Page 1 of 1

Questions on indicators and time frames

Posted: Mon Dec 12, 2011 1:06 am
by trending3029
I have been facing a lot of whipsaws lately with my breakout system. I was wondering is there is a filter / indicator to identify whipsaw market and thereby reduce risk or even avoid taking the signal (based on accuracy of whipsaw filter).

Long term, medium term and short term trend following meant using relevant indicator lengths. How does this work with daily, weekly and monthly data?

I mean does the data have an affect on the performance of systems generally from experience?

Please share some of your experiences within the relevant areas.

Posted: Tue Dec 13, 2011 3:11 am
by TrendsCatcher
How about using your P/L as a feedback input to your position sizing, e.g., if n losses in a row, decrease bet size by m%. Then if x winners in a row, increase bet size by y% until you are trading full size again. It is like the sampling rate concept, you are running low-resolution "preview" in a hostile environment (your P/L will tell you), only when the favorable environment comes along (your P/L will tell you), i.e., a bona fide trend, you use the highest sampling rate to capture the market and get a picture-perfect representation.

Posted: Wed Dec 14, 2011 10:19 pm
by michaelt
trending3029,

Another possibility is to judge the environment with a volatility measurement. You could look at current market volatility - whatever you want current to mean, versus a longer-term volatility. When the ratio fits what you think it should be - say, high enough for break-out systems to follow though, you can test whether that does improve results. Or maybe a high ratio is not good - that environment has more whipsaws - and you can use the environment ratio to reduce or eliminate trades (of that nature).

Posted: Thu Dec 15, 2011 8:10 pm
by trending3029
michaelt, TrendsCatcher thanks for your responses.

TrendsCatcher, your advice of reducing the bet size (fractional) is something that i currently follow. My rule uses % bet-size reduction for a fixed string of losses continuously.

michaelt, your advice of volatility measurement definitely makes logical sense, however using current market volatility (VIX) alone is something i probably would not be inclined to use. The reason i say that is I'd need to determine the correlation between the current market volatility (VIX) and the relevant asset (for which I've got the signal generated) and determine its viability based on historical data, and then use that as a precise decision maker. I would see this as curve fitting or making the model more rigid.

I've however read other posts in the forum to be able to use ADX as a filter and ATR for position sizing which make a logical choice.

The second part of my question was pertaining to Short Term Trend (STT) Following, Medium Term Trend (MTT) following and Long Term trend (LTT) following:

I've read a few posts on the forum which use indicator length to dictate
STT, LTT, MTT. I was curious to see if anyone had ideas based on use of data i.e (weekly, fortnightly, monthly, etc) rather than use of the indicator length.

Posted: Fri Dec 16, 2011 3:37 am
by Moto moto
when it comes to tweaking random indicators and numbers - I think you will find for every problem you fix, another rears its head.
Everything becomes a trade off between missing out, getting in too early, to late, too big, too small. Most will boil down to what you can live with, makes sense and is not going to blow you up....then hopefully it makes money.
Thats the search for the holy grail as they say and why people back test - to give an indication of what to expect given certain parameters.

Posted: Fri Dec 16, 2011 7:38 am
by babelproofreader
I've read a few posts on the forum which use indicator length to dictate
STT, LTT, MTT. I was curious to see if anyone had ideas based on use of data i.e (weekly, fortnightly, monthly, etc) rather than use of the indicator length.
You could make your indicator length adaptive to the cycle length in your data e.g. if your monthly data exhibits a cycle length of say 8 months, then 8 months * 4 weeks * 5 days = 160 days (or multiples thereof) indicator length for your LTT etc.

Posted: Fri Dec 16, 2011 10:54 am
by michaelt
"michaelt, your advice of volatility measurement definitely makes logical sense, however using current market volatility (VIX) alone is something i probably would not be inclined to use. The reason i say that is I'd need to determine the correlation between the current market volatility (VIX) and the relevant asset "

I wasn't suggesting VIX as the measure for your purpose. You can create other measures of volatility.

Rather, I meant to give an idea of a way to look at the market structurally and then use (and if necessary, create) measures that serve your purpose to keep your system acting when you want it to.

For example, perhaps your system approach works best when the market made new highs (lows) in the past 40 days (weeks). Then you can determine for how long that measure has use, and then allow your system entry signal up to that length of time.

You can substitute any measure to convert to a 'status' (where 'status' means it's ok to take entries). The measure could be VIX but it could be any idea you have about determining a market environment.

Posted: Mon Jan 02, 2012 4:56 pm
by trending3029
I'd like to thank: michaelt, babelproofreader, Moto moto and TrendsCatcher.

Michaelt, thank you for your suggestion regarding creating a measure for volatility.