Experiments in Technical Analysis

[SGM]

Hello Karthik, Murthy

I wanted guidance from you regarding two issues or may be both issues are different aspects of Trade Management.

1. Stop Loss
2. Taking a trade with older signals/ Re-entry

With all the systems we are discussing above, I have observed that using a smaller value for stop loss lets say 2 to 3% turns out to be disastrous for the over all performance of the system. Anything less than 5% will wipsaw and stop us out too soon, resulting in mediocre results. Even the trailing Stop should be above 10% preferably 12-15%. Thus we need to keep a stop loss of 5% and Trail Stop of 10-12%.

There are 2 consequences, the initial drawdown % will be relatively high and thus the faith to continue with the system to take all the signals will waver. Second the over all position size will reduce considerably and we will hardly be able to use any leverage, even if we are working in FnO segment.

The solution could be to use a smaller SL initially but keep tracking the position that has been stopped out. As long as the reversal signal is not given we keep taking multiple entries for the same trade, maybe using a SL order on the initial side of the trade.

That bring out another point, how do we enter a trade for which the entry signal is old, what would be the criteria for it? Can we use pullback and continuation (reversal of pullback)? Maybe we can use a signal based on distance to cross over, something similar to what Anant is using in his method. { D > Ref(D,-1) and Ref(D,-1) < Ref(D,-2)}

Maybe all this sounds a bit cryptic, but I just hope I am able to put across my doubts.

Warm Regards
Sanjay

 

[karthikmarar]

Hi Karthik,

Another point i thought would be useful to discuss is the number of indicators a person should use. one problem i face continously is my bias coming tilting decisions the way It wants it to. if i think of a stock to buy and start to study it. I see many indicators and some or the other indicator shows what i want to see. so basically we should keep a fixed number of indicators we refer to and not add every other time and that will help keep the bias view away.

True, most of us try to see what we want to see. Many keep looking at various indicators till they see what they want to see. Sticking to coupling of Indicators does help to over come this bias to some extend. Strictly following you system rules is a must.

Can we use fibonacci and bollinger with other indicators. i am using RSI with a smoothed rsi. i use MACD. and apart from this I am using support resistance and volume. would adding fibonacci interfere with support levels or can we use it to just see projections and with the indicators I am using will bollinger be a addon or will it not fit in with the other indicators. there has to be a set of indicators that work with each other( based on how they are calculated, periods and what they use to be calculated) and some that dont fit the team of indicators we are using. when i was doing my hotel management and were learnign about cocktails there were certain alcohols we could mix and some we could not based on certain criterias . Wonder if itwould be same with indicators. because a combination and weightage of diff indicators/alcohol in the right amount would make the difference

Any Views on this.

There is no harm in using Fibonacci and BB along with other Indicators, while Fibo levels are good for support/ resistance levels and projections, BB can tell us about possible reversal/continuation in the immediate trend.

Some combinations of Indicators do form a good cocktail. We will deal with them later. Right now we will concentrate on the wonderful exercise being co-ordinated by Sanjay.

Regards

Karthik

 

[karthikmarar]

Hi swagat

Nice to see you post on this thread. Also glad to hear the Mabiuts and the AMA system work fine on intraday data. Did you check on real time data?

Of course we can short when we get sell signals. But shorting when the overall trend is up can be dangerous. One has to be careful about this.

Looking forward to your chipping in with some systems. We can all learn new things.

Warm regards

Karthik

 

[murthymsr]

dear sanjay,

With all the systems we are discussing above, I have observed that using a smaller value for stop loss lets say 2 to 3% turns out to be disastrous for the over all performance of the system. Anything less than 5% will wipsaw and stop us out too soon, resulting in mediocre results. Even the trailing Stop should be above 10% preferably 12-15%. Thus we need to keep a stop loss of 5% and Trail Stop of 10-12%.

There are 2 consequences, the initial drawdown % will be relatively high and thus the faith to continue with the system to take all the signals will waver. Second the over all position size will reduce considerably and we will hardly be able to use any leverage, even if we are working in FnO segment.

my observation also goes the same way. often when i do not set any stop-loss, i am getting a better annualised return. this could be because of the cyclic nature of the stock prices. a stop-loss of even more than 5 % may be ok, even 8 to 10 % may even give better results.

but with a higher value of the stop-loss, the draw-down will also be higher. but, i feel, that this draw-down can be reduced by ensuring that we will be holding a minimum of 3 stocks and a max of 8 stocks (average= 5 stocks).

but how we can achieve this in our testing? may be the AFL may also the rank the tradeable set of stocks and will continouously replace the held stocks with 'better' ones such that the total number of stocks may be maintained within the range. but, of course, this will be a totally new subject and may be ignored for the time being.

another observation was that the number of trades executed by the back-test module is too low to be realistic. i feel that a short / medium term trader may normally does on an average 100 to 200 trades in a year ( means buying & selling of 50 to 100 stocks).

The solution could be to use a smaller SL initially but keep tracking the position that has been stopped out. As long as the reversal signal is not given we keep taking multiple entries for the same trade, maybe using a SL order on the initial side of the trade.

but this may mean human intervention and may vary depending upon the user's perceptions. the results cannot be entirely be attributed to the efficacy of the system.

That bring out another point, how do we enter a trade for which the entry signal is old, what would be the criteria for it? Can we use pullback and continuation (reversal of pullback)? Maybe we can use a signal based on distance to cross over, something similar to what Anant is using in his method. { D > Ref(D,-1) and Ref(D,-1) < Ref(D,-2)}

the first few days of data ( may be around 20) are used to arrive at the moving averages and other parameters after which only buy/sell signals can be generated. by taking a longer time period and ignoring the 'old signals', we may still have enough signals to make the back-test truely representative.

Maybe all this sounds a bit cryptic, but I just hope I am able to put across my doubts.

i am sure i have not cleared any of your doubts, but added my own ! To make the exercise a fruitful one, it may be desirable to have one or more 'acroos the chat room' meetings, to arrive at meaningful values to the test settings logical and common to all volunteers. the testing may be started only after such common settings are arrived at. even we may agree on a range of values for each setting and the tester will also comeout with the 'best' values within that range.

sanjay, you may organize such a session and all participents may be informed. and probably we need to get some more volunteers to test the systems.


Warm Regards
murthymsr

 

[karthikmarar]

Hi Sanjay / Murthy

Sharing my thoughts on the points raised by you

Stoploss.

As you have rightly pointed out smaller stop-loss value will take one out of the trade much sooner or we get whipsawed often. Having a larger value of stop-loss would also mean initial high (start-trade) drawdown.

As you suggested the best solution would be to have a smaller stop-loss during the initial period of trade after entry. Once the stock moves up to a value equivalent to the initial risk then stop could be raised to breakeven. Once the entry has been proved to be a good one and an uptrend has been established then we can switch-over to a wider trail-stop. It could be a percentage stop-loss or an ATR stop-loss or a last low pivot stop-loss. If we are successful in catching a large trend we can even switch over to a trendline stop-loss ( with a filter to avoid false trend line breaks). In AB these switch-overs can be easily programmed.

If the position sizing is based on the risk of each trade, then the lower initial risk will affect the position sizing. So it boils down to the question whether we gong to size based on the lower initial start trade risk or the overall risk (2% of total equity) we have decided.

Taking a Trade on Older signal / Re-entry

We can take a trade on older signal or re-enter using the pull back and continuation. The Mabiuts-K system does provide re-entry signals, but there are a little delayed as the condition peak(c,2,1) is used. So the re-entry signal is given only after the price has crossed the earlier peak on its continuation path.

We can use the stochastic Indicator effectively for re-entry. Ananths Histogram approach is very good for re-entry. Even the MACD Histogram turning up when the MACD is above the signal line ad moving up provides good re-entry signals. These can be discussed in detail later.

Warm regards

Karthik

 

[ragsid]

Praveen,

I think you have created the link "amibroker links" under amibroker, can we have all of these AFL's there and can you make that thread sticky?

Rags

PS. I felt both, your thread as well as the AFL's developed here are great (eg. mabiuts gives great signals, nobody has any doubt about it).

 

[swagat86]

hi frnds just tryin to get into perscpective ATR study. Put it in AUto analyser and hav a look. The gains are huge.

Not much idea of the ATr trading System.
May be u Expert guys are gonna work it out.

PLs dont leav this pls hav a look.

Thanks

 

[SGM]

Hello Friends

To make the exercise a fruitful one, it may be desirable to have one or more 'acroos the chat room' meetings, to arrive at meaningful values to the test settings logical and common to all volunteers. the testing may be started only after such common settings are arrived at. even we may agree on a range of values for each setting and the tester will also comeout with the 'best' values within that range.

Thats a good idea. As we wait for few more volunteers to join in, meanwhile we can start a trial run from friday/monday so that other things like how to report daily signals & trades and how to maintain records can be sorted out.

I think a single post after market hours (evening/night) can list the trades and the signals genrated by the system. No need to give detailed accounts/financial statements just total value of equity would be enf. Please post your views

and probably we need to get some more volunteers to test the systems.

Warm Regards
Sanjay

 

[asnavale]

Hi Karthik,

This is a post with respect to the AMA System which you posted earlier. I applied this system to two stocks and following are the results. I have taken the results of 3I-INFO from the chart you posted (for the period May 2005 to April 2006) and run the system on INDSWFTLTD for the period 1-1-06 to 28-11-06. I also tested both these two stocks under the respective periods with MABIUTS-H and two modifications of MABIUTS. The profits shown are gross, without any commission etc. The following results are given only for just two stocks and should not be taken as indicators of performance. They have to be evaluated for more stocks under different conditions.

3I-INFO:

AMA System: 14 trades (7Buy + 7Sell), Profit 30.51%

MABIUTS-H: 54 trades (27 B + 27 S), Profit 214.3%

MABIUTS-H (modification 1): 20trades (10 B + 10 S), Profit 146.88%

MABIUTS-H (modification 2): 26 trades (13 B + 13 S), Profit 153%

INDSWFTLTD:

AMA System: 14 trades (7+7), Loss 27.15%

MABIUTS-H: 42 trades (21 + 21), Profit 96.36%

MABIUTS-H(modification 1): 20 trades (10+10), Profit 55.86%

MABIUTS-H(modification 2): 14 trades (7+7), Profit 38.54%

-Anant

 

[karthikmarar]

Hi Ananth


Interesting .. please do post the modification of the Mabiuts-H system.

By the way, The AMA system does no render itself very well for back testing. There are many parameters to adjust like the Sensitivity and Fine-tune.

It is more like listening to music, we adjust the Bass, treble, balance to our individual liking so that we have the best listening pleasure. Likewise the AMA system parameters can be adjusted to ones own trading style. Also optimizing the parameters does not work well as the parameter directly reflect different trading styles.

Warm regards

Karthik