Intraday gaps

[Fatafim]

Hello,

Basically, I would like to use some intraday, say 15m, data. However VBT doesn't understand, that it should treat all the data continously, and it calculates the indicators incorrectly. How can I deal with that efficiently?

[polakowo]

You need elaborate, what do you mean indicators are not calculated correctly? Just pass 15min data, what's the issue?

[Fatafim]

Okay I'm gonna try to clarify it a little bit. I'm taking sma cross for 15 and 30 periods, the data is us100 in 15m from the 06-02-2022. When I visualise it on VBT, it looks like this:

Nevertheless, comparing it to TV, even the number of trades is different:

As I take it, I don't think the issue is the plot, it seems to me as if the MA was calculated partially on NaNs, causing it to calculate something like (NaN+2+3)/(NaN+1+1). Or am I wrong here? How can I deal with that?

[polakowo]

If you have NaNs, drop them.

[Fatafim]

In my dataset as such there is no NaNs

[polakowo]

Then where those straight lines on the first chart are coming from?

[polakowo]

If you use YFData, all questions related to data quality should be asked on the GitHub page of yahoo-finance.

[Fatafim]

Not at all, 12-13 of february was a weekend - TradingView does not contain this data neither, but it simply skips it and treats it as if these days didn't exist. I'm aware that there are periods which do not contain information, however that is the core of the problem - traditional market does have such gaps, unlike crypto for example. That's why I'm also confused by VBT - I was looking for some functionality that would take in into account, but (maybe I didn't see something?) I haven't found anything

[polakowo]

vectorbt skips this data as well, what you see on the graph is plotly simply interpolating because it cannot just skip dates on the X axis.

[Fatafim]

After looking more into this case I actually managed to overcome this problem:

fig = portfolio.trades.plot()
sma1.ma.vbt.plot(fig=fig)
sma2.ma.vbt.plot(fig=fig)
portfolio.positions.plot(fig=fig)
fig.update_layout(xaxis = dict(type="category"))

Indeed the indicators are correct, my fault. Unfortunately with this layout all the index is displayed, but I'm quite sure it's possible to 'overcode' it as well. Would it be possible to add this feature to VTB as a boolean parameter anytime soon? It would be most useful, particularly for markets that run for less than 8h a day.

[polakowo]

Worth checking: https://stackoverflow.com/a/65632893