SmartQuant Discussion

Automated Quantitative Strategy Development, SmartQuant Product Discussion and Technical Support Forums
It is currently Sun Apr 05, 2020 10:45 pm

All times are UTC + 3 hours




Post new topic Reply to topic  [ 4 posts ] 
Author Message
 Post subject: System robustness
PostPosted: Wed Dec 23, 2009 1:54 am 
Offline

Joined: Wed Dec 23, 2009 1:45 am
Posts: 2
In using indicators to create a system one often finds the use of moving averages, standard deviation etc. which are are subject to variation even with a small change in the underlying price series. What are the boards views on using more robust measures such as the median, trimean, interquartile range et al in the construction of indicators?


Top
 Profile  
 
 Post subject:
PostPosted: Wed Dec 23, 2009 8:13 pm 
Offline

Joined: Thu Jun 08, 2006 3:56 pm
Posts: 537
Location: BC Canada
FWIW, my experience agrees with yours -- small changes in prices affect indicator values. But this effect holds true for all kinds of indicators, regardless of whether you use close, median, weighted, etc. All anyone can say about the various BarData types (close, median, etc) is that their values change with price according to the mathmatics of the indicator.

In real life I've found that the BarData types that combine multiple bar price points (averaging open/close/high/low in some way) tend to be less volatile. Sometimes that is an advantage for some strategies, and sometimes it is not. It probably depends on your strategy and what you want to do with it.

For example, for a long slow trend following strategy, it probably doesn't matter what BarData type you use, because a single bar price is very small compared to the big trend price swings. In contrast, if you have a fast contra-trend strategy that emphasizes price levels more than moving average values, it might work better if you just use the Close price alone.


Top
 Profile  
 
 Post subject:
PostPosted: Sun Dec 27, 2009 7:24 pm 
Offline

Joined: Wed Dec 23, 2009 1:45 am
Posts: 2
"small changes in prices affect indicator values. But this effect holds true for all kinds of indicators, regardless of whether you use close, median, weighted, etc" Not so I'm afraid. The median of 1, 2, 3, 4, 5 is 3; the median of -1, -0.5, 3, 7, 100 is also 3. If, for example you are attempting to measure the central tendency of a trading range, the median is much more likely to be unaffected by outliers, spurious/bad data etc and if similar market conditions occur again, the hope is that maybe the median measure will be more similar to past price occurance's measurements than those given by other metrics i.e. the mean


Top
 Profile  
 
 Post subject:
PostPosted: Mon Dec 28, 2009 2:27 am 
Offline

Joined: Thu Jun 08, 2006 3:56 pm
Posts: 537
Location: BC Canada
Well, your short mathematical examples are surely accurate, but it seems to me that the examples are far from practical trading reality using moving averages etc. For example, what reasonable prices move from 1 to 5, or from -1 to 100 in 5 bars, quotes, or trades, or in 5 steps? And especially where moving indicator values are used (EMA, SMA, etc), the example values would still produce clear changes in output values with each new value. So I'm not sure what the examples are intended to illustrate.

You also say that "the median is much more likely to be unaffected by outliers, spurious/bad data etc", but to be precise, the median will of course be 100% affected by outliers and bad data, exactly in proportion to the median and indicator mathematics calculations, no more and no less.

My main claim is only that the values of essentially all practical indicators will vary with changes in price inputs, regardless of what combination of BarData values are used. Some combinations (eg median, weighted, etc) may cause the indicator values to change a little less, some a little more.

Isn't that the way it should be?---the mathematics of an indicator should force a change in the indicator output values when the input values change---else what good would a constant indicator be?

I could be wrong, but it seems to me that the only real question about any indicator is whether or not its input and indicator mathematics have the combined characteristics that you want for your strategies.

I suppose that's part of the strategy game---to find or make indicators that have the change (indicator) characteristics that you want for your strategies. And of course no one but you can say whether a particular set of indicator mathematics for inputs (close, median, weighted, etc) and outputs (change in indicator value) is satisfactory for your strategy.


Top
 Profile  
 
Display posts from previous:  Sort by  
Post new topic Reply to topic  [ 4 posts ] 

All times are UTC + 3 hours


Who is online

Users browsing this forum: No registered users and 1 guest


You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group