Why Are There No Double-Bottom Scans

[VR555]

A couple of searches for a scan for double bottoms has only shown indicators. Are there any scans for double bottoms on this site, or is the one built into thinkorswim the best?

 

[useThinkScript]

The double bottom indicators on the forum are repainters
https://usethinkscript.com/search/2552326/?q=double+bottom&c[title_only]=1&o=date

No, the forum does not support repainting scans
No there are no non-repainting double bottom studies, as the repainting logic is required.

The Problem:
There is no way of identifying that the low is the bottom until long after that bottom formed.
Sadly, confirming that the bottom formed last week, is not what daytraders are seeking.
They want to know that the double-bottom is forming now.

The Solution:
Repainters identify every new low as the bottom, in hopes that this is where the bleed ends.
When it is not, it erases that point, riding it down to the next tick, and identifies that as the bottom.
This is helpful for daytraders, who use this as immediate confirmation to their other indicators that are lining up; that a potential reversal is forming.

Meanwhile a scan would trigger on every downward tick, rendering its usefulness less than optimal.

Best usecase for repainters: use your other non-repainting indicators that are lining up as your scan filter and then review your chart for the double-bottom as confirmation.

 

[VR555]

The one I’m referring to is in Scan Hacker > Add Filter > Patterns > Double Bottoms. That may not be exactly right since I’m operating from memory, but it should be pretty close to show what I mean. Thank you, as always, for your help.

 

[useThinkScript]

ThinkOrSwim does not provide the ability to view or edit the underlying ThinkScript pattern code for the double-bottom filter.

When you click "Edit" on a Pattern filter in the Stock Hacker scan tab, you are only allowed to modify the GUI parameters (Pattern length, Direction, Trend, Breakout).​

To answer your core question:
The code powering this built-in pattern is proprietary and hardcoded into the platform's backend.
Therefore, there is no way to know whether it is repainting.

Hazarding A Guess:
In the editable parameters, you see breakout: complete or emerging

Perhaps that could mean that the complete-type is a lagging identification of historical double bottoms? and
emerging-type would more likely be repainting attempts at current double bottoms?


When you add this pattern to your chart, what type of behavior do you see?

 

[antwerks]

The one I’m referring to is in Scan Hacker > Add Filter > Patterns > Double Bottoms. That may not be exactly right since I’m operating from memory, but it should be pretty close to show what I mean. Thank you, as always, for your help.

Try this - start with broad scan and then narrow yourself or scan with another scanner like we were talking about in another post-

input lowLookback = 40;
input tolerancePct = 3.0;
input reboundMinPct = 5.0;
input shortMA = 21;

def c = close;
def h = high;
def l = low;
def ma = ExpAverage(c, shortMA);

def priorLow = Lowest(l[10], lowLookback);
def pctFromPriorLow = 100 * (c - priorLow) / priorLow;
def nearPriorLow = AbsValue(100 * (l - priorLow) / priorLow) <= tolerancePct;

def reboundAfterLow = Highest(h, 10) >= priorLow * (1 + reboundMinPct / 100);
def structureRepair = c > ma;

plot scan = nearPriorLow and reboundAfterLow and structureRepair;

What this is doing
It tries to avoid random falling knives by requiring:

• a prior low exists
• price revisits that zone
• there was at least some rebound off the first low
• price is starting to repair above short-term structure

This still may catch some sloppy names in broad ranges. That is why I’d rank them with additional columns.

Scan B: Undercut and reclaim​

This is the stronger one.

input lowLookback = 40;
input undercutPct = 2.0;
input reclaimMA = 10;

def c = close;
def l = low;
def priorLow = Lowest(l[5], lowLookback);
def undercut = l < priorLow and l >= priorLow * (1 - undercutPct / 100);
def reclaimed = c > priorLow;
def aboveShortMA = c > Average(c, reclaimMA);

plot scan = undercut and reclaimed and aboveShortMA;

This captures a very tradeable pattern family:

• prior low tested
• sellers pushed it marginally lower
• stock reclaimed the low
• right-side repair has begun

That often matters more than finding two mathematically equal lows.

Scan C: Higher-low repair​

This is for stronger names that do not need a full retest.

input lookback1 = 30;
input lookback2 = 10;
input shortMA = 21;
input longMA = 50;

def c = close;
def l = low;
def maShort = ExpAverage(c, shortMA);
def maLong = Average(c, longMA);

def leftLow = Lowest(l[lookback2], lookback1);
def recentLow = Lowest(l, lookback2);

def higherLow = recentLow > leftLow * 1.02;
def repaired = c > maShort and maShort >= maLong * 0.98;

plot scan = higherLow and repaired;

A lot of the best swing names do not print a clean second bottom. They just stop going down, print a higher low, and repair. You want those too.

 

[VR555]

Try this - start with broad scan and then narrow yourself or scan with another scanner like we were talking about in another post-

input lowLookback = 40;
input tolerancePct = 3.0;
input reboundMinPct = 5.0;
input shortMA = 21;

def c = close;
def h = high;
def l = low;
def ma = ExpAverage(c, shortMA);

def priorLow = Lowest(l[10], lowLookback);
def pctFromPriorLow = 100 * (c - priorLow) / priorLow;
def nearPriorLow = AbsValue(100 * (l - priorLow) / priorLow) <= tolerancePct;

def reboundAfterLow = Highest(h, 10) >= priorLow * (1 + reboundMinPct / 100);
def structureRepair = c > ma;

plot scan = nearPriorLow and reboundAfterLow and structureRepair;

What this is doing
It tries to avoid random falling knives by requiring:

• a prior low exists
• price revisits that zone
• there was at least some rebound off the first low
• price is starting to repair above short-term structure

This still may catch some sloppy names in broad ranges. That is why I’d rank them with additional columns.

Scan B: Undercut and reclaim​

This is the stronger one.

input lowLookback = 40;
input undercutPct = 2.0;
input reclaimMA = 10;

def c = close;
def l = low;
def priorLow = Lowest(l[5], lowLookback);
def undercut = l < priorLow and l >= priorLow * (1 - undercutPct / 100);
def reclaimed = c > priorLow;
def aboveShortMA = c > Average(c, reclaimMA);

plot scan = undercut and reclaimed and aboveShortMA;

This captures a very tradeable pattern family:

• prior low tested
• sellers pushed it marginally lower
• stock reclaimed the low
• right-side repair has begun

That often matters more than finding two mathematically equal lows.

Scan C: Higher-low repair​

This is for stronger names that do not need a full retest.

input lookback1 = 30;
input lookback2 = 10;
input shortMA = 21;
input longMA = 50;

def c = close;
def l = low;
def maShort = ExpAverage(c, shortMA);
def maLong = Average(c, longMA);

def leftLow = Lowest(l[lookback2], lookback1);
def recentLow = Lowest(l, lookback2);

def higherLow = recentLow > leftLow * 1.02;
def repaired = c > maShort and maShort >= maLong * 0.98;

plot scan = higherLow and repaired;

A lot of the best swing names do not print a clean second bottom. They just stop going down, print a higher low, and repair. You want those too.

Are these put into an existing scan as an additional filter?

 

[antwerks]

Are these put into an existing scan as an additional filter?

they are scans on their own but you can layer them in the scanning hacker like Merry was saying scan within a scan so to speak

https://imgur.com/FjOEiZU

 

[VR555]

mmm, I like how that would work. I'll start working with it in a while (pesky Dr. app'ts right now). Thanks so much for sharing!!!

 

[antwerks]

mmm, I like how that would work. I'll start working with it in a while (pesky Dr. app'ts right now). Thanks so much for sharing!!!

How was the Chili?

 

[VR555]

It was terrific. Thank you so much for the recipe!

 

[VR555]

I now have 12 scans that you have been very kind and generous to share recently. My challenge is deciding which ones to spend most of my effort on. Any guidance/advice on these scans would be much appreciated.
(PS - I'm posting this on the thread about 'Forex...' too since we've been working in both, to give a better chance for you to see soon)