Missfit of tlines

[TheGrowingPlant443]

I may have done something wrong but this may also be a bug.

I have data from yahoofinance of the ticker symbol AMC from (2021-07-22 09:30:00-04:00 to 2021-07-28 15:59:00-04:00). I want to use the function tlines() but I get something that is not suppose to happen.

Code:

import yfinance as yf
import pandas as pd
import mplfinance as mpf

tickername = "AMC"

Ticker = yf.Ticker(tickername)

df = Ticker.history(period='1d', interval='1m', start='2021-07-22', end='2021-07-28')

datapairs = [('2021-07-22 09:38:00-04:00', '2021-07-22 09:58:00-04:00'), 
             ('2021-07-22 09:58:00-04:00', '2021-07-22 10:02:00-04:00'),
             ('2021-07-22 10:02:00-04:00', '2021-07-22 10:15:00-04:00'),
             ('2021-07-22 10:15:00-04:00', '2021-07-22 10:24:00-04:00'),
             ('2021-07-22 10:24:00-04:00', '2021-07-22 10:33:00-04:00'),
             ('2021-07-22 10:33:00-04:00', '2021-07-22 10:36:00-04:00'),
             ('2021-07-22 10:36:00-04:00', '2021-07-22 11:22:00-04:00')]

mpf.plot(df, tlines=[dict(tlines=datapairs,tline_use='high',colors='g')])

My problem is that the fit is wrong how do I fix this?

[DanielGoldfarb]

@TheGrowingPlant443

Can you please, after df = Ticker.history(period='1d', interval='1m', start='2021-07-22', end='2021-07-28'), call

df.to_csv('AMC_20210722_20210728.csv')

... and then post the csv file here, or put it somewhere that I can get to it (or email to [email protected]) so that I can be sure I am working with the exact same data you are working with. I will take a look. Thank you. --Daniel

[TheGrowingPlant443]

AMC_20210722_20210728.zip

@DanielGoldfarb

Thanks for the quick response. I have linked to a zip-folder with the csv file in it.

[DanielGoldfarb]

This is indeed a bug that was reported previously here.

The explaination is as follows: Whenever plotting a time series with matplotlib, the datetimes must be converted to low-level matplotlib dates. Mplfinance does this for you. Normally this works fine, but when your dataframe's datetime index contains a timezone offset. matplotlib's date conversion always converts the time zone to UTC. (I'm not sure why matplotlib does that).

In order to avoid this automatic conversion to UTC, mplfinance detects if a timezone offset is present in the dataframe index, and if so, then it "localizes" the pandas Timestamp objects. This generally works fine.

The bug is this: when I added code to localize the pandas Timestamps, (See issue #236), I forgot to add code to also localize any datetimes passed in as tlines and/or alines specifiers. Effectively then, your trend lines are being drawn using price data that is offset by the datetime offset (in your case, 4 hours).

This will definitely be fixed eventually. In the meantime there are a couple of workarounds that you can try:

1. The simplest workaround is to set kwarg tz_localize=False when calling mpf.plot(). This will allow your tlines date specifiers (like '2021-07-22 09:38:00-04:00') to match the offset in your data. Your trend lines will be correct, but your trading day will appear to start at 13:30 looking something like this:

2. Alternatively you can write some code to remove the time zone offset from your dataframe index, and then specify your tlines datetimes using no timezone offset:

   df = Ticker.history(period='1d', interval='1m', start='2021-07-22', end='2021-07-28')
   
   # Remove offsets from timestamps in the dataframe index:
   df.index = pd.DatetimeIndex(df.index.tz_localize(None).to_pydatetime())
   
   # Now continue as before, but with no timezone offsets in the `tlines` specification:
   datapairs = [('2021-07-22 09:38:00', '2021-07-22 09:58:00'),
                ('2021-07-22 09:58:00', '2021-07-22 10:02:00'),
                ('2021-07-22 10:02:00', '2021-07-22 10:15:00'),
                ('2021-07-22 10:15:00', '2021-07-22 10:24:00'),
                ('2021-07-22 10:24:00', '2021-07-22 10:33:00'),
                ('2021-07-22 10:33:00', '2021-07-22 10:36:00'),
                ('2021-07-22 10:36:00', '2021-07-22 11:22:00')
               ]
   
   mpf.plot(df, tlines=[dict(tlines=datapairs,tline_use='high',colors='g')])

The result is:

[TheGrowingPlant443]

Thank you so much for the help. This works perfektly 👍