Ability to plot arbitrary (non-ohlc) data

[trueneu]

Hello! First of all, thank you, Daniel, and contributors for this great library. Works like a charm!

I like the concise default style of the graphs produced very much. However, I was wondering, is it possible to plot arbitrary data without prices? Usecase: for example, I wanna compare P/E ratios over time for 10 different companies. I'm not really interested in share price movement for this, but I'd still like to reuse the mplfinance's concise visual style, and not to recreate it manually with matplotlib. (needless to say, the type becomes irrelevant in this case, maybe other things as well).

Sorry for the newbie question, and best regards.

[DanielGoldfarb]

Hi Pavel, Thanks for your interest in and appreciation for mplfinance. It's an interesting question.

I suppose we could add a type, perhaps call it data so that you could write:

mpf.plot(df,type=data)

or something like that, to indicate that the dataframe, df, simply has columns that you want to plot, with no expectation for columns 'Open', 'High', 'Low', and 'Close'.

---

In the meantime there is a workaround that you can use. You can create a dummy dataframe for the OHLC data, and pass in your PE data via addplot.

Suppose we have a dataframe with 5 years of monthly PE data for 8 stocks:

print(pedf.shape)
print(pedf.head(3))
print(pedf.tail(2))

(60, 8)

               AAPL    AXP     BA     CAT   CSCO     CVX    DIS     DWDP
Date                                                                    
2011-01-31  48.4743  43.38  69.48   97.01  21.15   94.93  38.87  23.8533
2011-02-28  50.4586  43.57  72.01  102.93  18.56  103.75  43.74  24.9827
2011-03-31  49.7786  45.20  73.93  111.35  17.15  107.43  43.09  25.3794

              AAPL    AXP      BA    CAT    CSCO    CVX     DIS     DWDP
Date                                                                    
2015-11-30  118.30  71.64  145.45  72.65  27.250  91.32  113.47  35.0471
2015-12-31  105.26  69.55  144.59  67.96  27.155  89.96  105.08  34.6101

---

Create a dummy OHLC dataframe, with the same dates as the PE data:

dum = [float('nan')]*len(pedf)
dum[0] = pedf.iloc[0,0]  # At least one valid data point to avoid warning from mplfinance

d = dict(Open=dum,High=dum,Low=dum,Close=dum)
dumdf = pd.DataFrame(d, index=pedf.index)

---

Now plot the PE data using addplot, along with the dummy ohlc dataframe:

ap = mpf.make_addplot(pedf,secondary_y=False)
mpf.plot(dumdf, addplot=ap, ylabel='PE Ratio')

The result:

---

Note that Pandas DataFrames have a .plot() method as well.
One advantage to using DataFrame.plot() directly is that you automatically get a legend based on the column headings. Legends are not yet supported directly in mplfinance, but if we implement type=data we could do something simlar to what padas is doing here:

pedf.plot()

---

Of course, with mplfinance you can use mplfinance styles:

mpf.plot(df,addplot=ap,ylabel='PE Ratio',style='nightclouds')

mpf.plot(df,addplot=ap,ylabel='PE Ratio',style='yahoo')

mpf.plot(df,addplot=ap,ylabel='PE Ratio',style='blueskies')

Hope that helps. Let me know if you might be interested in contributing type=data to mplfinance.

All the best. --Daniel

P.S. By the way, just curious if you have seen this: https://github.com/DanielGoldfarb/pffap

[trueneu]

Hi @DanielGoldfarb , hands down this is the best reply to any github question I've asked so far. Appreciate it, saves me a lot of time. As for contributing the implementation, sure, I'm all up for it. I'll first need to spend some more time playing around with mpf and pd, though. Do we leave this issue open for the upcoming implementation discussions/review, or just open another one in the future? As for pffap, haven't seen it, will definitely check it out. Thank you.

[manuelwithbmw]

P.S. By the way, just curious if you have seen this: https://github.com/DanielGoldfarb/pffap

Wonderful training Daniel! Loved the video

[DanielGoldfarb]

Pavel, We can leave this issue open for the enhancement/implementation of type=plain_data or whatever we end up calling it. No rush, but let me know when you have an idea how you want to implement it. Worst case, we could always do something similar to the above workaround, just hide it inside mpf.plot() so the caller doesn't see it. But there may be a cleaner way too; I haven't thought it through yet. All the best. --Daniel

[alexpvpmindustry]

type=plain_data

this is interesting. i can try this too