Welcome to NexusFi: the best trading community on the planet, with over 150,000 members Sign Up Now for Free
Genuine reviews from real traders, not fake reviews from stealth vendors
Quality education from leading professional traders
We are a friendly, helpful, and positive community
We do not tolerate rude behavior, trolling, or vendors advertising in posts
We are here to help, just let us know what you need
You'll need to register in order to view the content of the threads and start contributing to our community. It's free for basic access, or support us by becoming an Elite Member -- see if you qualify for a discount below.
-- Big Mike, Site Administrator
(If you already have an account, login at the top of the page)
If you compare STD bands from SC to NT( using Fat tails shared indicators), there can be a significant difference when using bigger resolution/ larger time-frames. This is most apparent when looking at data over a small duration, eg the current day. Although using smaller time frames for accuracy is advisable, it could be more important for those using SC.
Also, the rolling VWAP study in SC now functions correctly, it used to plot the wrong look back period.
I think that both accuracy and CPU performance are issues with VWAPs.
The accuracy basically depends on two factors
-> input data used
-> model used for calculating the VWAP, if aggregate input data is used
EXACT VWAP
A 100% accurate VWAP can only be calculated from single transactions, tick by tick. This should serve as a reference. The inconvenience of using 1-tick data is the unnecessary CPU load.
VWAP calculated from aggregated data
When aggregated data is used to calculate VWAP and standard deviation bands, there will be an error term. The error term basically depends on
- the resolution of the input data
- the model used to calculate the VWAP for a single bar
Using an advanced model (weighted average of different data points within a single bar) will mitigate the impact of low resolution data and reduce the error term.
Finding the best compromise between accuracy and CPU performance
In my personal experience, the best compromise between accuracy and CPU performance can be obtained when 1-minute data is used as an input series. With 1-min data and a recursive algorithm, CPU load is extremely low and the typical error term is a fraction of a tick only.
Attached you will find three charts
-> VWAP calculated from tick data (100% accurate)
-> VWAP calculated from 1-min data (compromise)
-> VWAP calculated from 5-min data (primary bars)
Values for 2nd SD bands end of day
-> accurate VWAP bands: 212.40 / 210.72
-> VWAP bands calculated from 1-min data: 212.40 / 210.72
-> VWAP bands calculated from 5-min data: 212.38 / 210.73
This is just one example to show that the error term becomes significant, when you use input data with a lower resolution than 1-minute data. And yes, the inaccurate chart looks better than the accurate ones. But that is life.
I just pointed out the above because I was comparing VWAP bands in SC vs NT vs TOS, and noticed that SC seemed to be more dependent on using a lower resolution( When looking at data over a short duration). My thinking was that it was probably more due to the input used, as you're using "center of gravity" of each bar, where the most accurate in SC appears to come from using the "last".
I believe that is completely unnecessary. Just select "use underlying data" and SC automatically does the rest. If I'm wrong someone please correct me.
Using underlying data will make the VWAP equal to a 1 tick chart, however the bands still require a lower resolution since the STD computation is based on bars.
As fat tails has pointed you need to be using tick data if you want real bands accuracy( although most of the time there won't be much difference). I knew this, I just didn't realize how significant it could become.
Eg Here's my actual chart showing weekly SPY on a 15min vs an overlay study based on vwap/ bands from a 1 tick chart.
On the 1 tick overlay price tags 3STD, on the 15min it does not.
I did a quick comparison between Vwap with STDs with Study/Overlay from tick chart and the one using the underlying data. the Vwap is always the same, the STDs are different, this become more evident on higher timeframes.
Yellow lines are from overlay 1 tick chart, grey are from using the underlying.
Said that, I'm thinking if other studies that use the underlying data, such as the Volume at Price, TPO profile, Volume and TPO value area lines etc. may have the same problems, although I'm thinking that only Vwap's STDs have not been programmed to use the underlying data.
Correct, and the sc code does this when you specify use Base On Underlying Data. The sc code is there for all to see in studies3.cpp as scoff_VolumeWeightedAveragePrice().
Yeah VWAP is the same on any time-frame providing you select "based on underlying data". However the STD bands are calculated based on bar values, so when using "based on underling data" STD bands on a 1min chart will not be the same as STD bands on a 15min- even though VWAP will be identical. You need to use the overlay study if you wan true band accuracy using higher time frame charts. As pointed out above.