will be made by our TS, which we will train on Sample, select necessary results probably using OOS, and then we will give it the data, which in our opinion it will have to make dough. But we haven't got to that point yet, for the sake of purity of the experiment this data just hasn't happened yet...

У-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у-у!

How complicated things are with you!!!

And I thought you had a model of how things work in real life:

1) Sample - optimise, analyse results at the end, select some set of sets.

2) OSS - checking selected sets.

3) TEST - range of historical data, on which applying the finally selected FN set we simulate the behaviour in real life(moment of truth), checking whether we have earned the virtual dough or not.

4) And running TC-ku several such consecutive cycles, for example, (Sample-6months --> OOS-1months --> TEST-1months) * 12 cycles (with shift-1month),

and then summing up the results of the TEST --> periods shows us whether or not to continue in the real world.


.........................

I was hoping you had something like this......... ((

 
paukas:


Is that the one on the mashes?

We take, say, a 100. and a 20.

And above 100 we buy a rebound from 20.


You can limit yourself to one mashka and watch the bounce from the bead.

Maybe not relevant, I missed something about the 2 types under discussion, no time to look for it, but I have implemented a similar algorithm here. A simple strategy with its disadvantages (building up position volume)

https://www.mql5.com/ru/code/218

You can buy and sell + trailing. the whole strategy principle is shown in the diagram.

https://www.mql5.com/ru/forum/2777/page2#comment_38484

 
TheXpert:
Question for a puzzle - how can you tell if a pattern is a pattern?

Apply my motto on my avatar.

it must be as "obvious" to its owner as it is invisible to others.

 
The answer is wrong.
 
TheXpert:
The answer is wrong.
Your option is...
 
lasso:

I was hoping you had something like this......... ((


Yes, everything is exactly so, and we are talking about the same thing, only you have some specifics: consecutive cycles of training, for some reason it is 12, shifts and strictly in 1 month, etc. I do not rule out that it is all tied to your TS, and we also value these practical observations as an invaluable experience. We have not "matured" enough in this topic. We do not want to teach your TS specifically, do we?

And the test sample will of course be there, but it's not right to look back on it in the learning process. The training itself hasn't been dealt with one iota...

 

Read what you are told more carefully. Check if it exists.

The second question is, how can you do it without a history?