|
PureBytes Links
Trading Reference Links
|
John,
Below is a first pass at the Linear Regression Slope code for the Divergence
indicator:
LC30:=LinRegSlope(C,30);
LI30:=LinRegSlope(RSI(9),30);
AA30:=(LC30>0 AND LI30<0);
BB30:=(LC30-LI30)-Ref((LC30-LI30),-1)>0;
LC60:=LinRegSlope(C,60);
LI60:=LinRegSlope(RSI(9),60);
AA60:=(LC60>0 AND LI60<0);
BB60:=(LC60-LI60)-Ref((LC60-LI60),-1)>0;
AA30 AND AA60 AND Sum(BB60,3)=3 AND Sum(BB30,3)=3
I've limited this to looking for negative divergences on rising closing prices.
For the other three possibilities, the changes would be "cut and paste" from the
above with some > or < comparsion changes.
In the code above, I'm looking for positive price slope with negative RSI(9)
slope. I'm also requiring that the difference between the two has been
increasing for the last three periods. I do this for both the 60 period slopes
and the thirty period slopes. Then I'm requiring the test to be true for both
cases. The result of this formula will be a 1 when a divergence exists.
Some arbitrary values used above that need to be experimented with or changed
based on preference. "3" periods of widening divergence, "60" and "30" periods
for slope comparison, RSI using "9" periods.
I've 'eyeballed' this on a daily chart or two, and it seems to call out
divergences. The next step would be to take the HHV of the last x periods and
compare it with the same for 30 or 40 periods before. Select only those with a
comparative difference of some chosen percentage. (e.g. The HHV of the close
is within 3% of the prior HHV 30 periods ago, but the HHV of the RSI(9) is more
than 10% different from its HHV 30 periods ago).
That filter should help limit some of the above divergence signals and make it
more reliable.
Dave Nadeau
Fort Collins, CO
John Adair wrote:
> Hi Dave
> I like your suggestion for llinear regression to pick up divergence. Yes
> lets you try to code it for us.
>
> -----Original Message-----
> From: owner-metastock@xxxxxxxxxxxxx
> [mailto:owner-metastock@xxxxxxxxxxxxx]On Behalf Of Dave Nadeau
> Sent: Saturday, December 02, 2000 5:36 PM
> To: metastock@xxxxxxxxxxxxx
> Subject: Re: Divergence
>
> John,
>
> I haven't coded this and run it out yet, but here's what I'd try:
>
> I would combine a comparison of Linear Regression Slope with the HHV (or
> LLV's)
> for different periods of time. Using 1 to 70 periods as you suggest, I'd
> probably first screen for those tickers which have made new highs (or lows)
> in
> the last 20 periods, or last 20-40 periods. Then I'd compare the
> LinRegSlope of
> the prices with the LinRegSlope of the RSI. I'd then subtract the two and
> eliminate any difference less than some threshold (say 0.5 or
> 1.0--experiment
> with the results to find the right amount). I'd probably take it a step
> further
> and compare the LinRegSlope of 70 periods with say 30 periods then 10
> periods of
> each series, just to narrow down the pattern.
>
> Once I've established a divergence in slope, I'd then compare the recent
> HHV(of
> say the last 20 periods) with the HHV of the 20-40 bars ago period with the
> 40-70 bars ago period. For a negative divergence, I'd want HHV1, HHV2, and
> HHV3
> of the prices to be constant or increasing, while those of the indicator
> should
> be decreasing. ...and just the opposite for a positive divergence.
>
> If anyone on the list is interested in the actual coding of this, we can
> keep
> the thread alive and post some Metastock Formula Language. Or try some
> other
> approaches.
>
> Dave Nadeau
> Fort Collins, CO
>
> jhmtn wrote:
>
> > One of the more interesting indicators of price reversal for me is the
> > divergence that occurs between a momentum indicator and the price action.
> > For example, if through some time period both price and RSI make a high
> and
> > then price rises and RSI drops, then divergence has occured between the
> > indicator and the price and a reversal in price action "might" occur.
> >
> > I have not been able to figure out a manner of looping thru the past 1 to
> 70
> > periods of price and momentum indicator values to figure out if the prices
> > are showing divergences or not using Metastock.
> >
> > Does anyone have such a formula or know where i can find it?
> >
> > Thanks. ................................... John
|