candlesticks screenings stocks analyzer

Maciek Paciarski

Hi,
I am trying to create an analyzer class based on the formations of candlesticks contained in the TA-lib library which can be used together with the Backtrader. This class determines whether some stocks contains recently a given candlesticks formation and consequently it lets me create a set of companies which recently include such a formation and a set of companies which don't. However, I encountered a wall I'm trying to overcome, I would be grateful for the tip on how to overcome this , namely :

TA-lib candlesticks pattern recognition gives the result in the form : -100, 0 or 100 .
Inputs are prices : open , high , low close .

class Screener_CandleSticks(bt.Analyzer):

    def start(self):
        self.candles = {data: bt.talib.CDLHARAMICROSS(data.open, data.high, data.low, data.close)
                        for data in self.datas}

    def stop(self):
        self.rets['within'] = list()
        self.rets['outsite'] = list()

        for  data, candle in self.candles.items():
            note = data._name, candle[0]
            if candle==100
                self.rets['within'].append(note)
            else:

                self.rets['outsite'].append(note)

question: am I on the right track giving the comparative condition:

if candle==100

or maybe other values should be called here?

a small tip would be priceless.

dasch

from https://github.com/mrjbq7/ta-lib/issues/74:

The -100 denotes a bearish Tristar pattern where the middle candle body is above the other two. Conversely +100 denotes a bullish Tristar pattern where the middle body is below the adjacent ones.

So this may apply to these, too

dasch

-100 = bearish pattern
+100 = bullish pattern

Maciek Paciarski

@dasch
Thanks for your attention.
this entry which I used here throws me an error:
```

if candle==100
                 ^
SyntaxError: invalid syntax```

it looks like the syntax should look different when using the analyzer class rather than the
strategy class

dasch

You would need to check candles for the signal, since this will be something like a numpy Array or some kind of list. Maybe look into the data in candles to see what to do with candle ...

Maciek Paciarski

@dasch Hi
That is, the dates cannot be a type object -->'backtrader.feeds.csvgeneric.GenericCSVData'. Data rows (open , high,low,close) must be classes 'pandas.core.series.Series' to use class object like 'bt.talib.CDLHARAMICROSS' in the analyzer of the backtrader . do i reason well ?

dasch

Oh i see, I did not run your code. but by reading:

https://www.backtrader.com/blog/posts/2016-07-26-talib-integration/talib-integration/

your approach should work.

but what exactly do you want to?

if you just want to look if there is such a formation in your data (without looking for the date) you can do the following in stop:

    def stop(self):
        self.rets['bullish'] = list()
        self.rets['bearish'] = list()
        self.rets['outsite'] = list()
        for data, line in self.candles.items():
            note = data._name
            bullish, bearish = False, False
            for i in range(0, len(line)):
                if line[-i] == 100:
                    bullish = True
                elif line[-i] == -100:
                    bearish = True
            if bullish:
                self.rets['bullish'].append(note)
            if bearish:
                self.rets['bearish'].append(note)
            if not bearish and not bullish:
                self.rets['outside'].append(note)

The code above will look through the line, going from newest data to oldest and check if there was a formation, if there was, it will be stored in a bool var (bullish, bearish).
Then it will store the result in the rets list

Maciek Paciarski

@dasch
Thank You so much for help . I wanted to create a stocks screening similar to the example on the backtrader's blog but with candlesticks formations in the main role. Thank You again for your attention.

dasch

glad it helped.

pozdrawiam

dasch

@Maciek-Paciarski the example in the blog uses the last price.

https://www.backtrader.com/blog/posts/2016-08-15-stock-screening/stock-screening/

since you are checking the whole line, not only the last entry, you would need to iterate through all candles. with var i you can also get the date from data and other values. you would access it the same as the line you generated in init:

data: bt.talib.CDLHARAMICROSS(data.open, data.high, data.low, data.close)

so the line with bt.talib.CDLHARAMICROSS (think of a list) contains all occurrences of the pattern for every candle in data.

with data[-i] you would get the candle on which the pattern recognition was triggered

        for data, line in self.candles.items():
            note = data._name
            for i in range(0, len(line)):
                if line[-i] == 100:
                    bullish = True
                    print("BULLISH", data.time[-i])
                elif line[-i] == -100:
                    bearish = True
                    print("BEARISH", data.time[-i])

with

            for i in range(0, len(line)):

you go through the line starting with the newest entry to the oldest entry.