Signal clarity

[dloomis]

Re: Reply to post 'Signal clarity'

Still using the same cross over signals? Didn't see them on the parameter
string in the study window

----- Original Message -----
From: <[email protected]>
To: <[email protected]>
Sent: Wednesday, December 17, 2003 5:22 PM
Subject: Reply to post 'Signal clarity'


> Hello dloomis,
>
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>

[dloomis]

Also looks like you only ran 1 contract in the back test, usually I have been running 50 so the dollars show up instead of points.

[ACM]

David
The parameters used for the backtest are the ones mentioned in the message with the attached efs and, yes, I used 1 contract only so as to show points.
Although the strategy seems to provide slightly better results and the equity curve is more acceptable the system still needs a lot of work IMHO.
Alex

[FibbGann]

Here is the bisggest efficiency problems that I have found.

The indicators really need to have a smoothing factor to them.
(A MA of each used instead?) due to the whipiness of the indicators.

Here is an example of another problem: Lets take CCI in theory...

Say the CCI is at -120 and crosses up through the -100 to give a buy signal. Then the CCI goes to +90 then rolls over and comes back down. The problem now here is, The CCI did'nt reach the upper boundry to give us a sell signal on the way back down. This can be based on all indicators in reference to there triggers. So we either need to be able to Identify the change in the angle (degrees) of the indicator or have some type of median point of all indicator to indicate a secondary buy/sell indicator.

I would really like all your thoughts on this. Which seems to be the biggest default to this study.


Fibbgann

[dloomis]

Re: Reply to post 'Signal clarity'

Well, let's take these points one at a time...

The indicators really need to have a smoothing factor to them.
In your example, the cci went from -120 thru -100 (giving a signal)
then proceeded to +90 then went back down.
I don't see how adding a MA would help that situation, any ma would
probably lower the +90 reading and would still not give a reversal signal.
I am not sure why you think this is a bad thing.

Identify the change in the angle (degrees) of the indicator ...
We can do that easily enough, but it introduces another complexity - not
only do we have to have the 'right' parameters for cci, now we need to
'choose' the right slope. I am not saying this is wrong, just trying to
keep some simplicity here.

have some type of median point of all indicator ...
This is the same as saying the signal points should be different. Say
for example we wanted to use a reading of 0 on the cci for our signal
(instead of +/-100 currently). Just change the parameters already available
in the Edit Studies window to reflect the level you want the signal at.
Question - Are you getting away from the 'textbook' usage of cci if you so
this? Is that a good thing?

That's my 2 cents, fwiw.

PS - Have you gotten a system with more points than Alex has - about 280 or
so? Isn't that one of the prime determinants of a good system - the bottom
line?

[gspiker]

What might be interesting to try, along with the MA if you want to try that, is to put a filter on it. (This is without really looking at the results too closely)

1) Something like you will only take the signal if in next two bars after the signal that you get a bar in the direction of the signal

2) Some sort of offset MA that prices have to rise above (for a long) or close below (for a short).

Often these types of rules can prevent whiplash and not delay the signal too much.

G

[gspiker]

[

Isn't that one of the prime determinants of a good system - the bottom

Yes, but that includes more than pure profit. Unless you got unlimited funds and good intestinal fortitude draw down and biggest loosing streak is also important.

G

[ACM]

All
In the last strategy I posted the system has changed somewhat in the sense that it no longer is an OB/OS system but rather a contrarian one in as much as it tends to sell as the indicators are crossing upwards the centerline of their excursion and to buy when they are crossing down.
I would agree that applying MAs to the indicators may (note the conditional) be beneficial in as much as it may accomplish two things.
First it should smooth out the signals thus eliminating some of the excess trades.
Secondly it may introduce the offset/delay that Garth refers to. If it does not then we can anyhow introduce that offset ourselves.
Given that I have most of the studies with MAs already done I will put together a version using those. David can then add the rest of the logic as it is now.
In the mean time if anyone wants to continue the testing to see if other values produce better results you are welcome to it.
Alex

[FibbGann]

Using a MA really does'nt have a draw down that I can see at all!
You can visually test if you like (I did) it looks to be alot better.
As to the CCI reference, these indicators don't always meet there targeted strategies, which is adding to the drawdown. This is the reason I mentioned possibly having a secondary signal notifying us that the indicators did not reach the potential targets that we had liked. Alot of times they do, But, if visually backtesting you can see instances where indicators did not go above then cross below or sell signal. This givng us unwanted draw down.

I have been following the CCI and STO MA's that Alex created, modifying the periods of the MA. Looks to be alot smoother. It is not my intent to keep modifying this strategy all the time. But, I beleive we all want the most efficient indicator as possible.


Fibbgann

[FibbGann]

I have continued to run backtest in every possible scenario I can come up with. This is what lead me to these other ideas.


Fibbgann

[ACM]

FibbGann

you can see instances where indicators did not go above then cross below or sell signal

That is the reason why I modified the strategy and moved it away from an OB/OS system. And in fact the drawdowns have come down somewhat and the equity line has smoothed considerably.
Anyhow, as I said before, I think it should not be too difficult to create the combined strategy using MAs of the same indicators. Most of the coding is already done.

Alex

[dloomis]

Re: Reply to post 'Signal clarity'

FYI

No trades today in the 3signal system - still short from 1072.75 yesterday

[FibbGann]

Still short from 1072. Obviously not a good sign considering today.
Yikes!

Fibbgann

[Gavishti]

That is an example of why gspiker and I suggested that you put a different indicator in your mix. Nearly every indicator I use is saying to short the market. A few say it is questionable or that I should be long. If you are going to use multiple indicators, they should not be all of the same variety. This will help get rid of the problem of being on the wrong side of the trade during big momentum moves like we have had this week. You may miss the move entirely, but at least you won't be on the wrong side. Then the signals that do get through are good ones.

Once you have all of that figured out and coded you can add something to deal with the big momentum moves as an exception.

Which indicators will make the best mix is something a lot of people have been trying to figure out. There are many approaches to the problem of getting good signals and no consistent answer. You can find as many opinons as there are books on technical analysis and trading. Plus all the chat, forums, and magazines. You have to find a combination that makes sense to you.

Good luck. Your signal will most likely be correct, just early. The question becomes how much pain can you endure in this system?

[FibbGann]

Gav,

Thanks for your reply. We are aware of this, we just have to make a few tweaks. Frankly, I like the indicators we are using and think they give good signals. Alex has already made a revision, that after thinking about was a great idea. Just making notes on signals here.

Fibbgann