Econometrics: State-space model forecasting

 

Made a model of the state space and by not making a prediction one step ahead.

Laying out the chart.

The black one is eurusd and the red one is the one step forward forecast. You can see from the chart that there is a one-step forecast. It was obtained on the previous 48 bars H1. The previous forecast - it was obtained on the previous bars. I.e. the red line is out of sample.

Problem. How to get to the trading system. It seems to make good predictions. I do not understand the market entry and exit conditions.

I am ready to implement the ideas and post the result.

 
EconModel:

The problem. How to switch to a trading system. Seems to be making good predictions. I don't understand the entry and exit conditions.

Bummer!

If it "predicts well", then enter the market according to the prediction.

 

do a spread analysis - divergence of the two curves. What is the average value for a 4-digit spread?

If within the trading spread, then no.

 
PapaYozh:

Bummer!

If "predicts well", then enter the runoff according to the prediction.

I don't really get it.
From the picture: is the red one below the black one a short?
 
Demi:

do a spread analysis - divergence of the two curves. What is the average value for a 4-digit spread?

If within the trading spread, no.

Here is the end of the chart. Is the red below the black - short? But just to the left, the red is also lower and the price is up.....
 
EconModel:
I don't really get it.
From the picture: is red below black a short?

in theory, yes.

in practice it would make you lose at the speed of the spread.

)))) if you were trading both red and black, you would be trading at the "collapse" of the spread, and this way you would be losing



 

Enter in the direction of the forecast. If the predicted value is higher than the current value, buy, if lower, sell. And then, on the next bars to hold or roll.

But I do not believe in the possibility of such an accurate prediction, incredible picture, as if there was not a mistake with looking into the future.

 
Integer:

Enter in the direction of the forecast. If the predicted value is higher than the current value, buy, if lower, sell. And then, on the next bars to hold or roll.

But I do not believe in the possibility of such an accurate prediction, it is an unbelievable picture, as if there was not a mistake with looking into the future.

Long if the forecast is higher than the previous one or higher than the fact, i.e. also one step back.
 
Integer:

...however much the mistake of looking into the future.

The algorithm is as follows: a sample is taken (now 48 bars), the forecast for one step is counted on it, it can be counted for several steps. A new bar and it is considered again, i.e. the forecast is always considered on previous bars. This is how the red line is obtained.
 
EconModel:
Long if the forecast is higher than the previous forecast or higher than the fact, i.e. also a step backwards.


Two options you can try: forecast higher than the actual value and forecast higher than the previous forecast.
 
Integer:

Two options to try: prediction above the real value and prediction above the previous prediction.

Should the truth be in the tester?

Under R I could try something, but the EA would have to take time out.