What 4 factors do you think the price depends on? - page 9

New comment
 

Screenshots of the MetaTrader trading platform

EURUSD, M30, 2016.05.25

Alpari Limited, MetaTrader 4, Demo

Resistance line, further downward trend

EURUSD, M30, 2016.05.25, Alpari Limited, MetaTrader 4, Demo


 
Maxim Romanov:
That's the thing, there are no stable patterns, the lifespan of any found pattern is finite, or it will change, start to manifest itself differently. And the hardest thing is to understand when the pattern has disappeared from the market. Within the framework of this subject, the ratios will change, we can take all possible correlations and the ratios will change.
I agree, that's the point.
 

And there are many patterns to be found, so let's think about how to know if a pattern has changed/disappeared. I have only come up with one way so far. Keep several trading strategies at the same time and set each one to a drawdown level, after which this trading system is no longer used. A new strategy is put in place of the old one.

Maybe I can come up with a more effective way, particularly with the ratios that Yusuf is developing.

 

Developed a new system called "The nagging Husband and the cunning Wife".

 
Maxim Romanov:

And there are many patterns to be found, so let's think about how to know if a pattern has changed/disappeared. I have only come up with one way so far. Keep several trading strategies at the same time and set each one to a drawdown level, after which this trading system is no longer used. A new strategy is put in place of the old one.

Maybe it is possible to come up with a more effective way, in particular with ratios that Yusuf is developing.

First, answer the question "was there a pattern"?

Is it when the model selected variables on history and established a relationship between them, and the forward showed an immediate exit FOR the maximum drawdown of the model.

 
Дмитрий:

First, answer the question "was there a pattern"?

This is when the model selected variables on the history and established a relationship between them, and the forward showed an immediate exit FOR the maximum drawdown of the model.

Right, and how do you know if it was a pattern or a coincidence?

Usually I develop a theory first, then I make a trading strategy based on this theory and check it on historical data, but it's so slow, I need to speed up the search for patterns and automate it.

 
Maxim Romanov:

Yeah, but how do you know if it was a pattern or a coincidence?

I usually develop a theory first, then make a trading strategy based on that theory and test it on historical data, but it is so slow, I need to speed up the search for patterns and automate them.

I wrote - forward. If the divergence of the model price on the forward is within the divergence of the price and the model on the training sample - the model is adequate. If the divergence on the forward immediately "goes" beyond the divergence on the training sample - the model is inadequate.

For example, if the maximal deviation of the price from the model on the training sample was maximum +/- 100 points, and on the forward one it immediately "flew" beyond 200 points, then the model has no predictive value.

 
Дмитрий:

I wrote - forward. If the divergence of the model price on the forward is within the divergence of the price and the model on the training sample, the model is adequate. If the divergence on the forward immediately "goes" beyond the divergence on the training sample - the model is inadequate.

For example, if the maximal deviation of the price from the model on the training sample was maximum +/- 100 points, and on the forward one it immediately "flew" beyond 200 points, then the model has no predictive value.

Again it is inaccurate. I have a model that predicts values of a future bar (O, H, L, C) and sometimes it is very accurate, up to 1 point accuracy on a forward position. But at other times it may not be accurate at all and it is not very useful because I do not know in advance if the forecast will be accurate or not.
 
Maxim Romanov:
Again it turns out to be inaccurate. I have a model that predicts future bar values (O, H, L, C) and sometimes it is very accurate, up to 1 point on a forward position. But at other times it may be not even close and it is not very useful because I do not know in advance if the forecast will be accurate or not.
I remembered a kind of question: Which is better, a clock that always shows inaccurate time, or a clock that shows the exact time 2 times a day?
 
Maxim Romanov:

And there are many patterns to be found, so let's think about how to know if a pattern has changed/disappeared. I have only come up with one way so far. Keep several trading strategies at the same time and set each one to a drawdown level, after which this trading system is no longer used. A new strategy is put in place of the old one.

Maybe I can come up with a more effective way, particularly with the ratios that Yusuf is developing.

Now I have set myself a private goal to identify the patterns or behavior of the ratios in unexpected reversals, to try to separate a correction from a reversal. But, so far we haven't decided on the main 4 factors. Let's take, if there are no other suggestions, as before, F1 - $, F2 - Au, F3 - F (frank), F4 - f (funt), Y - E (eouro) and the variant: F1 - O, F2 - H, F3 - L, F4 - V, Y - C (eouro). Let's find out whether there are significant or tangible and more or less constant signs of a reversal and/or correction in the behaviour of the 5 factors of the 4-factor linear regression equation or not.
123456789101112
New comment