Skip to content

Short Term RSI as a market timer

May 30, 2009

Short Term RSI, or Relative Strength Indicator over very short windows of time, popularly RSI(5) and RSI(2) are widely talked about in the online trading literature. The hypothesis is, to detect short term often BTST or STBT opportunities on one’s favour by tracking extreme values of RSI.

Said in other words, we detect periods of extreme oversold-ness and use it to buy. And when we detect days when price moves into overbought levels , we short. How do we detect extremeness? By detecting extreme values of RSI(2).

Two things can be easily discerned:

  • RSI(n) is a smoothened version of RSI. Higher the ‘n’, smoother the RSI. Since n here is 2, hence we can see extreme movements very often
  • More importantly, usual levels to have long or short bias is 80-20 levels. We tweak that to form extreme levels like 95-5, 90-10 or even assymmetric levels like 95-25, 7-75 etc

Now lets put some perspective into it. My analysis tells me in NIFTY, since 06/10/2005 to 16/04/2009, around 1000 days, we had 271 days of extreme RSI(2). The extreme levels being 90-10, proper sweet symmetric levels. Nice, cute and easy.

Note, the period includes one bull run and one bear run.Lets compare the very next day returns. The returns are divided into two categories:- seasonal and secular. i.e during a bull run how well do oversold indicators work, and vice versa [this is seasonal mean next day returns] and during the entire period how did oversold indicators fare. Same for overbought levels.

T+1 profit for Overbought Levels: RSI(2)>90

Secular Mean: 0.46 points
Seasonal Mean:17.70 points

T+1 profit for Oversold Levels: RSI(2)<10

Secular Mean:18.89points [2k5-2k9]*
Secular Mean:7.08points[2k6-2k9]**
Seasonal Mean:26.46points

Now there are some nitty gritty details in it. For OS levels, there are two secular means and that is to provide you with an overall balanced picture. We must not forget two things ,we were living in the midst of an extraordinary bull run from 2003- 2007*. And clearly that bias or effect  can be seen on the mean also. To provide an equal and clear balance to both, I have provided with a data slice which has equal parts of bull and bear run. 2006-2007 is a raging bull, while 2008-2009 is mad bear**. But the effect is clearly seen. If our market doesnt turn by this year, I am sure by the end of this year we will see the means skew even more, towards randomness.

Now, we can at the very outset, see this, that in seasons the contrarian indications/ levels have a clear advantage.

This was the hypothesis. What about forming a system around it? I will take up two variations, symmetric levels and unsymmetric levels of RSI(2) extremeness.

Lets work with unsymmetric levels first: Our first approach is to, check the barebone performance without any position sizing whatsoever, to check and have a basic idea about maximum adverse excursion, favourable excursion, CAR/Max DD ratio etc etc etc. Now the feature of this test is to buy when RSI(2) dips below 7, and exit when it goes above 90, short when it goes above 95, and cover when it goes below 10. Simple enough, eh? Some of the few selected statistics are:


Net Profit %age -32.6%
Total Winners 69.77% [48.84(L)+20.93(S)]
Max Trade DD -50%
Max System DD -53%

The results were so hopeless, that I decided not to harangue over other important metrics for something which is so clear. The next iteration involved profit booking, and it was very conservative trading. Booking profits at the first sign of trade going in our favour and that shows in the results. The second iteration statistics are as follows:

Net Profit 11.82%
Total Trades 47
Winners 65.96%[25.53%[L]+40.43%[S]]
Max Trade DD -36.43%
Max System DD -36.79%
CAR/Max DD 0.08
Profit Factor 1.16
Risk Reward Ratio 0.15

Clearly, even this  strategy is nowhere to fall over upon. But yes its profitable, nevertheless testing. An important thing people may notice and ask is  there were more short side winners than long side winners, which sort of contradicts our assumption that mean returns of  OB levels weren’t that lucrative. But you forget that the stats given above at the very beginnng discussed the very next day returns.

Now coming to symmetric levels, with buy at 5, exit at 95 and short at 90 cover at 10 weren’t even remotely profitable. So I will spare my dear reader the torture of watching such statistics.

On a serious note, I would encourage you to test on your own NIFTY data and discuss any possiblities of turning it into a better system. You can mail me[ mail id given in about page] or leave a comment here.

Till Then,
Adios

P.S: On public demand I am okay to provide with MAE/MFE charts. Do ask if you require.

Advertisements
No comments yet

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: