Econometrics: State-space model forecasting - page 4

 
anonymous:
The deviations of the real price from the predicted price are too small.

A very original conclusion. So if something is predicted by something with absolute accuracy with no deviations, then the prediction is not applicable. Original.
 
anonymous:

Can't find the function

na.locf(p, na.rm = F)

In which package? which class? maybe some other way?

 
Your forecasting methods are strange in general. You can't be that accurate.
 
Integer:
In general, your prediction methods are strange. Well, it cannot be so accurate.

Is 18 pips for H1 accurate? There are plenty of candles on H1 with that length.

That is why I started this thread.

 
Integer:

A very original conclusion. So if something is predicted by something with absolute accuracy without variance, then the prediction is not applicable. Original.


of course not applicable.

The spread is traded. If the current price is 1.3200 and the forecast is 1.3200 for one step, how can we trade here?

If the spread is within the trading spread - also not applicable.

The Topikaster wrote on the first page - analyze the spread.

Prediction accuracy is nothing. The variance is everything.

 
Integer:
Your methods of forecasting are strange. You can not be so precise.

Well by one bar only. I have another difficulty in assessing - visually it is difficult to assess even the potential profitability. The truth is nearby - in the tester. However, it seems he (avtar) hasn't even looked into mql, and so he will make clever R nonsense for fifty pages, until someone like you offers a free mql-realization. Looks like his professor subcortex is counting on it... ))

) Anyway, another Yusuf. )))

 
FAGOTT:



Prediction accuracy is nothing. The variance is everything.

More accurately, not the accuracy of a:

prediction error = prediction - fact

It's one step away from variance. I don't see a contradiction. I use MSE.

 
EconModel:

More precisely, not accuracy but:

prediction error = prediction - fact

It's one step away from variance. I don't see a contradiction. I use MSE.

It's not the forecast error, it's the spread you're trading. That's what you need.

TS would be a grail if the price always returns to the forecast. But in this case you will have at least 50/50 and you will lose the spread.

 
FAGOTT:


of course not applicable.

The spread is traded. If the current price is 1.3200 and the forecast is 1.3200 for one step, how can we trade here?

If the spread is within the trading spread - also not applicable.

The Topikaster wrote on the first page - analyze the spread.

Prediction accuracy is nothing. The variance is everything.


You guys are interesting econometricians ... as if you fell from the moon, talking about something in your small talk.

If the current price is 1.3200 and the forecast is 1.3200, it is clear that one shouldn't do anything.

The size of the bar is larger than the spread; therefore the forecast of the bar direction is sufficient. In the first picture almost all directions are correctly predicted, just fantastic.

 
FAGOTT:


Of course it does not apply.

The spread is traded. if the current price is 1.3200 and the forecast is 1.3200 for one step, how can we trade here?

If the spread is within the trading spread - also not applicable.

The Topikaster wrote on the first page - analyze the spread.

Prediction accuracy is nothing. Dispersion - everything.

A controversial thesis. Quite.

In limit trading, you can and should try to turn the variance to your advantage. I do not believe in the profitability of such tactics on a random series, but I am almost 100% sure it can be used for accurate forecasting.

That is why I have a counter thesis: variance is nothing. Prediction accuracy is everything. ))