Friday, October 2, 2009

Reducing Whipsaws in Moving Average Systems

Aurobindo has an interesting question:
" If 3 MVA crosses 50 MVA 2 times without yielding profit then on 3rd crossing enter the market. We may miss some opportunites but we will increase the chances of profit and avoid whipsaws. Usually after the end of a long trend , the next trend signal will definitely yield in a loss thats due to a fibonacci retracement. "

My Notes:
Instead of taking every moving average trade, what happens if we trade only after two losses? The assumption is that the two losses take care of the inevitable whipsaws, thus increasing the chances of the next trade.
I put this into a tradestation system. My code says:
Buy when price closes above the 5 day MA and sell when it closes below.
Take all trades after at least 2 losses.
Here are the test results on End of Day Nifty near month futures from 2000 to Oct 1, 2009. Costs and slippage have been taken at 5 points per trade.

System : Raw (basic crossover)
Net Profit: +2105
Drawdown: + 1259
Number of trades: 526
Net per trade: 4.00
%profitable: 36.3

System:Trade after 2 loss
Net Profit: +3054
Drawdown: 487
Number of trades: 174
Net per trade: 17.56
%profitable: 41.95

Conclusion: trading only after 2 losses reduces whipsaws and increases the efficiency of the moving average system. Aurobindo's system used 2 averages, while I did my test on MA and close. I think similar results will be found on 2 MA. I will do the test over the weekend, and, also test on intraday data. Like most trend following systems, winners are concentrated in small clusters. Only system traders can trade these systems.
One caveat: while I did try to check my programming, I wrote the code quickly so there is a small possibility of errors. I will recheck and come back over the weekend.

A word about Google. My last post was on wednesday. But my adsense earnings yesterday (this comes when readers click on ads displayed by google in my blog) were above average, suggesting that probbaly the right ads were clicked. I do not know which ones come in this category, but Google's technology is awesome.


Lalit said...

Hello Sir
This is Lalit. I met u at Investor Camp held in Chennai on Sat. A very Big fan of urs. and do said that i am doing my CA Final. Hope u remember me now. Just wanted to ask u up that do u have any office of urs at Chennai or any such plans to have so. since Finance has always been my Passion and hence i taught to join and work under u sir.
Waiting for u r reply eagerly.

E-mail id-
Ph No- 09884863049

gurvi said...

Sudarshan ji,
Regarding Advertisements:-
As you might remember I always say:-
I click at least 5 ads when I visit your blog. (At least to give you sth back)
I'm sure other readers also do the same :)


wildeazoscar said...

A question on the basis of your being associated with the ATA---

"What are the job-prospects of a person with the CFTe certificate (not MFT) in India?"

On second thoughts----
" if someone in a secured career with an active interest in trading wishes to take up trading as his source of subsistence by quitting his job(which is nowhere related to any trading activities), would a CFTe certification help him earn a job, if after some period he wishes to work and at the same time trade?"

jagjit said...

My count:-->

Hi sukhaniji,
As everybody is giving his own wave count of nifty i decided to put on the table what i feel. Please do make a chart representation of it if you think its necessary as i'm unable to attach charts with this message.

And please do point out where i went wrong if you think count is not correct.

There are two possibilities--on the basis of what i've studied about Elliot Wave:

Possibility Number One:--

1.Wave count should starts from an important point which may not be the lowest low of the nifty. One such point is 2539 so i'd start my wave count fron there.

On that basis i consider W1=4693->2539

2.W2=4693->3918 which is <38.2% W1

3.W3 started from 3918 and is still in W3 can't be the smallest wave so it has to be >W1 Therefore minimum target for W3 comes out to be >3918+(4693-2539)
=> W3 > 6072

Possibility Number Two:--

I've read anohter twisting point about the Elliot wave theory--that some times W1 is in the form of a rising wedge and following conditions should hold for that--

W1 sohuld be the largest wave

1.W2 should be <38.2% W1
2.W3 should be <61,8% W1
3.W5 should be the shortest of the three thrust waves

(Odds:If W3 is 38.2% W1 then W5 will be 61,8% W3)

4.W1 "must be" 161.8% W5

Now if we see the move of nifty from 2539->(W1)4693->(W2)3918->(W3)4739->(W4)4353 it comes out in the form of a rising wedge and the first two conditions hold.

Condition 3 doesn't hold as W5>W3 in this case.

But if the "must be" condition 4 is to be held then Target for this W5 comes out be 5684

(Also if we make a rising triangle joining 4693-3918-4743 and considering the breakout from 4750 target of this triangle is 5532)

So the target of the nifty in the coming months could be around 5500-5700 before a major corrective move begins.

Please tell me which count can be the nearer to the correct count.

Jagjit singh

kaps said...

Elliot wave count update from my view:We may be in ABC flat movement on the yearly/monthly charts.
WaveA => 6357-2252/2539(zigzag)
WaveB => 2539 - continuing
Wave B further splitting into abc.
wavea=> 2539-5100
waveb=> 5100-3600
wavec=> 3600-6200
(taking a=c)
So If indeed we are in ABC flat correction on monthly basis,we should see Nifty going to 3600 levels before rising again to 5000-5500+ levels.