![MQL5 - Language of trade strategies built-in the MetaTrader 5 client terminal](https://c.mql5.com/i/registerlandings/logo-2.png)
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
clearly that's not what I meant. Do you calculate the prediction error as a RMS?
There is no forecast itself and you all are counting its error, isn't it strange? Let us first make a prediction, even if it is wrong to some extent, or rather an equation or function to make a better prediction, and then dance from there.
The author has already written many times how he gets the predicted value and the error.
it cannot be a constant in every single trade. And the slope of the price from the forecast (error) may converge to a certain constant, if the error distribution is stationary
I can present the exel version of the indicator for you to check according to your methodology. I have exhibited the indicator several times, you can extract the information you are interested in from the code, including about the Gamma function.
There is no forecast itself and you all are counting its error, isn't it strange? Let us first make a prediction, even if it is wrong to some extent, or rather an equation or function to make a better prediction, and then dance from there.
A constant (almost a constant) is the goal of the model construction, if it fails, then the model does not exist in this quotient area.
Here for example we decide to do the bullshit)))) and predict a random walk. We start at zero, increments +1/-1 with 0.5/0.5 probability. Best prediction for any number of steps is current position. So if we have zero, we have 100 or 1000 steps, then our best prediction is zero. But how does the error of this prediction change? The RMS error increases in direct proportion to the root of the number of steps. Error (RMS) for 100 steps = 50, but for 400 steps it is 100. This is in the case if there is no reversion or trending. If there is reversibility, the error will grow slower than the root of the number of steps. If trending, it is the opposite
For example, we've decided to do the bullshit))) and predict a random walk. We start at zero, increments +1/-1 with probabilities 0.5/0.5. Best prediction for any number of steps is current position. So if we have zero, we have 100 or 1000 steps, then our best prediction is zero. But how does the error of this prediction change? The RMS error increases in direct proportion to the root of the number of steps. Error (RMS) for 100 steps = 50, but for 400 steps it is 100. This is in the case if there is no reversion or trending. If there is reversibility, the error will grow slower than the root of the number of steps. If trendiness, it is vice versa.
Not interested in the random walk error (no demolition) - it is not predictable and is a sign of stopping the model construction process. It is the end of the story by definition.
I'm explaining the point to you with an example and you're saying "prediction is not possible". The prediction of anything is possible. Another thing is whether this prediction makes practical sense. In the case of Sb from the point of view of minimizing the skop (error) the best prediction will be the current value of the series. Of course, this does not mean that you can make money on a forecast.
I'm giving you an example and you're saying "prediction is not possible". Anything is possible. Another thing is whether this prediction makes practical sense. In the case of Sb, from the point of view of minimizing the cost (error), the best prediction will be the current value of the series. Of course, it doesn't mean that you can make money on the cb.
I'm talking about the methodology (not invented by me) of model building: the initial non-stationary quotient must be decomposed into components until you get a stationary residual. This requirement is well understood on an intuitive level (which is very important), because the stationary residual can be replaced by a constant equal to any of the values: mean, skew, variance, spread - anything and everything is possible in the case of stationarity.
Did I argue with that?