JonnyBoy 

Firstly, it is very simple to calculate

SMCJB 

The problem is it's actually not. It's easy to put on a real time chart but that's the only way it can be calculated. If I pull up a CL daily chart how do I add the the 20 day average of the VWAP? I can't because the only way to know the VWAP is to have calculated it already. Sure I could actually have 20 days of 1 min buys and estimate the VWAP for each day from that. But what about when I'm pulling 10 years of daily bars to perform some calculations? I can calculate most (all?) indicators, but not the VWAP.

While VWAP should technically be an average at the level of single trades, as a practical matter I think it is usually computed by taking fairly small time intervals (such as the 1-minute data you mentioned), often averaging something like OHLC to get a more representative price, and basically using that as your price and the volume of the bar in your calculation. This means it's not exactly an "average trade," but the question is whether it is a close enough approximation.

I recall reading something like that in @Fat Tails' NT version of the indicator, although the details temporarily escape me. Approximating the calculation this way would be good enough, at least it seems so to me.

A multi-day VWAP -- say, a weekly VWAP -- would not compute the separate VWAPs for each day, which I think is what you are saying, but would just run the weighted average for the underlying bar series straight through beginning on Monday and ending on Friday, then start over on Monday again.

I don't know if I understood you, but I don't think you would need to calculate individual days' VWAPs. Perhaps I missed your point.

If you wanted to do a 10-year VWAP you would need a rougher approximation than 1-minute bars unless you wanted to be overwhelmed by the volume of computations. But if you're using a large timeframe you can have a large allowance, I would say.

I'm not sure I quite got you point. Does this address it or am I simply confused about what you said?

Bob.

JonnyBoy 

I think I understand what you are saying about the placement of VWAP at intervals other than the data series specified on the chart. For that very reason I coded myself a VWAP anchor indicator. I can place a VWAP anywhere on the chart and it will forward project it for me. For example, attached is the 20 day VWAP for CL. This may or may not be what you are referring to, but I use it in the ES to anchor at LOD, HOD, month, week etc. or as you said a specified period in time you are interested in.

I've done this too. Simple, and you get a VWAP from any arbitrary start point.

Bob.

Anchoring a VWAP at market turns makes much more sense as others have mentioned. The buy/sell group responsible for turning the market usually defends the VWAP levels.
I have tried to anchor it at market pullback to mixed results. It does give a good picture of a tiring trend though.
Also, I have seen people use it as a trade evaluation at eod, by anchoring it at entry price.
Great points everyone.
Cheers.

Sent using the NexusFi mobile app

Just wanted to put some different thoughts into this thread too, I did a study back a few years, which is probably still much the same today;

Master Homework and Statistics Thread Average variance between the Session Mid and VWAP for ES RTH 2015. Difference calculated each 1min for the duration of the session. Mid being the rolling (Session High - Session Low) / 2 VWAP based on tick data I wanted to see if there was any …

It looked at ES RTH difference between Session Mid and VWAP; 75% of the time it is less than 4ticks difference, 97% is less than 3pts difference. (this was for 2015)

It's not making a conclusion in any direction, but you could perhaps go further and look into times where VWAP is extended away from Mid, what does this mean, etc..

But in the discussion of execution, you could for all purposes just look at above/below session Mid and be mostly the same as VWAP for execution performance.

Thanks @JonnyBoy and @bobwest

To clarify I think VWAP would be like the ultimate daily pivot calculation hence why I would like to be able to use it. When I say 20 day average of the VWAP I specifically mean average(Daily VWAP, 20) and not a VWAP calculation over 20 days. When I talk about 10 years of data, I don't want a 10 year VWAP calculation, I would like 10 years of data with daily OHLC AND daily VWAP. (Be interesting to feed daily VWAP into machine learning algos).

SMCJB 

Thanks @JonnyBoy and @bobwest To clarify I think VWAP would be like the ultimate daily pivot calculation hence why I would like to be able to use it. When I say 20 day average of the VWAP I specifically mean average(Daily VWAP, 20) and not a VWAP calculation over 20 days. When I talk about 10 years of data, I don't want a 10 year VWAP calculation, I would like 10 years of data with daily OHLC AND daily VWAP. (Be interesting to feed daily VWAP into machine learning algos).

Ah, OK now I got it. A completely different idea, to use the daily final VWAP as an input for an average of x days.

I can see how this would be useful. A little complicated to figure, because you'd have to run the VWAP calculation for each day and store the values, then take the average.

Light slowly dawns....

Bob.

------------------

Edit: Now I get this part too:

SMCJB 

I've always been surprised VWAP isn't more available, to the extent I'm a little surprised exchanges don't publish official VWAP numbers.

It would be a lot simpler if they just did that, so you could just grab them up. It is strange they don't... I guess you've checked for it, but are you sure? Seems like a simple, basic statistic that should just be there.

SMCJB 

Thanks @JonnyBoy and @bobwest To clarify I think VWAP would be like the ultimate daily pivot calculation hence why I would like to be able to use it. When I say 20 day average of the VWAP I specifically mean average(Daily VWAP, 20) and not a VWAP calculation over 20 days. When I talk about 10 years of data, I don't want a 10 year VWAP calculation, I would like 10 years of data with daily OHLC AND daily VWAP. (Be interesting to feed daily VWAP into machine learning algos).

Constructed in 5 minutes on Sierra Chart:

srgtroy 

Constructed in 5 minutes on Sierra Chart:

reviewed in less than 1 minute with ninjatrader. conclusion > not the desired outcome

Dueling platforms?

Explanations might be more helpful....

Bob.

bobwest 

Dueling platforms? Explanations might be more helpful.... Bob.

indeed.

but you know I had to make a smart remark. otherwise @srgtroy would be really disappointed

anyway I just replicated his chart. but I have no clue what the logic behind is.

if I would try to find a solution, I probably would try something with the vwma instead of the vwap. maybe @Fat Tails has a better idea?