Registered User
Joined: 12/21/2004
Posts: 902
|
Back in August, Diceman created a new custom indicator that "counts" how many times in a particular period a particular condition is met. I have found this indicator to be extremely interesting and versatile, and since it never got a thread of it's own, I would like to devote a thread to it, and challenge the other participants in this forum to contribute their own variations on the indicator, as I challenged Apsll later on in the thread. The original thread can be located here:
Sir Fan-Pattern
The custom indicator takes the form:
ABS(PCF condition)*n
(smoothing average set to n, simple)
where n = number of days in the time period under consideration where the PCF condition (i.e. price is greater than MA or vol>100,000, or MACD-H>0, Sto>80, etc.) is true.
I think this particular sort of counting indicator can be very interesting in a turmoiled market such as we've had the past 2 months in identifying stocks that have demonstrated a stabilty (or instability) of sorts.
We haven't had Tobydad posting in a few months, but when he was posting, he was extremely generous in sharing a style that made very successful use of price and volume indicators (which I dubbed the "Tobydad profile"). Tobydad used a TSV26 - do a search of "TSV26" and read the several posts from last spring where Tobydad very kindly described and shared his indicators and entry and exit strategies - it makes for very interesting reading.
Anyway, in the beginning of April, Tobydad mentioned the following:
QUOTE (tobydad) I am presently in HTI....
My primary indicators above all others are a 20day bollinger band on price, a 30-day linear regression on price, and exponential TSV26 with a 13-day bollinger band on it. Of course volume and upper and lower trend lines help clarify any picture.
On 3/23 I noticed that the TSV 26 and its BB had been above the 0 line for over 90 days (a show of strength in a so-so market...not to mention stocks seem to have memory, they like going back where they've been).
Now I would describe the current market as "so-so", and the Diceman Counting Indicator lends itself to readily identifying stocks where the TSV 26 has been over the 0 line for a long period. Since TSV over 0 indicates strength, I would say that any stocks which have weathered the past 2 messy months with positive TSV values should be pretty good picks.
I prefer to use a daily zoom=6 (65 days), which will bring us back to mid-June when things were still rosy.
The formula for exponential TSV26 for 65 bars:
ABS(XAVG(TSV,26)>0)*65
smoothing average 65 simple
I put it in as a custom indicator that isn't visible
Sort your watchlist by "actual value" and the "Sort value" is the number of days that the condition is true (i.e, in the last 65 bars, the number of days that the TSV26 is over 0)
In the S&P 500, I came up with:
JNPR (65)
BMET (65)
BIIB (64)
NVDA (64)
CSCO (64)
Doing some other sorts that I use, here are some others that are all 65 for 65:
RNO
CDEC
CPHD
FSTR
DRYS
NVT
BOOM
OI
SCUR
KTEC
GMST
I would use this not necessarily as a primary indicator, but in conjunction with searching for something like high OBV or high MACD-H, or in evaluating stocks that meet a particular entry criteria. Not every great stock has to be a "65" - the point is to find stocks that maintained their strength through at least the past 2 months - the 65 just matches my personal use of zoom-6.
We all have different indicators that we like and different styles that we use that incorporate a variety of time periods and combintions.
As the Diceman Counting Indicator is so very useful and versatile, I would like to challenge forum participants to come up with and share their own other ideas of ways to use this indicator.
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
Actually that last link wasn't to the thread where Diceman made up the indicator - try this one:
Sir Fan-Pattern
|
|
Registered User
Joined: 3/21/2006
Posts: 4,308
|
Hohandy, How would you wright this type of formula for the amount of days Money Stream is above its 22 day exponential moving average?
The only thing that I am not crazy about with this type of indicator is that the indicator, lets say TSV can be 40 days above the zero line out of 65 (and the formula indicator will register the value of 40), but TSV can also be currently in a two or three day nose dive due to heavy selling even though the value was still 40. Do you get what I mean. When you are sorting a large watchlist you will think that 40 days over zero out of the last 65 days is good but that will not give you current information as I stated above.
What is your thoughts about that?
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
Apsll - that formula would be
ABS(MS > XAVG(MS,22))*n
(smoothing average set to n, simple)
I agree with you that the indicator won't give you any information about the current situation (unless, of course the condition is satisfied throughout the entire time period in question) - which is why 1) so far I've only played with indicator at the extremes (either very high or very low incidences of the formula condition being met) and 2) I think the best uses of the indicator in conjunction with other indicators. Say a primary sort would be all stocks with MS moving above it's XAVG today, and then using this indicator on that list.
|
|
Registered User
Joined: 3/21/2006
Posts: 4,308
|
Thanks Hohandy.. I am using the indicator to see how many days out of 20 that the Money Stream was above its moving average. I think that using a lower setting like 15 or 20 days is better than say a 60 or 100. There is less chance of the value representing the the back-side of the indicator. If the value maxes out at more than 20 days then the indicator will just stay horizontal. Knowing this, then when you are doing a large sort although a value less than 20 could still be on the back-side (or down-side) of the indicator, it is less probable then if you used a higher number for the *N factor..
Thanks for your help and sharing. I will be playing with this for a while, (Hold all my calls).
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
Apsll - that quote of Tobydad's about the TSV being above zero for over 90 days as a sign of strength has stuck in my head ever since he posted it (and it was quite interesting going back through all the posts last night until I found it) - so that's why I particularly went with the TSV being above 0 for the longer period (I also have had good success with looking for stocks with MACD above 0 and price above 13 and 26emas for long periods with various PCFs before Diceman created this indicator).
But we all have different styles and methods and I don't think any particular one, or way of using this indicator, is more valid than any other. Which is why I wanted to create this thread and challenge people to come up with different and various ways to use it - we have some very creative people contributing to this forum, who, if they participate, should come up with some pretty varied and interesting ideas.
|
|
Registered User
Joined: 1/1/2005
Posts: 2,645
|
Suppose we have a Boolean condition A. For example, A may be (C>AVGC50). The following two counts may be of interest:
1) The number of days out of the last N days the condition A has been satisfied.
2) The number of consecutive days including today the condition A has been satisfied.
The approach of diceman using a Custom Indicator or the use of a Percent True Indicator are equally adept at finding 1). The results of 1) are primarily used for determining the probability that condition A occurs.
We can write a PCF to yield the count in 2). The results of 2) are primarily used in Trading Strategies.
There is a case where the result of 1) can be used to satisfy a requirement on the result of 2). Suppose we require that the count from 2) be greater than or equal to N. Then, if the count in 1) is N, the requirement is satisfied.
In the examples given so far, it appears to me that one would be more interested in the count from 2) than the count from 1).
Thanks,
Jim Murphy
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
bustermu
1) Would depend on how long-term
the trend you were looking for
was.
(if we are looking for multi-year
growth, for example)
2) Can be satisfied by limiting your
search to the length in question.
(since the equations are so easy to
edit)
Thanks
diceman
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
Here's an application that Apsll may find useful in his Bonanza Bottom setup and that also addresses Jim's point: Not every Trading Strategy necessarily has to have a condition be consecutive or in effect "today"
Consider the Bonanza Bottom - a bottom formation with (one to three day) volume anomalies such as high BOP or TSV>0 accompanied by high volume spikes which give confirmations that buying has been going on.
So how about, taking something that will measure a down trend such as a 100LR, applying it to a watchlist and generating another watchlist with long term downward sloping 100LR. Then, using the Diceman Counting Indicator and identifying those on the list with most instances of high BOP (say BOP>60) or number of days with TSV above, or better yet, for finding spikes, crossing the zero line.
So how about, for example,
RDN
TMA
CFC
LEN
BSC
|
|
Registered User
Joined: 3/21/2006
Posts: 4,308
|
Hohandy, I do not know if it was your intention, but the stocks that you have here are all residential construction or mortgage investment firms. They all took a hit at the same time because of the prime lending fiasco.
Over-all I like your approach, I have seen before when a stock will have high BOP and poitive TSV while in a down trend, this scenario usually turns out pretty well. I never thought about scanning for these conditions before. I hesitate however to associate this formation with a Bonanza Bottom. If you think about it, a down trend is not a bottom. But I am on board with designing indicators to look for this type of formation.
Since you came up with the idea then you get to name it.
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
One way to use the combination of long and short-term:
A) Use short-term to demand a condition be true.
B) Use long-term to develop a batting average
for a condition.
-------------------------------------------------------------------
Lets say we want to do a count of stocks where
C>C4 (the close is greater than 4 days ago)
To avoids stocks that have fallen through the floor.
We could count:
L>AVGL21
(the low is above its 21 day moving average)
we could use the past month:
ABS(L>AVGL21)*21
(smoothing average set to 21 simple)
There are 71 stocks in the SP-500 that
equal 21. (flag those)
Now sort:
(C>C4 for the past year):
ABS(C>C4)*250
(smoothing average set to 250 simple)
------------------------------------------------------
Out of the 71 stocks that were flagged.
NOV is number 1 with 171 times.
(171/250)= 68.4% of the time.
MOLX is the worst with 111 times.
(111/250)= 44.4% of the time.
(obviously this type of thing could be used
to find strength or stocks that were weak
and have just gotten strong)
Realize that I'm not recommending these
scans. It is simply to illustrate the concept
of demanding something be true in the
short-term. Then developing a batting average
for the long-term. (from those stocks)
Thanks
diceman
|
|
Registered User
Joined: 1/1/2005
Posts: 2,645
|
diceman,
The above results could perhaps be accomplished a little more directly using a Custom Percent True Indicator instead of a Custom Indicator.
It should be noted that a Custom Percent True Indicator may be convient, but it is superfluous in the sense that nothing can be accomplished by its use that cannot be accomplished by the use of a Custom Indicator.
Thanks,
Jim Murphy
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
Here's another interesting application of the formula:
Say I'm testing for momentum and one of my indicators for "momentum" is number of up days with heavy volume (i.e., I use volume greater than 1.5x45emav)
I can set this for a longer period such as the 65 days (3 months) or zoom=6, or look for recent short-term, i.e., the last 10 days.
ABS(v>1.5xavgc45 and (c>c1))*10
(smoothing average =10 simple)
Some interesting results:
JRJC 7
AXTI 6
MHJ 6
INXI 5
MG342 (Farm Products) 5
SVA 4
SPIR 4
FXC 4
SLP 4
SNDA 4
RDWR 4
GTU 4
|
|
Registered User
Joined: 12/21/2004
Posts: 902
|
And some more...
One thing that we sometimes look for is a turnaround day on a new low on heavy volume. The following uses parameters
(heavy volume) 1.5x45xmav
(touching a low) L<MINc100 (in this case I'm picking a 100 day low)
(and rebounding intraday) (C - L) / (H - L)>.50 (closing in the upper half of the candle)
Setting this to search for number of times during the past 6 months:
ABS ((C - L)/(H - L)>.50 and (v>1.5*xavgv45) and (L<MINc100))*130
This can be further tweaked - let's look for stocks that have rebounded off of the 50ema - on heavy volume - either the low or the close touched within .005 of the 50ema, and it finished in the top half of the candle on heavy volume, number of times in the past 6 months:
ABS ((C - L)/(H - L)>.50 and (v>1.5*xavgv45) and ((L<1.005*avgc50 and L>.995*avgc50) or (C>1.005*xavgc50 and C<.995*xavgc50)))*130
Of course any of the parameters can be tweaked to individual preferences, but these are examples of how this indicator can be adopted to test for popular measures of support.
|
|
Registered User
Joined: 1/28/2005
Posts: 6,049
|
"It should be noted that a Custom Percent True Indicator may be convient, but it is superfluous in the sense that nothing can be accomplished by its use that cannot be accomplished by the use of a Custom Indicator."
------------------------------------------------------
bustermu
I look at it the other way. I have always favored the
custom indicator. It can do what the percent true can but it can also do what the percent true cant.
Since it is not boolean it can be used
for analog measures.
We can also(with ABS) in effect create
a digital analog.
(analog that we can turn on and off)
This lends itself to more flexibility and manipulation.
The poor percent true can only perform its
intended function.
Thanks
diceman
|
|
|
Guest-1 |