PercentRank

AlfredSisley

Hi All,

Looking for clarification regarding the PercentRank indicator.

Specifically, looking to understand if it's possible to get the PercentRank of another indicator.

Would be used in a multi instrument strategy, something like:

def __init__(self):

        self.o = dict() 
        self.holding = dict() 
        self.ema = []
        self.percentrank = []

        for i, d in enumerate(self.datas):   
            self.ema.append(bt.ind.EMA(self.datas[i],period=2))
            self.percentrank.append(bt.ind.EMA(self.datas[i],period=2),200)

The error I've been getting has been the following:

TypeError: __init__() takes 1 positional argument but 2 were given

The documentation states that 2 arguments are required (data + lookback period).

Is this possible? If so, how would I go about doing this?

Thanks in advance.

Alfred Sisley

run-out

Could you clarify what you are trying to do? Your error is most likely coming from

self.percentrank.append(bt.ind.EMA(self.datas[i],period=2),200)

You have defined self.percentrank as a list and you are trying to append two items to it, where only one is allowed.

I'm not sure where you are going with this. Provide some more information and we'll try to help you out.

AlfredSisley

The error is coming from that line 100%.

Trying to get the percentile of any indicator value over the last x periods.

Was originally trying to see if there was a way to somehow access all indicator values (value at each time period) in the data and apply percentile to see on any given day where it stands relative to history.

Would prefer the later but would settle for the former.

Let me know if you need further explanation.

Thanks!

Alfred Sisley

run-out

For one indicator comparing againt itself and going back n periods.

# In the strategy class...
 def __init__(self):
        self.prank = bt.ind.PctRank(period=5)

# results in with 0.00 the bottom and 0..80 the top rank: 
2005-05-17T00:00:00, o 12.26	h 12.28	l 11.98	c 12.21	v 61633100	prank 0.40
2005-05-18T00:00:00, o 12.21	h 12.43	l 12.13	c 12.33	v 44940000	prank 0.40
2005-05-19T00:00:00, o 12.36	h 12.64	l 12.35	c 12.43	v 40463700	prank 0.80
2005-05-20T00:00:00, o 12.45	h 12.64	l 12.40	c 12.55	v 25230900	prank 0.80
2005-05-23T00:00:00, o 12.62	h 12.77	l 12.60	c 12.70	v 47390500	prank 0.80
2005-05-24T00:00:00, o 12.66	h 12.86	l 12.64	c 12.80	v 43792600	prank 0.80
2005-05-25T00:00:00, o 12.72	h 12.79	l 12.61	c 12.75	v 32280900	prank 0.60
2005-05-26T00:00:00, o 12.82	h 12.98	l 12.77	c 12.92	v 31827400	prank 0.80
2005-05-27T00:00:00, o 12.86	h 12.90	l 12.79	c 12.85	v 21594400	prank 0.60
2005-05-31T00:00:00, o 12.77	h 12.87	l 12.69	c 12.80	v 28423900	prank 0.20
2005-06-01T00:00:00, o 12.79	h 12.97	l 12.77	c 12.89	v 33329000	prank 0.60

@AlfredSisley said in PercentRank:

Was originally trying to see if there was a way to somehow access all indicator values (value at each time period) in the data and apply percentile to see on any given day where it stands relative to history.

Could you clarify this maybe with an example?

AlfredSisley

Absolutely.

For example, let's say I had the ROC indicator for my instruments which would look something like this in a multi-instrument strategy:


def __init__(self):
    
        self.roc = []

        
        for i, d in enumerate(self.datas):   
            self.roc.append(bt.ind.ROC(self.datas[i], period=self.params.period))

Where percentrank would come in is when I would want the ROC output compared to historical output. This way I can assess the indicator from a historical perspective.

My initial thought was to do something like this:


def __init__(self):
    
        self.roc = []
        self.percentrank = []
        
        for i, d in enumerate(self.datas):   
            self.roc.append(bt.ind.ROC(self.datas[i], period=self.params.period))
            self.percentrank.append(ROC_INDICATOR_ABOVE, period=PERIOD_FOR_TICKER)

Notice above I've tried adding (in partial pseudo code) percentrank to the ROC indicator for the period of the entire dataset of a ticker.

In essence, this is just trying to take an indicator of an indicator.

Let me know if you need more information.

Appreciate the help!

Alfred Sisley

run-out

@AlfredSisley

How about this if I understand you correctly.

def __init__(self):
    
        self.roc = {}
        self.percentrank = {}
        
        for i, d in enumerate(self.datas):   
              roc = bt.ind.ROC(self.datas[i], period=self.p.period
              prank = bt.ind.PctRank(roc, period=self.p.period)
             
              self.roc[d] =roc
              self.percentrank[d] = prank

Then when you are in next, you can access the values like:

for d in self.datas:
    # then your prank value per data
    whatever_list_or_var = self.percentrank[d]

AlfredSisley

This is exactly what I was looking for.

Thank you very much.

Alfred Sisley

AlfredSisley

@run-out

One quick follow up question. Is there a way to access all historical close values in the init for each ticker?

Thanks

Alfred Sisley

run-out

@AlfredSisley said in PercentRank:

One quick follow up question. Is there a way to access all historical close values in the init for each ticker?

For what purpose?

AlfredSisley

Looking to somehow use PctRank to get bottom percentile of historical ROC data (historical = entire data set).

run-out

If I understand you correctly you want to calculate data ahead of the back test using the entire data set. If you were to use this data to trade you have look ahead bias. Backtrader doesn't do this.

Unless I understand you wrong.

AlfredSisley

@run-out You are correct. Needed to go back to the drawing board as a result. Thank you.

Sam Smith

Why choose competent writers paper writer to help you compose your term paper help? Below are some of our guarantees that will keep you coming back to order more essays.