[wje3]

Hi Everyone;
This is my first post here. I have been reading Jerry's threads on Trading with Market Statistics with great interest. I have a few questions:
I am considering using market statistics yet using it for End of Day trading. I remember reading somewhere within the discussions that Jerry felt that this methodology would work on any time frame. I would be using a 20 day time period and closing prices. The 20 day period is somewhat arbitrary yet it is one month of trading days.
My questions pertain to using Excel to generate the relevant data points.I am no expert in Excel and usually rely on my children to walk me thru the formulae, but once done it's pretty straight froward.
Will Excel give me the correct data points, especially the SD of the VWAP ? The PVP, and VWAP calculations are straight forward. But after all the lengthy discussion about variance and calculating the standard deviation of the VWAP, I want to make sure that what Excel says is the SD is valid.I would use the 20 days of VWAP to calculate the SD.
I'm not a statistician/mathematician nor an Excel expert. I'm just trying to understand any limitations of the statistical tools that Excel uses.
Thanks in advance for your help.
Bill

[BlowFish]

Yes if the formulas are correct. Does excel have a weighted SD? You might need to do a bit of dinkering to get that.

I would be as concerned about PVP (and lesser extent VWAP). If you are using the closing price of the daily bar you will be lumping all the volume for the day at that one price point (the close). If you can sample data every hour (or better still every minute) to get the volume distribution at price things will be more accurate. No reason not to do the calculations once per day but I would personally be happier with a sampling rate that is higher.

[wje3]

Hi BlowFish;
Thanks for your answers.
Regarding the once a day sampling rate,lumping all the volume in for the closing price-yes I am a little concerned about that as it isn't very "granular". Yet, the underlying theory is the same as with two minute bars,if I understand it correctly. With two minute bars, all the volume is attributed to the close of that interval. With daily bars, it's the same underlying concept. I could do an average price for the day which might be more reflective of the days volatility.
Actually, what I am more concerned with is the statistics of the standard deviation. Why do I need a weighted SD. If I am trying to determine the SD for the VWAP, that series of numbers(the VWAP) is already weighted. I'm just trying to determine the SD of a series of numbers. Why does the SD need to be weighted as well?
Thanks for your help.
Bill

[BlowFish]

Hi Bill, sounds like you understand the sampling issues. If you are looking 20 days back thats only 20 samples. As you correctly observe Jerry samples every 2 minutes. He talks about what he considers enough data....from memory he likes to see at least the bands open up wider than the bars themselves. I dunno I would have thought things are only just going to be settling down after 20 bars, it's a fairly small sample.

I did some comparisons (a long while back) of Jerry's 2 minute sample and raw tick data (so completely accurate) usually the VWAP & SD's where negligibly different but occasionally there would be a 2 or 3 ticks difference. The PVP is a bigger issue. It is very sensitive to how you allocate the volume how you sample and accumulate the peak can completely alter the PVP. This is because it jumps around based on the highest peak. Incidentally the way Ensign (which Jerry uses) allocates volume to PVP is that it pro ratas it to every level/tick between the high and the low of the bar. Allocating all the volume to a single price will likely give quite some odd results.

Jerry talks about why the SD's should be weighted by volume in the threads.

[wje3]

Hi BlowFish;
Thanks for the information. I agree that the 20 day sampling is inadequate. There are large vacant areas on the volume histogram. I'm going to continue looking at the feasibility of using this on a larger time frame, e.g. daily
I'll also re-read the threads on the weighted SD.
Thanks again for your help.
Bill

[BlowFish]

Hey Bill,

Even if you want to look at the longer time frames if you can build what you need with more data (like hourly perhaps) you should be golden. You might want to look at Ninjatrader you can get it free and there are indicators for VWAP SD and PVP in the public domain. of course there may be other reasons that you are leaning towards excel in which case that might not make sense for you.

[wje3]

Hi BlowFish;
Funny you should mention NinjaTrader. I just downloaded it last night and I'm going to start playing with the charts. 20 days of hourly data might work but I'll take a look at it over the next few days.
Thanks for the suggestion.
Bill

[BlowFish]

Keep us posted how you get on