Econometrics: one step ahead forecast - page 53
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Yusuf, what does Martin have to do with an advisor? It's a million light years away from an EA.
начиная с лота 0,01 наращивать его в три раза до получения положительного результата
Please carefully read the threads on martin and find out how long a losing streak can occur in real "systems" based on martin.
And I ask you to give this kind of advice there and not here.
I understand it and in hope to get such a set of unlike, "orthogonal" models.
At the moment I have linear and non-linear models available to me in regressions. Thinking the bottleneck is trend highlighting.
let's get to the bottom of your model - what properties a series should have and how this can be used to make your model make sense.
Since you are using smoothing and the forecast value lags behind recent prices, your model implies the use of series reversion only. Scientifically speaking, mean reversion. This property is paramount to the profitability of your model. It always implies:
1. The presence of some "fair" level to which prices return. It can be called the centre of price attraction. There are models with a fixed level, and there are models with a dynamic level. This level is a regression of NR. Therefore it is dynamic.
2. The presence of overbought and oversold levels relative to a fair level. They are also calculated differently. They are calculated differently, taking into account volatility, extremums, etc.
These items 1 and 2 are not taken out of the blue, they should best describe the properties of the reversion of a particular real series. If point 1 has been defined, then point 2 is not so important.
And the main thing is when this reversibility appears. Besides recoverability there are other properties - trendiness for example. So we need filtering - when the market has the used property. Consequently, when your model is effectively traded.
Yusuf, what does Martin have to do with an advisor? It's a million light years away from an EA.
Please carefully read the threads on martin and find out how long a losing streak can occur in real "systems" based on martin.
And I ask you to give this kind of advice there and not here.
Let's get to the bottom of your model
With great pleasure. Descriptive view: kotir = trend + noise + seasonality + cyclicality + outliers. Making analytical model: kotir = trend + noise. no seasonality on forex, cyclicality I do not know, outliers (news) I do not believe.
Analytics: kotir = NR(-1 to -4) + residue from NR(-1 to -2). I explain: I take 4 previous bars of HP indicator (trend) and two previous noise indicators. I optimize the number of lags. Look at the noise chart:
We see that the residual fluctuates around zero, i.e. fluctuates around HP indicator
Here's a closer look
Why not... That's right... A negative result is also a result. The main result in this case is that the person finally got rid of the obsession, which was consuming energy and time. And moved on, knowing that the previous path had been a dead end.
Let's get to the bottom of your model
With great pleasure. Descriptive view: kotir = trend + noise + seasonality + cyclicality + outliers. Making analytical model: kotir = trend + noise. no seasonality on forex, cyclicality I do not know, outliers (news) I do not believe.
Analytics: kotir = NR(-1 to -4) + residue from NR(-1 to -2). I explain: I take 4 previous bars of HP indicator (trend) and two previous noise indicators. I optimize the number of lags. Look at the noise chart:
We see that the residual fluctuates around zero, i.e. fluctuates around HP indicator
you have already cited this many times, but it is only part of the prediction. In a previous post I already wrote about the rest
2. The presence of overbought and oversold levels relative to a fair level. These too are calculated in different ways. Taking volatility, extremums, etc. into account.
These items 1 and 2 are not taken out of the blue, they should best describe the properties of the reversion of a particular real series. If point 1 has been solved, then point 2 is not so important.
And the main thing is when this reversibility appears. Besides recoverability there are other properties - trendiness for example. So we need filtering - when the market has the used property. Consequently, when it is efficient to trade your model
I have become convinced that in regressions this or any other period, even optimised ones, will eventually punish the trader. It's impossible to guess future market periods, that's the problem. That's why now I think to look on the market as on the game with random outcome, but, apparently, to have a small advantage one should enter and exit by recommendations of TS and not to hope that necessarily it will lead to profit. I think one cannot do without the option of sparing martingale to cheat the market and gain profit, for example, starting from 0,01 lot increase it three times till getting positive result and then come back to 0,01. Is there such a variant of the EA?