Mplfinance: Bug Report: Candlestick too short

[toanalien]

When values of y-axis vertical too small 0

[toanalien]

I workaround and found solution:

ymin = df.loc[df['Low'].idxmin()]['Low']
ymax = df.loc[df['High'].idxmax()]['High']

mpf.plot(df,
    type='candle',
    style='starsandstripes',
    ...
    set_ylim=(ymin, ymax)
)

[DanielGoldfarb]

Thiên,

Thank you for reporting this issue. Can you please provide the exact data and exact code you used for this plot so that I can be sure to be able to reproduce the bug exactly? Also, please provide the output from mpf.__version__.

Thank you. --Daniel

[toanalien]

Thank @DanielGoldfarb quick reply!

mpf.__version__)
0.12.4a2

I fetched data from Binance

Date,Open,High,Low,Close,Volume
2020-05-19 09:00:00,2.11e-05,2.118e-05,2.106e-05,2.113e-05,324959.0
2020-05-19 10:00:00,2.114e-05,2.119e-05,2.107e-05,2.118e-05,400205.0
2020-05-19 11:00:00,2.119e-05,2.119e-05,2.094e-05,2.11e-05,1004006.0
2020-05-19 12:00:00,2.114e-05,2.133e-05,2.109e-05,2.121e-05,735355.0
2020-05-19 13:00:00,2.121e-05,2.125e-05,2.112e-05,2.116e-05,225666.0
2020-05-19 14:00:00,2.118e-05,2.122e-05,2.102e-05,2.111e-05,200795.0
2020-05-19 15:00:00,2.112e-05,2.132e-05,2.103e-05,2.13e-05,491363.0
2020-05-19 16:00:00,2.128e-05,2.132e-05,2.11e-05,2.124e-05,418583.0
2020-05-19 17:00:00,2.124e-05,2.136e-05,2.122e-05,2.127e-05,420179.0
2020-05-19 18:00:00,2.129e-05,2.131e-05,2.116e-05,2.12e-05,258836.0
2020-05-19 19:00:00,2.12e-05,2.133e-05,2.117e-05,2.124e-05,177158.0
2020-05-19 20:00:00,2.125e-05,2.133e-05,2.121e-05,2.125e-05,300338.0
...

https://gist.github.com/toanalien/62a3740b934c21c5e6d30691f97c9a20

[toanalien]

Issue when I add a plot scatter.

[DanielGoldfarb]

Thiên,

Thank you for posting your the data.

Without seeing the other data that you are using for your addplot scatter, I cannot know for sure, but I suspect that the _max and min of the scatter data_ has a much larger spread that the OHLCV data that you posted.

If so, then adding secondary_y=True to your addplot scatter should fix the problem.

I would very much like to see the addplot scatter data, because "in theory" mplfinance tries to determine whether or not to automatically use a secondary_y for the additional data. In this case it appears possibly to be getting it wrong. That is the reason we have to ability to force secondary_y to be either True or False (instead of "auto").

I would also recommend you use the version 0.12.4a0 since that is currently the most stable version, and is the version currently published to Pypi.

---

Fyi, here is my code, trying to reproduce your problem, and the plot it produces:

$ ipython
Python 3.7.4 (default, Aug 13 2019, 20:35:49)
Type 'copyright', 'credits' or 'license' for more information
IPython 7.8.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: import mplfinance as mpf

In [2]: mpf.__version__
Out[2]: '0.12.4a0'

In [3]: import pandas as pd

In [4]: idf = pd.read_csv('issue#142_data.csv',index_col=0,parse_dates=True)

In [5]: idf.shape
Out[5]: (41, 5)

In [6]: idf.head(3)
Out[6]:
                         Open      High       Low     Close     Volume
Date
2020-05-19 09:00:00  0.000021  0.000021  0.000021  0.000021   324959.0
2020-05-19 10:00:00  0.000021  0.000021  0.000021  0.000021   400205.0
2020-05-19 11:00:00  0.000021  0.000021  0.000021  0.000021  1004006.0

In [7]: idf.tail(3)
Out[7]:
                         Open      High       Low     Close    Volume
Date
2020-05-20 23:00:00  0.000021  0.000021  0.000021  0.000021  385498.0
2020-05-21 00:00:00  0.000021  0.000022  0.000021  0.000021  532283.0
2020-05-21 01:00:00  0.000021  0.000022  0.000021  0.000021  172228.0

In [8]: mpf.plot(idf,volume=True,type='candle',style='starsandstripes')

[DanielGoldfarb]

I did some more testing and I am able to reproduce the problem by creating a scatter plot of just highs. So the order of magnitude is clearly the same. Use secondary_y=True or set_ylim as a workaround for the issue. Either one one works (no need to use both).

This appears to be the exact same issue as reported here.

I think the issue is with matplotlib itself, and how it may handle ylimits differently for scatter plots versus other types of plots.

I tried to reproduce the issue _outside of mplfinance_, to find a very simple case that I can then post to the matplotlib issues page. Originally I thought I had reproduced it, but then was not able to again. (I must have changed something). I then put the issue aside to work on other stuff, intending to come back to it later. That's where it stands now.

[DanielGoldfarb]

I forgot that @free-soellingeraj figured out that upgrading to matplotlib v3.2.1 fixes the issue.

I have confirmed on my own machine that sudo pip install --upgrade matplotlib fixes the problem.

[toanalien]

Thank you for your quick reply. Problem solved by upgrade matplotlib.