A New Way to Look at RSI(2) and ATR(10)

This was the lead paragraph of today’s economic blockbuster - Home building posted a small gain in April, but permits for future construction plunged by the largest amount in 17 years, a sign the nation's housing industry is still in a deep slump.

When is the best time to buy something? I’ll give you a hint “buy low - sell high.” That’s right – when it is hitting record lows and it is something that you are pretty sure will recover, eventually.

Today’s filter is a new one and once again is a very simple one –

show stocks where close is between 15 and 35
and average volume(90) > 500000
and rsi(2) reached a new 4 week low
and atr(10) reached a new 4 week low

You really can’t get much simpler than that can you?

It tests very nice too both in the short term (4 day swing) and in the mid term (30 day trades).

Short term – 69% win rate, 2.30 reward/risk ratio and a 95% ROI that’s compared to an SPX ROI of 1.54% for the same period. In other words, the market was depressed when this filter was tested.

The net change over 20 and 30 days was 3.56 and 3.63 respectively so this filter has some legs as they say.

The interesting thing about this filter is that as I go back through the previous day’s selections it shows a significant number of wins. Sometimes it takes a couple of days, sometimes a few weeks, but eventually the stocks selected seem to turn up. The best selections however are those that have the highest volume and that are beginning to ascend on the day after the filter finds them.

Here are a couple of examples.


and this one


Now why do you think that this filter is so appropriate to today’s blockbuster? Because this report means that the crisis is nearly over and any small increase in housing permits next month will make the housing stocks go boom. Bank on it.

Another 2-Hour Chart

Here's yet another example of what you can do with 2-hour charts - Again it is AKS but what I wanted to show you was a simple three-pronged set-up.


What you see here is an RSI(2) < 2 coupled with a tweezer bottom (the black diamond below the third candle informs that formation) coupled with an S1 pivot point.

Note that a purchase on the tweezer bottom candle - say at the cross over on the S1 at 29.88 and this morning it hit 31.60. That would have been an excellent return for a couple of days of swing trading.

RSI Selections

Well I spent some time looking at RSI(2)<5 selections over the last month or so and of the top 5 each day 3 were winners and 2 were losers on average - But that begs the issue of how did they do when the market was down and in the last 18 trading days the market was down twice and both times the filter outputs did very well.

I did find that if the filter picked a stock below the lower donchian band it did better than one that was coming off the top band. This would probably work the same with bollinger bands too - so maybe a two parter is a better way to go.

At any rate the market is down as of Friday, is probably going to go up today and my top 5 selections are - WIT, LYV, EXR, ANW, and MHO.

I'm going to put these on the watch list and if they start stirring I might take some day trades in them.

Not an invitation to speculate in the stock market - just an observation.

SPX RSI(2) < 5 -

Red Hue mentioned another bit of research on Trading Markets regarding the best time to buy stocks being from the 23rd of the month to the 1st.

I agree with this. I've done similar research and discovered that market, in very large generalities, is lower towards the end of the month and higher in the first couple of weeks. It has to do with supply and demand - demand is higher in the beginning of the month because of the retirement fund money (billions) pouring in and option manipulation money (that's a joke for Jimmy Crack Corn Pone) and when these two things run out the market quiets down, generally, and stocks, in the absence of demand, fall.

I don't think you could design a viable system around time of the month (Ms Market has PMS - run! run!) (that ought to fix me right up with my female fan) (George Clooney has female fans - I've got fan). The reason being too many moving parts - too many economic announcements and earnings announcements and just the world situation in general.

But the idea did get me thinking. What if you designed a method that relied on the market being below 5 on the RSI(2)? So you would only buy when the market, as represented by the SPX or other derivative, was down and you only sold (short) when the market was up.

It makes more sense to go with the flow of the market than go against it. I'm going to try to program that into my filter model and see what happens. Dogwood - if you are reading this might be something more easily done in wealth-lab so maybe you could give it a try.

Williams %R

There are many ways to skin the technical analysis cat and two of them are the Relative Strength Index (RSI) and the Williams %R.

First the RSI. TAZ Trader reports on a six rule method from Trading Markets that is an "improved RSI(2)" method. Because I am always interested in any method that can be programmed into my modeler I adapted their method to my own. Their method exclusively followed the SPX or its derivatives while mine selected individual stocks. Now that isn't necessarily the same but for all practical purposes I don't buy the market so if it doesn't work on individual stocks I don't want to fool with it.

TAZ Trader then goes on to talk about his method that uses Williams %R. Apparently he prefers a three period Williams %R to the RSI(2).

So I wrote a couple of filters and here are my results -

The Trading Markets method output a 65% win ratio with a 105% ROI - this isn't bad.
The Williams %R(3) method output a 62% win ratio with a 150% ROI - this is just fine.
The RSI(2) method output a 72% win ratio with 162% ROI - that is excellent.

Here is the RSI(2) method -

show stocks where close is between 15 and 35
and average volume(90) < 500000
and rsi(2) < 5
and volume 1 day ago > volume 2 days ago
and volume > volume 1 day ago

Now I tried including a volume statement in both the Williams and the Trading Markets methods (although it isn't called for) and it only lessened their efficiency and effectiveness.

But here is an interesting aside - even though these are supposed to be extremely short term filters - if you hold the stocks you select using them you will hit some home runs in the next 20 days. The Williams %R filter had a whopping 6.58% net change over 20 days and the RSI(2) turned in a respectable 4.31. The Trading Markets method was a more pedestrian 2.5 not bad but nothing to get excited about.

I'm going to stay with the RSI(2) < 5 method as it seems to be better on the individual stock basis. But there is nothing wrong with using either the Williams %R or the Trading Markets methods if that's what you want to do.

RSI Revisited

A number of trading posts talk about the RSI(2) as a means to quick profit and for all practical purposes I believe it to be so. I did a post some time back that got good results from an RSI(8) but I think that if you adapt to the techniques that other traders are using you are probably better off in the long run. This is because you can exploit their weaknesses - and that is what you want to do.

This morning I read an article published by Trading Markets and reprinted in Yahoo Finance regarding several ways that they select stocks for trading.

One of these was using the RSI(2) and I quote - 2-Period RSI Below 2: These are stocks that have a 2-period RSI reading below 2 and are trading above their 200-day moving average. Our research shows that stocks trading above their 200-day moving with a 2-period RSI reading below 2 have shown positive returns, on average, 1-day, 2-days and 1-week later. Historically, these stocks have provided traders with a significant edge.

Needless to say in my never ending quest to rid the world of Simple Moving Averages this one caught my eye. I said - well - I have to test this theory out - so I wrote a filter.

It's a simple filter as filters go -

show stocks where close is between 15 and 85
and average volume(90) < 500000
and rsi(2) < 2


And I back tested it. I got some pretty nice results - 68% win percentage and 129.73 ROI. This is not too shabby.

Then I modified the filter to add a line - and close > ma(200) and back tested that. My results were 64% win percentage and 51% ROI. Wow - talk about a fall off in profit potential.

I modified it once more to change the line to close > ema(90) - the results were even worse - 55% win percentage, 44% ROI.

One more modification to close < ma(200). This time the results were almost the same as they were with no moving averages involved at all - 68 and 131 (a little improvement in the ROI).

A final modification just to close the loop to close < ema(90) and that improved the ema results to 66% and 127% ROI.

Bottom line - the moving average doesn't matter - RSI(2) < 2 with a play on the highest volume output that day is a good short term winner without considering a moving average. But if you are going to use a moving average of any kind - go low not high - in other words - below is better than above. Which just serves to reinforce our common refrain - distressed stocks do better than high flyer's in the short term.