PureBytes Links
Trading Reference Links
|
It doesn't matter, Lionel.
While the relative strength comparative can be used to rank, the
relative values it ranks on have no absolute or objective value that
can be used in a formula or in any systems tester for MS testing.
It doesn't matter what formula is used to calculate it, the values are
only good for ranking. Ranking alone has little definitive predictive
value in determining which stocks are going to go up in the future,
only which stocks had the most momentum in the past.
One theory is the stocks with the most momentum in the past should
have the most momentum in the future. I haven't seen any tests results
that support that.
My tests results do show consistently that if a stock has a reason to
go up like strong fundamental or quantitative strength, then picking
the stocks from the QA and FA sub-group that have had the strongest
momentum during the previous 120 bars look back does a pretty good job
of predicting that they will perform well in the future, provided the
market trend is up.
Personally I haven't seen any method of manipulating the RSC numbers
so they can be used in formulas or in the tester (at least as all of
the MS testers are now written).
That's why I use the external relative strength like Investors
Business Daily but calculated using SpyGlass. Those values hold their
relationship based on a scale of 1 to 100 so they can used in the
tester and in formulas.
It's a matter of personal taste and purpose. When I tested a variety
of momentum formulas like the one I last posted, most anchored
momentum, the 2nd derivative of the price curve (both estimated and
calculated), NDI (which is a measure of the ADX), and the 3 week price
change along with some of the other methods mentioned in the other
posts, the external relative strength was the best, the 2nd derivative
of the price curve was next and then all of the others including the
formula I mentioned using the ROC came about the same.
I've also looked at the slope of the RSC, whether it is rising or
falling, the time periods any stock has been in top of the rankings
and several other variations on the RSC. Shorter look back periods had
more relevance than longer ones. Something between 13 weeks and 26
weeks almost always seemed to work the best.
That's my experience. However, whatever works for someone is good enough.
Super
--- In equismetastock@xxxxxxxxxxxxxxx, "Lionel Issen" <lissen@xxx> wrote:
>
> Super:
>
> Couldn't we use a moving average of the relative strength to smooth out
> these variations?
>
>
>
> Lionel
>
>
>
> From: equismetastock@xxxxxxxxxxxxxxx
[mailto:equismetastock@xxxxxxxxxxxxxxx]
> On Behalf Of superfragalist
> Sent: Saturday, August 25, 2007 1:21 PM
> To: equismetastock@xxxxxxxxxxxxxxx
> Subject: [EquisMetaStock Group] relative strength comparative
>
>
>
> The problem with relative strength in general, regardless of what
> someone uses to calculate it, is the measure of strength it gives is a
> relative number only and does not mean the same thing from day to day.
>
> For example, if the relative strength of 100 stocks is calculated and
> sorted, today the top strength rating for the strongest stock might be
> 58 and tomorrow it might be 38. While it is true that the sorting does
> rank the stocks by their relative strength, the problem is numerical
> value associated with the strength is meaningless and can't be used in
> a tester to determine the effectives of relative strength as a filter
> for a particular trading system. At least not with the tester
> available for MS.
>
> In addition trading rules can't be written that define an entry with a
> relative strength rating above X because X is a relative number.
>
> I use the external relative strength which compares a specific stock
> to every other stock in the pool of stocks being looked at. That pool
> can be the entire market or 100 stocks, etc. If a stock is out
> performing 98% of all other stocks last week and 98% of all stocks
> this week then those two percentages mean exactly the same thing.
>
> This is relevant because various performance thresholds can be tested
> in the tester across varying look back periods to find the external
> relative strength filter that is the most effective with any
> particular trading system.
>
> If someone wants to compare the relative strength of stocks to an
> index like the S&P500, simply include the SPX index or SPY ETF in the
> pool of stocks. All of the stocks that have higher strength than the
> SPY are performing better and all the stocks that have a lower
> external relative strength numerical ranking are performing worse.
> This too can be tested or programmed into a set of trading rules. In
> other words, someone could write an equation to ignore all trades
> signals on stocks that were performing worse than the index.
>
> I use the old SpyGlass program to calculate the relative strength.
> It's now called Fire. The external relative strength calculations that
> SpyGlass does are nearly identical to the ones IBD does. IBD uses a
> larger pool of stocks than I do, but I have cross compared the lists
> and the ranking differences weren't statistically significant.
>
> I use the old SpyGlass because it's really the only method I've found
> that has the appropriate dll's to do the calculations consistently and
> correctly. I've heard someone else was writing a dll for this purpose
> but so far I haven't seen one. The external relative strength
> calculations in SpyGlass are about as easy as it gets.
>
> In my various systems tests I have found that using the external
> relative strength as a filter improves the performance of almost all
> of the trading systems I tested it in.
>
> However, overall the filters that I've found that have improved
> performance the most are a market conditions filter that keeps the
> system from trading against the trend, a prescreening of trade
> candidates with fundamental and quantitative constraints like those
> used by Value Line, IBD, Ford Equity Research and S&P platinum and
> neural fair value screens, some sort of simple volume filter, and
> finally the use of external relative strength.
>
> Those four things had the most impact on the hundreds of trading
> systems I've tested them in. Based on the results, I've written about
> a simple trading system in Roy's newsletter with the test results from
> those items on one system so the impact is objectively measured.
>
> Here is a simple relative strength equation that has a look back
> period of 120 bars and is weighted 40% more for the last 30 bars, etc.
> When the stocks are sorted according to the numerical values coming
> out of this equation, the sorted list compares very closely with other
> software that calculates relative strength. Again the problem is the
> numerical values have no comparative values except the relative
> comparison for that day. You can use threshold values in tests or
> equations as I explained before.
>
> ROC(C,30,%)*0.4+ROC(C,40,%)*0.2+ROC(C,60,%)*0.2+ROC(C,120,%)*0.2
>
> Based upon my trials and tests with this equation, if I just wanted to
> be able to sort a group of stocks from the highest to lowest
> performers over some look back periods, I would just use this equation
> because it gives very similar results to all the other methods of
> calculating relative strength that I could find and I bought or
> programmed every program I could find that did those calculations.
>
> Have fun! Ain't volatility wonderful...
>
> Super
>
>
>
>
>
> [Non-text portions of this message have been removed]
>
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/equismetastock/
<*> Your email settings:
Individual Email | Traditional
<*> To change settings online go to:
http://groups.yahoo.com/group/equismetastock/join
(Yahoo! ID required)
<*> To change settings via email:
mailto:equismetastock-digest@xxxxxxxxxxxxxxx
mailto:equismetastock-fullfeatured@xxxxxxxxxxxxxxx
<*> To unsubscribe from this group, send an email to:
equismetastock-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|