[Archive!] FOREX - Trends, Forecasts and Consequences (Episode 6: August 2011) - page 174

New comment
 
wmlab:


I have already got it up to 1.44 =) I have not got any profit. I should have made 10-20 predictions and gathered statistics. If at least 7-8 out of 10 will come true - I should write my Expert Advisor according to the indicator =)

I can send you a message. I've already written it. Maybe not the best way to implement the idea, but it works and you can run it in the tester.

I lost on all timeframes, but the auto may improve the Expert Advisor as well.

 
wmlab:

That's why I was asked to fix the sample range. For example, put on H1 for the last three days, made a forecast for the day and already the pattern is not moving anywhere. It is impossible to play on a moving sample.
Can you be more specific? What is a moving range? I.e. in the example above, how to decide which of the two trends should be considered exemplary? The idea is that the sliding prediction should not change radically - if the system contradicts itself and requires us to choose from the buy-sell options (which are always there anyway), then what is it for?
 
gold up 100 units in 2 days....doll. should be getting cheaper... market won't settle down until 1900...
 
marketeer:
Can you be more specific? What is an exemplary range? I.e. in the above example, how do we decide which of the two trends is exemplary? The idea is that the sliding prediction should not change radically - if the system contradicts itself and requires us to choose from the buy-sell options (which are always there anyway), then what is it for?

Apparently the exemplary range depends on the pair. For EURUSD I have chosen a trading day boundary. I choose two or three days on H1 - and I can see the dynamics for today. In addition, I'm going to improve the indicator so that it can choose its own parameters according to certain considerations. For example, stability of forecast... These are ideas for now.
 
wmlab:

Apparently, the exemplary range depends on the pair. For EURUSD it is possible for me to choose a trading day boundary. I choose two or three days on H1 - and I can see the dynamics for today. In addition, I'm going to improve the indicator so that it can choose its own parameters according to certain considerations. For example, stability of forecast... These are ideas at the moment.
As for stability, I've been averaging my forecasts on several consecutive bars.
 
wmlab: For example, the stability of a forecast... are still ideas.

A normal, quite reasonable idea, by the way.

Marketeer: About stability - I've been doing forecast averaging on several consecutive bars.

It seems to me that stability should not be imposed by ironers. It should be a selection criterion.

 
lotos7:
gold up 100 units in 2 days....dol should be getting cheaper... the market won't settle down until 1900...
Everything is going down - the quid is going up - gold is going down...
 
Mathemat:

A normal, quite sensible idea, by the way.

It seems to me that sustainability should not be imposed by ironers. It should be a selection criterion.

How about using confidence intervals? For me the probability on the graph is too high. I think the 3 most probable scenarios should be plotted in different colours and their probability and quality of forecast fulfillment should be saved in the log. Then it is possible to use the tester and forward tests.
 
Mathemat:

It seems to me that sustainability should not be imposed by ironers. It should be a selection criterion.

And what is wrong with such a criterion? If adjacent predictions overlap weakly enough, then any one of them seems unlikely. From the algorithmic point of view it seems simple enough to sum the proposed trajectories, and if they do not agree with each other - we get a horizontally directed cloud(the confidence interval will be wider than the length of the "start-target" movement). If there are other ways of determining stability, please tell me.
 

The ruble has jumped a lot... Is it just me ?

1...167168169170171172173174175176177178179180181...202
New comment
Reason: