Trading Regime Analysis Using RWI

[Do Or Die 10]

Hi,

 

Michael Poulos shot into fame in early 90’s when he published an article about RWI. He showed that Corn has the strongest tendency to trend while Wheat has the least. RWI is not talked about much because Poulos was more interested in his trading rather than writing books to commercialize it. It can be used to rank stocks for their trend strength. As a comparison, RWI is conceptually better in mathematical construction than ADX.

 

RWI ‘ranks’ an instrument’s tendency to trend by comparing it to a random walk. This is very useful for Regime Analysis because traders can apply trend following strategies to instruments which show a greater opportunity to trend. Similarly traders shift to mean reversion strategies for instruments which tend to move around a mean (‘more stationary’).

 

The earlier articles in this series are:

Trend Following Vs Mean Reversion: Trading Regimes

Introduction to Understanding Volatility

Trading Regime Analysis Using RSI

Relative Strength - Internal

 

Note that within an instrument, regime shifts may appear favoring trend following or mean reversion strategies. However, in this article I want to discuss the trending tendency of an instrument in general. Some instruments tend to exhibit stronger trending tendencies than others always.

 

Take the stocks for example. Which sector stock do you think are most trending on EOD? Consumer Goods and Technology are the most strongly trending, while Utilities and Transportation are least trending. This ofcourse is drawn on ETFs, but a similar model can be drawn for individual stocks. The general conclusion here is Technology stocks favor trend following strategies while Utility stock favor mean reversion strategies.

The above chart shows RWI averaged over 100 periods for some ETFs. Since RWI shows a trending tendency, the wilder the RWI extremes for an instrument, the greater is it’s tendency to trend.

 

Let’s look at now what makes RWI effective for such conclusions:

• It compares trends directly with random drift
  
• Objectivity is maintained by not requiring an arbitrary lookback interval (for example, the length of a moving average)
  

RWI has two components-

Random Walk Index of Highs (measures uptrend) RWHI= (H-Ln) / (R*sqrt(n))

Random Walk Index of Downtrends (measures downtrend) RWLI= (Hn-L) / (R*sqrt(n))

Here ‘Hn’ and Ln’ are Highs and lows n days back respectively and R is the average range.

 

The sqrt comes into picture from Mathematician W. Feller’s proof that a “random walk” generated by tossing a coin (one step forward if heads, one step backward if tails) would show a displacement from the starting point, depending on the square root of the number of tosses.

 

RWI is calculated as the difference between RWHI and RWLI. It therefore effectively measures the tendency to trend compared against randomness.

 

The closer the RWI value remains around Zero, the lesser the tendency of an instrument to trend.

 

Here are some major indices ranked according to their trends over past one year:

Index Ticker	Index Name	RWI Average
IXUT	NASDAQ Telecommun	0.02
HUI	AMEX GOLD BUGS IN	0.11
XAU	PHLX Gold/Silver	0.13
TYX	Treasury Yield 30	0.16
IXBK	NASDAQ Bank	0.19
AORD	ALL ORDINARIES	0.22
DJU	Dow Jones Utility	0.30
DJT	Dow Jones Transpo	0.36
IXTR	NASDAQ Transporta	0.37
IXIS	NASDAQ Insurance	0.40
DJA	Dow Jones Composi	0.43
IIX	AMEX INTERACTIVE	0.45
NWX	AMEX NETWORKING I	0.50
SOXX	PHILADELPHIA SEMICONDUCTOR INDEX	0.52
DJI	DOW 30 INDEX	0.53
IXFN	NASDAQ Other Fina	0.56
N225	NIKKEI 225	0.57
VLIC	Value Line Geomet	0.66
XAX	AMEX COMPOSITE IN	0.67
NYA	NYSE COMPOSITE IN	0.68
RUT	Russell 2000	0.70
IXK	NASDAQ Computer	0.72
RUI	Russell 1000	0.73
SML	S&P SMALLCAP 600	0.73
NDX	NASDAQ-100	0.74
RUA	Russell 3000	0.75
IXIC	NASDAQ Composite	0.76
PSE	NYSE Arca Tech 10	0.77
BTK	AMEX BIOTECHNOLOG	0.78
OEX	S&P 100 INDEX	0.78
DWC	DJUS Market Index	0.79
GSPC	S&P 500 INDEXRTH	0.80
MID	S&P 400 MIDCAP IN	0.82
XOI	AMEX OIL INDEX	0.82
NBI	NASDAQ Biotechnol	0.89
IXID	NASDAQ Industrial	0.91

The RWHI and RWLI indices are calculated over a range of lookback lengths using the largest value returned for today’s indicator. Thus, we let the market determine the lookback interval, rather than use a fixed arbitrary one as many current indicators do.

 

Now we know, trends exist in all freely traded markets. The swings in RWI are cause by short-term exception to the rule that trend exists.

 

This leads us to other very important use by plotting a short-term RWI against a longer term. The short-term is a good over bought/oversold indicator while the longer one is a good trend indicator.

 

Posting a comment will only take you 2 minutes, but it will be the strongest motivation for me to share something better.

[umfan92 10]

Posting a comment will only take you 2 minutes, but it will be the strongest motivation for me to share something better.

 

Well I'm always reading! Thanks again for the thread. I am very interested in what you were saying. The only problem is that I couldn't find the RWI as an indicator in any of the charting websites I use. Also, I don't quite fully understand using the RWI yet.

 

So far, I have been more interested in the mean reversion strategy, but from what you have said so far, I understood that it would best be to use both depending on what kind of market in which you are trading.

 

So, the RWI is an indicator that shows whether a certain stock is trending. If the RWI is lower, then it would be best to use a mean reversion strategy? And if the RWI is higher, then a trend-following strategy would be recommended?

 

I don't typically like posting comments because I don't want to hijack the thread and I don't want to sound stupid, but if it encourages you, then I'll follow along and comment from time to time. Keep up the awesome work Do or Die! Thanks.

[Do Or Die 10]

So, the RWI is an indicator that shows whether a certain stock is trending. If the RWI is lower, then it would be best to use a mean reversion strategy? And if the RWI is higher, then a trend-following strategy would be recommended?

 

Yes, you can use RWI for selecting a particular stock or futures to trade.

 

Every stock within itself goes through different trading regimes.If you refer to the RS-internal thread, and regime shifting example using RSI, you can use almost any oscillator to shift regimes from and trade appropriately.

 

Random Walk Index of Highs (measures uptrend) RWHI= (H-Ln) / (R*sqrt(n))

Random Walk Index of Downtrends (measures downtrend) RWLI= (Hn-L) / (R*sqrt(n))

Here ‘Hn’ and Ln’ are Highs and lows n days back respectively and R is the average (daily) range.

 

also replied to your PM.

[phantom 16]

Hi D/D,

 

I noticed that your indicators for MSFT use settings of 9 and 40.

 

Any particular reason you chose these settings?

 

 

Thanks,

Phantom

[umfan92 10]

Thank you for the response.

 

Do you think you could post an example or two about how to use the RWI. I just don't really quite get it. I understand the theory, but I still don't know how to use it. I looked at the chart you posted of MSFT and couldn't see wha to do with it. When the RWI is low, it's supposed to be a sideways market, right? So mean reversion would be a better method. And the opposite goes for when the RWI is high. But in the chart, it seems to me that the RWI mimics the price chart.

[steve46 12]

Hi D/D,

 

I noticed that your indicators for MSFT use settings of 9 and 40.

 

Any particular reason you chose these settings?

 

 

Thanks,

Phantom

 

Its called "Optimization"

 

it doesn't work but thats what its called.

[Do Or Die 10]

Hi D/D,

 

I noticed that your indicators for MSFT use settings of 9 and 40.

 

Any particular reason you chose these settings?

 

 

Thanks,

Phantom

 

Hi Phantom,

 

No particular reason, as mentioned I wanted to plot:

1. Longer-term RWI for trend strength (on EOD most traders use the mid-MA as 40 or 50)

2. Short-term RWI for overbot/oversold levels (I would suggest a value 7-10 trading days because thats usually the period on EOD during which stocks get over extended or struggle in retracement).

 

If someone trades on Weekly charts, he may like to use 200 and 40 perios RWI. In general, the ratio between short-term and long term may not be greater than 1/3 because retracements arn't supposed to be mroe than 30% in a healthy trend.

 

thanks,

DD

[Do Or Die 10]

Its called "Optimization"

 

it doesn't work but thats what its called.

 

I'll point to some resources soon on what is optimization.

 

FYI, it's not optimization. Also, optimization is not just helpful, but essential for automated trading. What does not works is curve-fitting.

 

Cheers.

[steve46 12]

I'll point to some resources soon on what is optimization.

 

FYI, it's not optimization. Also, optimization is not just helpful, but essential for automated trading. What does not works is curve-fitting.

 

Cheers.

 

Of course you are welcome to your opinion. Mine differs....and as to the use of Random Walk, again my opinion is that it is close to useless...If at some point you start to discuss the use of hidden Markov models and GARCH I may look in again. Until then

 

Cheers

[Do Or Die 10]

Do you think you could post an example..

 

Sorry about the delayed reply. I have been on a family vacation for past few days; will post examples shortly.

 

Where can I find the RWI on a charting website?

 

Don't bother, I suggested RWI for regime analysis only. You can use RSI or many other methods for regime analysis. Once you get hold of a strategy software you can go ahead playing with custom indicators.

 

So if I was using a mean reversion method with the RSI and the RSI raised about 30 after becoming oversold, I would buy expecting the price to increase. Could I also do the same with the RWI?

 

Yes.

 

You can try on 60 min or EOD timeframe with one of the utility/transportation stocks. As a little exercise try paper trading with overbot/oversold levels when a stock is range bound or moving in a congestion area.

 

Consumer Goods and Technology are the most strongly trending, while Utilities and Transportation are least trending...

 

Cheers.

[Do Or Die 10]

Of course you are welcome to your opinion. Mine differs....and as to the use of Random Walk, again my opinion is that it is close to useless...

 

May I request you to elaborate on why you hold that opinion? forums are for discussions.

 

thanks.

[steve46 12]

Well it seems that we have had a short discussion. In answer to your question. If there is a problem it is with the implementation. The market I like to trade (S&P futures) seems to cycle in and out of stationarity. When it does that, its been my experience that the basic principle of RWI doesn't seem to work well. I think (I am not sure) that markets that are more stable (do not cycle as quickly) may be better candidates for that type of approach..Those markets might include commodities for example.

 

By the way, I believe a version of RWI is available on Investor/RT charting software. I think they call it "Random Walk Index" for those interested in pursuing the subject.

Edited August 2, 2011 by steve46

[Do Or Die 10]

Well, I think leaving one-liner opinions is similar to leaving tips where to buy/sell.

 

Some markets tend to show 'less random' cycle of stationarity, but they all cycle in and out of stationarity. In that sense, RWI will not be useful at all. I only mentioned RWI as a general trend strength indicator comparative to ADX. I also think that anyone who has done these kinds of studies will be using custom indicators.

[steve46 12]

Yes well now that you have had the chance to express yourself, what we end up with is just what I have said..that the RWI is essentially useless in markets that cycle through stationary and non-stationary movement.

 

and in the process you got in your little barb about one-liners...good for you... and now folks you know why generally I don't have patience with this kind of post.

 

Those who want to move a bit further down this road will need to take some advanced math. If you have that covered, the next thing to do is "Google" the following terms... "Markov" "Hidden Markov models" "Garch" "Tim Bollerslev" and "Estimators"

 

Its a long road and I am guessing few folks here will want to go there but thats where you have to end up if you want to find something that works (instead of RWI)...

 

Good luck people.

[Do Or Die 10]

Steve46,

 

Yes, I'm an arrogant prick full of overweening pride who is nowhere as smart as he thinks he is. And yes, I am using a slow old laptop running decade-old software on a thirty-year-old paradigm.

 

Now that I have got your brain working could you please get back to:

http://www.traderslaboratory.com/forums/psychology/10512-sin-predicting-anticipatory-trading.html#post125013

 

Until then, please stay in the confines of:

http://www.traderslaboratory.com/forums/technical-analysis/10597-early-warning-again-2.html#post126099