Not the Grail, just a regular one - Bablokos!!! - page 258

 
Joker:

I would say the most important one. In the picture I posted earlier, I purposely left out the 3rd zone. This is where the solution to the problem lies.

The superposition principle in this case insures the TS many times. If a spread model enters the decision zone and goes down, while it has already been activated, other spread models will enter the decision zone and pick up the TS, diversify the joint position and pull up the equity where it is supposed to go.


Happy Holidays.
 
Joker:
https://www.mql5.com/ru/code/1146

There are several approaches, the main ones being:

1. Normalizing all instruments in relation to a single instrument

2. Normalisation of all instruments relative to some independent function

All methods are ultimately based on regression analysis and I have already given a link to a matlib for calculating this

I use rationing relative to an independent function...

 
Joker:

There are several approaches, the main ones being:

1. Normalizing all instruments in relation to a single instrument

2. Normalisation of all instruments relative to some independent function

All methods are ultimately based on regression analysis and I have already given a link to the matlib for calculating this

I use rationing relative to an independent function...

thanks. I will think
 
Joker:

There are several approaches, the main ones being:

1. Normalizing all instruments in relation to a single instrument

2. Normalisation of all instruments relative to some independent function

All methods are ultimately based on regression analysis and I have already given a link to the matlib for calculating this

I use rationing relative to an independent function...

It's just that I've been poking around a bit, and so far I haven't observed the effect you mentioned - that the normalized spreads keep going in channels (directed in various ways, but in channels nonetheless). Immediately after going outside the co-integration area they start to dangle as they wish.
I blame the "method" of cointegration
 
Joker:

I use normalisation relative to the independent function...

Here's the fun part. We have so far guessed at a 3 on 1, straight line and oscillator regression count. The dick has a separate optimal spread.
What other ideas do you have and who doesn't care?
 
b2v2:
Here's the fun part. We have so far guessed at a 3 on 1, straight line and oscillator regression count. Dick has a separate optimal spread.
What other ideas are out there and who doesn't feel sorry for them?
Please explain what is "on 3 on 1"? Thanks to Joker's latest update the overall picture has become much clearer (thanks a lot and a big bow to the ground from "earth people" for that). At the moment I am plagued by a mathematical question. Namely which method to normalise the channel with. In his time Joker made a reference to the brainchild of Chrenfix - indicator Trindytsykly. He recommended to study his works. It uses the "normalization relatively to the independent function", if I'm not mistaken. But is this tool suitable for our purposes? Joker's mathematics may be quite different. And hence the results are somewhat different.
 
IronBird:
It's just that I've been poking around a bit, and so far I haven't seen the effect you mentioned - that normalised spreads keep going in channels (directed in all sorts of ways, but in channels nonetheless). Immediately after leaving the co-integration area they start to dangle as they wish.
I blame the cointegration "method".
What method did you use to normalise?
 
seedormatrasch:
Can you please explain what "on 3 on 1" means? Thanks to Joker's latest posts the whole picture has become much clearer (thanks a lot and hats off to him from "earth people"). At the moment I am plagued by a mathematical question. Namely which method to normalise the channel with. In his time Joker made a reference to the brainchild of Chrenfix - indicator Trindytsykly. He recommended to study his works. It uses the "normalization relatively to the independent function", if I'm not mistaken. But is this tool suitable for our purposes? Joker's mathematics may be quite different. And hence the results are somewhat different.

Don't reinvent the wheel:

CAlgLib::LRBuild will save the fathers of Russian democracy...

( Colleagues, with your permission I will leave this thread. I gave you all the necessary information. )

 
What is 3 on 1.
Who needs to:
1. take 4 instruments and do a linear regression of 3 on 1 of them. For example gbpusd, audusd, nzdusd on eurusd. There are exactly 4 options, as you can easily guess. You can choose 4 out of 7 majors 35. Total of 140 variants.
2. Regression can be done on a straight line y=ax+b.
3. Regression can be done on a sine wave or on +1,-1,+1,-1.
4. Dick solves another problem - making a spread from instruments with minimal variance.

Regression can be done on ANY function.
 
LRBuild in alglib is just building a linear regression. What function, however, I'm no longer comfortable asking. I'm not comfortable asking which function it is. Maybe Joker counts all 140 variants. For a computer it is not much.