> I didn't realize the error should scale between 0 and 1. I've made 
>modifications in the attached version to reflect it.

To be honest, I do not know myself what the appropriate scale is because the 
mesurement noise is one of the parameters that is specific to the application. 
It depends on volatility and, probably, is proportional to the price variance 
over the estimation period. So, if a slow filter is computed over a period, 
which is 10 times longer than the period for a fast filter, I would expect the 
measurement noise parameter to be also 10 times larger.

A while back you seemed to get pretty good "fast" smoothing using the default 
measurement noise set in the JKalman at 0.1, so I think that is probably the 
type of measurement noise values we can expect for ES. 




________________________________
From: Eugene Kononov <[email protected]>
To: [email protected]
Sent: Tue, November 30, 2010 6:41:08 AM
Subject: Re: [JBookTrader] Re: Status of Kalman filter?



In the constructor in TensionKalman, it has
> super(fastNoise, fastNoise); 
>should it not be:
>super(fastNoise, slowNoise); ?
>
Yes, that was a typo.


>addParam(FAST_ERROR, 0, 20, 1, 0);

>has 0 as the last parameter. I understand the significance
>of the first four parameters for a search algorithm.
> I am not sure what the last 
>parameter, value, signifies within the optimization algorithm context. 

The first three values, "min", "max", and "step" are used by the optimizers. 
The 
last value is the actual parameter value used when the strategy is running in 
the back testing, forward testing, and trading modes.



>The other area where the problem may lie is in the optimization step.
> I would expect the measurement_noise values to be somewhere between
> 0 and 1. For fast_error it may be 0.01 and for slow_error perhaps ten times 
>that, 0.1.

I didn't realize the error should scale between 0 and 1. I've made 
modifications 
in the attached version to reflect it.



-- 
You received this message because you are subscribed to the Google Groups 
"JBookTrader" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/jbooktrader?hl=en.



      

-- 
You received this message because you are subscribed to the Google Groups 
"JBookTrader" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/jbooktrader?hl=en.

Reply via email to