Non-fitting system - main features - page 31

 
grasn >> :

I certainly can't claim to be psycologically completely healthy, I don't think any of us pass the schizophrenia test. But still, where did we write about there being some sort of grail? Find at least one post of mine with such a promise.

Who says there was a promise? Surely you must be hiding it somewhere?!!

Ж8-)

 
traxart >> :

Have you looked here ? There's a lot of them.


Distribution of pills and appointments for treatments, one is already recovering, that's how it is in our department, normal guys, adequate :)


 
Neutron >> :

Hello, Nikolai!

Oh, what a bunch of people. Tell us what, how and where.

P.S. Concerning the given example - I agree - it is not absolutely correct and is voiced, only, as "to think". By the way, how it is possible to "accelerate against the applied force"? Like, you hit a movable wall and move through it (well, so it turns out)?

Hi Sergey! By and large everything is almost the same as before, I just do not have enough time for my favourite hobby :).

About acceleration against the applied force, for some reason I have an image of an aeroplane moving in the direction of the jet of its own engines. But if we remember that air resistance will accelerate such plane, we have to admit that it will have to use the engines for braking. So outwardly the picture would look quite familiar. :)

 
I have a question for the experts. Suppose there is a system with a positive MO. It is decided to use N number of filters to improve it. For testing you can add filters in series (TC + filter1, then TC + filter1 + filter2, then TC + filter1 + f2 + f3...), but you can add each filter separately, and test only it and the system (always TC + filterN). If it improves the system, leave it, if not, scrap it. After determining several winning filters, compile them together (TC+filter1+filter2+...). Which of these approaches is more correct?
 

You have to run through the layout options, too.

It may turn out that tc+filter5 is better than any layout

 

You are going to use filters together, so they will affect the initial BP in series (what will pass through the first, will not pass through the second, etc.). So, the sum will not change due to the interposition of the slots, and the system can be tested both ways - there is no difference, but... there is a time needed for adequate testing. It (time) should be sufficient for gathering statistics for all filters, and if you add filters one by one (TC + filter 1, then TC + filter 1 + filter 2, then TC + filter 1 + filter 2 + filter 3 ...) then the average life span of a person may not be enough for the latter. Therefore, it is more correct to test the filters individually and identify the leaders and then operate them together.

Don't forget that profitability of trade (points per unit time) is directly proportional to points earned and inversely proportional to the time during which these points were earned. So increasing the number of filters (indicators, etc.) we linearly increase prediction accuracy (points gained) and exponentially increase time between transactions. The latter is decisive for profitability and reduces to zero the practicability of increasing the number of filters beyond a certain one.

In short, the optimum number of filters used is 1-2. This is a subtle issue and requires complex analysis.