StockFetcher Forums · Filter Exchange · Help with a Reversal Filter Please !<< 1 2 3 >>Post Follow-up
2 posts
msg #32428
Ignore rgl123
6/24/2004 11:32:13 PM

"Market Reversals and How to Spot Them" by Matt Blackman maybe this will help better in understanding "sushi roll".

2 posts
msg #32429
Ignore rgl123
6/24/2004 11:54:14 PM

I see that the StockFetcher doesn't allow a person to go to the link as it is done on Yahoo message board.

You can find the article by going to "Google" type in "sushi roll" and scroll down to the last entry on the page. "Sushi Roll". Clicking on it sends you to "Investopedia. com" and the article.

2,025 posts
msg #32430
Ignore alf44
6/25/2004 12:08:13 AM


Believe me, I understand the "sushi roll" !

I've read that article and other related articles many times. The concept is simple and straight forward !

What isn't as simple (at least for me so far) is writing a scan that exactly defines the components of the "sushi roll" set-up ! I can get it pretty close but so's not quite there. That also seems to be the case for the other attempts that have been posted on this thread. cegis has some great ideas but so far they have not nailed the set-up.

I'm still working on it and welcome more input from those reading this thread !


235 posts
msg #32432
Ignore cegis
6/25/2004 10:45:18 AM


I'm new to the sushi roll, but got my definition from investopedia, which you point to. Although SF doesn't make URLs "hot", you can always cut and paste them into your browser's address field (as I did to review the description).


Glad my filter is pointing in the right direction. What part of the set-up do you think is still missing? From the article referred to by rgl123, the two points missing from what I suggested are the "narrow range" of the inside box (which I mentioned, but didn't attempt to define), and the possible use of RSI divergance as a confirmation. Let me try to address them both in this attempt:

set{in5hi,high 5 day high 5 days ago}
set{in5low,low 5 day low 5 days ago}
set{out5hi,high 5 day high}
set{out5low,low 5 day low}
set{in5rng,in5hi - in5low}
set{out5rng,out5hi - out5low}
set{in5pct,in5rng / price}
set{out5pct,out5rng / price}
set{ratio,out5rng / in5rng}

set{rsidown,count(10 day slope of rsi(14) is less than 0,1)}
set{rsiup,count(10 day slope of rsi(14) is greater than 0,1)}
set{trendup,count(slope of support(15,10,2) is greater than 0,1)}
set{trenddown,count(slope of resistance(15,10,2) is less than 0,1)}
set{selldiverge,trendup * rsidown}
set{buydiverge,trenddown * rsiup}
set{divergance,buydiverge + selldiverge}

out5hi is greater than in5hi
and out5low is less than in5low

and ratio is greater than 1.2

and divergance is equal to 1

and price is greater than 1
and volume 25 day low is greater than 25000

draw support(15,10,2)
draw resistance(15,10,2)
draw rsi(14)
do not draw out5hi
do not draw in5hi
do not draw out5low
do not draw in5low
do not draw divergance

add column in5rng
add column out5rng
add column ratio
add column in5pct
add column out5pct
add column rsiup
add column trenddown
add column buydiverge
add column rsidown
add column trendup
add column selldiverge
add column divergance

The first group of set{} commands get the needed high and low values, and calculate the range and range percentages for the inside and outside boxes.

The second group of set{} commands determines if there's a divergance between RSI and the trend (defined as the slope of the support or resistance line). This is done by checking the slope of the rsi, and the slope of the trend, then checking to see if the slopes are opposite in value, indicating the divergance. This uses the technique that employs the count() function to give a "logical" value of 1 (true) or 0 (false), based on if the condition is true over the past 1 day. Multiplication is then used to perform the AND function, and addition is used to preform the OR function. (The "only" problem that I see with this definition of divergance is that the "10 day slope of rsi(14)" does not necessarilly use the peaks (or valleys) to "draw the line" of which the slope is determined. It would be much better if I could use something like "rsi(14) resistance(15,10,2)" [i.e., the "resistance" function, applied to rsi(14)].) As an aside, note that I used resistance(15,10,2) for the down trend, and support(15,10,2) for the uptrend. The ",2" at the end of the parameter list allows the trend to be broken within the last 2 days, which is a form of confirmation...

The first two filtering conditions select stocks with the inside box followed by the outside box.

The next condition is the "narrow range" attempt. In this case, I'm suggesting the outside range is at least 20% bigger than the inside range.

The last condition requires the divergance. It may be useful to remove this condition, but keep it's calculation (to show in the columns) so that the divergance is not required (perhaps another confirming signal would apply).

I added a price and volume conditions, which you can change (or remove) as suits your needs. (I don't like trading thinly traded penny stocks.)

The "add column" statements are primarilly for illustration purposes.

So, how's it look? If you think something is still missing, let me know (be specific! Perhaps a specific stock as an example?), and I'll see what I can do... (I don't think this is perfect, but with it's pretty high hit rate [about 85 for last night], it may well give several stocks that fit the set-up quite well.)



2,025 posts
msg #32455
Ignore alf44
6/28/2004 12:52:11 PM

Sushi Roll anyone ?

First a few thoughts:

In my original attempts I was trying to define 10 days of action. The first five days (days 10 to 6) would have a Range that would be inside (or engulfed) by the Range of the last (most recent) five days (days 5 to 1 being the current day). Further, the Low on "day 5" would be the lowest Low of the last 10 days and the High on "day 1" would be the highest High of the last 10 days. Sounds simple enough.

One of the first problems I had was isolating "the lowest Low of the last 10 days 5 days ago". As some here on this thread said...the way I had written it in my scan would just ensure that the Low 5 days ago was the Lowest Low for the 10 days leading up to the Low 5 days ago. That was correct. I was wrong.

The next problem was in my thinking about just what "5 days ago" really meant. When you think about it..."Day 5" of the last 10 NOT the same as "5 days ago" in terms of writing a StockFetcher scan. Duh ! It seems obvious now...and it should have been obvious...but, to zero in on "day 5" you have to write the scan to read "4 days ago" because the current day is "day 1." Again...DUH !

So...with that in mind...

the Sushi Roll setup (for a Long Reversal) would be:

Fetcher[Show all stocks where High reached a new 10 day High and Low 4 days ago is below the Low and Low 4 days ago is below the Low 1 day ago and Low 4 days ago is below the Low 2 days ago and Low 4 days ago is below the Low 3 days ago and Low 4 days ago is below the Low 5 days ago and Low 4 days ago is below the Low 6 days ago and Low 4 days ago is below the Low 7 days ago and Low 4 days ago is below the Low 8 days ago and Low 4 days ago is below the Low 9 days ago and Average Volume(10) is above 100000]


the Sushi Roll setup (for a Short Reversal) would be:

Fetcher[Show all stocks where Low reached a new 10 day Low and High 4 days ago is above the High and High 4 days ago is above the High 1 day ago and High 4 days ago is above the High 2 days ago and High 4 days ago is above the High 3 days ago and High 4 days ago is above the High 5 days ago and High 4 days ago is above the High 6 days ago and High 4 days ago is above the High 7 days ago and High 4 days ago is above the High 8 days ago and High 4 days ago is above the High 9 days ago and Average Volume(10) is above 100000 and Price is above 7]


I'm not making ANY claims as to the profitability of these they appear here. They're not perfect ! They do however incorporate all/most of what Mark Fisher defined a Sushi Roll to be. I will be adding some criteria to try and further improve them. I think there IS something here worth building on. I would welcome suggestions as well.

Thanks to cegis and everyone for all your help !


2,025 posts
msg #32467
Ignore alf44
6/28/2004 10:39:53 PM

Thanks to...

...cegis, xraywiz, patefern & rgl123.

Further thoughts from ya'll or others ???

alf44 :8^)

235 posts
msg #32471
Ignore cegis
6/29/2004 9:55:39 AM


Glad I could help.

And, since you asked...

Another way to check that the low (or high, depending on long/short) 4 days ago is the lowest would be the phrase "low 10 day low is equal to low 4 days ago". This, however, does allow any of the other 9 lows to equal the 10 day low. I don't know if this is an issue with "sushi rolls"...



7 posts
msg #32502
Ignore patefern
7/4/2004 12:30:26 PM

PTEN is on a watch list I have and it looks like the beginning of a sushi roll, as I understand it. This stock did not come up in any filter and it is the pattern I look for which is a bit fustrating. Can you or cegis or anyone write a screen for this pattern. Thanks in advance

235 posts
msg #32505
Ignore cegis
7/5/2004 9:31:41 AM


Disclaimer: I'm no expert on sushi rolls, so take this with a grain of salt...

From what I can tell of PTEN's 7/2 EOD chart, it does not have a sushi roll pattern at present (days 10-6 not engulfed by days 5-1 [6/22 low of 16.01]; low 5 days ago not lowest low; other?). So, it's not surprising that it doesn't show up on any of the filters in this thread. But I suppose it COULD be "the beginning of a sushi roll." Obviously, that'd depend on how the next few days play out.

So, I'd have to ask, exactly, what pattern is it that you seek? If you can describe it, you are more than likely able to write a filter for it. (Although, there are filter limitations that make some things not "doable".) I'd be more than happy to help, too. Besides describing the pattern in detail, I will ask that you also post what you have tried, and why tou think it is not working.

I'd also suggest starting a new thread, unless your pattern is in fact related to sushi rolls.



22 posts
msg #32515
Ignore xraywiz
7/6/2004 8:42:37 AM

A couple of comments:
In your last two filters for long and short reversals, you are right in setting up the filter according to the literal def of the "roll."
I think this is an excellent filter to build on!
But try to run both of them, first for today, then for 8 days ago (offset 8 days). You will note significant differences (okay, what else is new). But seriously, the differences are telling in several ways. One is the much lower number of hits for shorts then (20?), compared to now (65), and the slightly larger number of hits for longs then compared to now. This says of course that the filter is highly sensitive to the overall market (as it should be).
The other is the greater relative movement in the predicted direction with the shorts compared to the longs. Both of these are telling us something about the overall market psychology.
My suggestion, therefore:
Add a volume (momentum) or activity-related parameter (RSI), to relax the effect of the price alone while adding an activity indicator.
Perhaps someone could tell us how to incorporate the Fibonacci levels as such a filter?

StockFetcher Forums · Filter Exchange · Help with a Reversal Filter Please !<< 1 2 3 >>Post Follow-up

*** Disclaimer *** does not endorse or suggest any of the securities which are returned in any of the searches or filters. They are provided purely for informational and research purposes. does not recommend particular securities., Vestyl Software, L.L.C. and involved content providers shall not be liable for any errors or delays in the content, or for any actions taken based on the content.

Copyright 2016 - Vestyl Software L.L.C.Terms of Service | License | Questions or comments? Contact Us
EOD Data sources: DDFPlus & CSI Data Quotes delayed during active market hours. Delay times are at least 15 mins for NASDAQ, 20 mins for NYSE and Amex. Delayed intraday data provided by DDFPlus