System parameters

[TabJockey]

No no you misunderstand me.

The first thing to take notice of and the most important aspect of robustivity is the right sort of relationship between parameter sets (already defined).

Comparative performance using multiple parameters = robust

Comparative performance using multiple markets / instruments = increased probability of the system being robust

Non comparative performance using multiple parameters = not robust

Non comparative performance using multiple markets / instruments = inconclusive

Comparative performance using multiple markets / instruments AND multiple parameters = very robust (ie. where do I send my cheque?)

 

[Billyb]

Didn't want to start a new thread for this question-

What is acceptable liquidity on the ASX for a trend following system? I've noticed that in my system, when I tighten the liquidity filter even only a little bit over $150000 turnover (volumes times Close price), results begin to suffer significantly. Not really worried about getting into the stock, just worried whether getting out will be a problem. Would not be placing any more than $5000 in any one trade..

 

[TabJockey]

Didn't want to start a new thread for this question-

What is acceptable liquidity on the ASX for a trend following system? I've noticed that in my system, when I tighten the liquidity filter even only a little bit over $150000 turnover (volumes times Close price), results begin to suffer significantly. Not really worried about getting into the stock, just worried whether getting out will be a problem. Would not be placing any more than $5000 in any one trade..

Depends if you are trading at the open or the close. Allot of trades are executed at below the published open price, so I have found that simulated trades on open prices only need a small margin of error.

The fact that your results suffer significantly when you filter out those low liquidity stocks means that if you live test the system, it quite possibly will give crap results. Sounds like a warning sign to me.

 

[Billyb]

Depends if you are trading at the open or the close. Allot of trades are executed at below the published open price, so I have found that simulated trades on open prices only need a small margin of error.

Thanks for your reply. I should have mentioned how the system is quite long term. I can see how this would be a major problem with a short term system. However, I don't think it should matter too much with systems where the average holding time is longer term (weeks or months).

The fact that your results suffer significantly when you filter out those low liquidity stocks means that if you live test the system, it quite possibly will give crap results. Sounds like a warning sign to me.

I think it's more because low liquidity stocks are the stocks that have more potential to trend strongly and become the big winners. I don't think it's a problem with my system, it's just how stocks behave eg BHP (high liquidity) vs AUT (lower liquidity) since 2009.

 

[tech/a]

Generally I agree with Tab J
That's why you put in a liquidity filter.
But I also agree that if a small cap increases in price it generally increases in liquidity.
The problem in my view happens in the middle where if you get severe volatility you could well find a lot of slippage getting out of a small cap.
Then again $5000 shouldn't be much of on issue.
I use 300k and have no problems finding prospects.
If a stock is growing it will soon reach the liquidity filter.

 

[Wysiwyg]

Didn't want to start a new thread for this question-

What is acceptable liquidity on the ASX for a trend following system?

From my learners viewpoint it is 'generally' the high turnover blue chips that meet the liquidity filter of above 150k. What needs to be considered is the look back period for this sum. A small cap. 'generally' has a rush of higher turnover then a lean period as opposed to a large cap. that meets the liquidity filter all the time.

In stock selection using position score (see position score in documents) I use the correlation function which is supposed to rank stocks correlated to the All Ordinaries. Notice how I have incorporated a linear regression slope above 0 for 20 periods on the All Ords. to rank stocks accordingly. These parameters can be altered and ultimately when testing history, the period look back has a significant bearing. Also I have a shorter period on the liquidity filter so those initial rushes for less traded stocks have a chance at selection.

Index = LinRegSlope(Foreign("^AORD", "C", True), 20) > 0;
PositionScore = Correlation(Index, C, 10);

Turnover = MA(C*V, 10) > 50000;

 

[Wysiwyg]

In stock selection using position score (see position score in documents) I use the correlation function which is supposed to rank stocks correlated to the All Ordinaries.

Note the stocks are ranked using position score and not the ultimate selection criteria. Had to make that clear.

 

[Billyb]

I use 300k and have no problems finding prospects.

300k! I wont worry too much then haha.

In stock selection using position score (see position score in documents) I use the correlation function which is supposed to rank stocks correlated to the All Ordinaries. Notice how I have incorporated a linear regression slope above 0 for 20 periods on the All Ords. to rank stocks accordingly. These parameters can be altered and ultimately when testing history, the period look back has a significant bearing. Also I have a shorter period on the liquidity filter so those initial rushes for less traded stocks have a chance at selection.

Index = LinRegSlope(Foreign("^AORD", "C", True), 20) > 0;
PositionScore = Correlation(Index, C, 10);

Turnover = MA(C*V, 10) > 50000;

Nice idea

 

[Billyb]

By the way Wysi, the blue chips on my data have turnover much more than 150k - they are in the 10s if not 100's of millions

 

[rnr]

Didn't want to start a new thread for this question-

What is acceptable liquidity on the ASX for a trend following system? I've noticed that in my system, when I tighten the liquidity filter even only a little bit over $150000 turnover (volumes times Close price), results begin to suffer significantly. Not really worried about getting into the stock, just worried whether getting out will be a problem. Would not be placing any more than $5000 in any one trade..

Have you considered using a tiered liquidity filter?

e.g.

Close >= 0.20 but <= 1.00 then Liquidity >= $150,000
Close > 1.00 but <= 2.00 then liquidity >= $250,000 etc, etc