Some signs of the right TCs - page 25

New comment
 
...:
That is, the invariance condition restricts the use of different mani-management options.

It does not.

 
Алексей Тарабанов:

Vladimir, the move between what is the current trend?

Why "current"? It has already happened, it has passed. It has a beginning and an end - the points of extremums or reversals.

On the subject (not on this issue) I would like to say that when we develop difference schemes for solving boundary and initial value problems for partial differential equations we should also check the conservatism property such as whether the scheme maintains the balance sheet properties that are consistent with the real process described by the equation. In difference schemes, the law of conservation of internal energy (heat) for the heat conduction equation, the law of conservation of mass of each of the transported components for the diffusion equations, and the laws of conservation of mass and momentum for the hydrogasodynamics equations.

https://mash-xxl.info/info/22191/:

" A difference scheme constructed in such a way that the conservation law is satisfied for each unit cell and is not violated as a result of summation over all cells, i.e. is also satisfied for the whole study area, is called conservative. Such schemes can be successfully applied to equations with non-smooth and discontinuous coefficients, when arbitrary meshes are chosen, etc. The use of conservative schemes, as a rule, leads to an increase in the accuracy of the solution. [c.63]"

When creating trading algorithms, why not similarly check compliance with some laws, e.g. conservation of moments of extremum occurrence?

 
fxsaber:

I am completely unclear as to what you mean by this. Take ANY symbol, run my EA on it. Then flip the symbol and run it again. Compare.

It takes one minute to check it: compile the EA and run two passes.

And the profit here will not be in pips, but in some incomprehensible units.

Again, how do you set the spread size?
 
and in general, what are we eliminating here?

for example, testing over 20 years.

and 20 years ago a financial instrument was worth 1.0000, and now it's worth 2.0000.

and our Expert Advisor has the size of the input parameter as 1 point.

But it is clear that at a price of 2.0000, the size of this parameter should already be 2 points.


Of course, we could take the size of the input parameter as a percentage of the current price, but this is also an option.
 
Vladimir:

Why "current"? Already established, having taken place, passed through. It has a beginning, it has an end - these are points of extremums, or reversals.


Why do I need a trend that has already passed, if I am only going to buy/sell something? I want to understand where price is going now, not where it was going before the previous extremum.

 

igrok333:
а прибыль же здесь будет не в пунктах, а в каких-то непонятных единицах.

Here are the details.

again, how do you set the spread size?

You don't need one.

 
Алексей Тарабанов:

Why do I need a trend that has already passed if I'm just about to buy/sell something? I would like to understand where price is going now, not where it was going before the previous extreme.

The same reason why testing in MT is used, why they accumulate archives of quotes, why they make other programs to imitate trading in the already passed trends, whythey create indicators and even why they just show charts of past price history. Don't you ever look at past courses, only forward? There is nothing to see, there are no charts or OHLC tables yet.

 
fxsaber:

I have never practiced it. Moreover, I consider such TS to be mathematically flawed, as it is impossible to do a TSR study on them through the Optimiser.

For example, OnlyBuy-TS cannot reveal the potential of the 1/S&P500 symbol through the Optimizer.


There are several stages of formation of a battle Advisor, I will highlight several indicative ones

  1. Research TS. It is the one that should be mathematically correct.
  2. On the basis of point 1, patterns are found.
  3. Perhaps some of them pass the filter of their requirements.
  4. A combinational TS is built based on these patterns, where various fitting variants are possible, including separate settings for Buy and Sell.
I'm talking about the research TS. Combat stumps are 5% of algotrader's work.

1. Well OnliSell will reveal it then, won't it?

2. it's a good description, the 1st point is interesting. In the classics it's like this:

- Hypothesis

- Experiment testing the hypothesis, screening hypotheses and building a working model

- Model fitting.

What I'm saying is that in general the hypothesis stage - they should be different for buy and sell for efficiency reasons of finding patterns (which in general case are different up and down)

But your right of course to use one and the same. As I understand you are aiming at generalising the strategy and scratching out every pip, then maybe it makes sense, on pips movements the patterns lie on a different plane than on larger trend movements obviously.

 
Aleksey Mavrin:

1. OnliSell will reveal it then, won't it?

It's on the same level as OnlyBuy on a straight symbol.

I really like that I could be wrong. I'll have to try optimising for each direction separately and see what happens. Now that's unsubstantiated assertions on my part.

 

I tried my robot on synthetic instruments and the results were interesting. The settings are the same everywhere

1) GBPUSD m1

2) 1/GBPUSD

3) 2*GBPUSD

4) 2+GBPUSD

5) GBPUSD^2

It turns out that addition of the coefficient 2 has not affected the drawdown and yield at all. Multiplication by the coefficient 2 has increased the drawdown and yield. With the reverse instrument the strangest thing, the yield fell and the drawdown increased by 2 times. But I understood the problem there, my algorithm stumbled on rounding errors, it should be corrected in the next version.

1...181920212223242526272829303132...37
New comment