Trading with Market Statistics. IV Standard Deviation

[walterw]

	Quote:	Dogpile
	Check this out... (first note that yes, my code is a bit flawed but I don't think by much since there is little movement here in the VWAP number so the numbers are not far off...) Look how my std dev bands compress and expand. This shows classic market behavior in that it is oscillating between 'balance' and being 'out of balance' -- a core market profile concept.... (this chart is using rolling 90-minute period for its std dev (45-period chart on a 2-min chart)-- its acting like a bollinger band does -- using rolling data -- but based on VWAP rather than a moving average). More interesting is look how price breaks lower out of the triangle/balancing and forms a classic 'bear flag' prior to breaking lower. I used to hate coils -- but the VWAP std dev bands show a market that is doing its thing -- coming out of balance then going back into balance -- then coming out again... knowing which environment you are in is crucial. you can see how the market shot out of the balance with a jolt down (it actually gapped). this was followed by a mini bear-flag which set up excellent location for a short... Yes, I traded this pattern today. http://bp0.blogger.com/_5h-SWVGx6Ms/...Coil+Break.bmp

interesting contraction expansion pattern there with vwap and bands..

[jperl]

	Quote:	Dogpile
	this chart is using rolling 90-minute period for its std dev (45-period chart on a 2-min chart)-- its acting like a bollinger band does -- using rolling data -- but based on VWAP rather than a moving average).

Dogpile--glad to see you are using VWAP in your own way and that you find it useful. I hope it works for you.

I have a problem using rolling 90-minute periods or for that matter any N-period method for computing the SD or N-period technical analysis in general. (This is includes all moving averages, CCI's, stochastics, MACD's, RSI's and any other method that requires a period length, and yes, sad to say Market Profile Analysis which uses 30 minute periods and an arbitrarily defined value area). The period length is arbitrary and would have to be readjusted when market properties change as they do daily. This is why I am presenting this general statistical method of viewing the markets. It's independent of period length of your chart and only depends on your starting time.
As we delve deeper into this you will see the utility of using this generalized statistic.

[jperl]

NEWBIE now has an arsenal of tools to trade with, all generated from the volume distribution function. He knows how the distribution is skewed by comparing the VWAP to the PVP, and he knows how volatile the market is by including SD bands above and below the VWAP. The SD now determines is exit strategy. If he enters at the VWAP, his exit will be 1 standard deviation above the VWAP (for long trades ) or 1 standard deviation below the VWAP (for short trades).

Here is an example from today's ER2 price action of a trade that NEWBIE takes with a VERY LARGE SD. Watch it to see how NEWBIE trades it. In this video NEWBIE has the opportunity to make 4 or 5 points because of the large SD. But he doesn't. He properly exits early. Watch it and understand why he exits where he does.

ER2shortJuyly25

[Dogpile]

<<The period length is arbitrary and would have to be readjusted when market properties change as they do daily.>>

yes I agree with you but I think the point of indicators is simply to aid you in identifying the current 'structure' of the market. my 90-minute indicator is not signalling a trade -- it is simply helping to clarify the current structure. I could have identfied this by simply drawing a triangle. But seeing the bands come together around the VWAP jumps out at me visually and 'speaks to me.' as Jim Dalton says ~"the human brain is particulary adept at identifying visual patterns" -- indicators are good but clearly only good in context.

[walterw]

Thanks Jerry for this great presentations ¡¡ they really have a tremendous educational value... I like your Newbie character ¡¡ he is the personification of a calm and relaxed in control trader... cool stuff keep it up... thanks Walter.

[Dogpile]

Jerry, can you explain in a mathematical or logical way why the PVP matters again for your directional bias? are you just saying that the VWAP can be considered the the 'real' point of value and so logic calls for the VWAP to be the 'eventual PVP'? Thus, if PVP is above the VWAP, expectation is for 'next PVP' to be something that is closer to the current VWAP and therefore odds favor the short side? and vice versa... is that right? if you could explain it in a different way that would be much appreciated. thx in advance

[nickm001]

Hey Frank, I have posted my code in Tradestation forum for calculating cumulative SD of the price , centered about VWAP. This gets you closer to what Jerry is talking about.

https://www.tradestation.com/Discuss...Topic_ID=66875

[Dogpile]

<<I have posted my code in Tradestation forum for calculating cumulative SD of the price , centered about VWAP. This gets you closer to what Jerry is talking about. >>

thanks nick...

you can see how the bands you have act more like 'keltner channels' (steadier band-width) and the bands I did act more like 'bollinger bands' (expand and contract). each has its own interesting characteristics which I have studied for years. The flaw with both of these classic indicators is that they both assume the current 'moving average' to be a point of 'value' -- but this is often way off the mark. VWAP is a much, much better estimate of current value so this is really quite a powerful concept we are discussing here. I am going to leave both sets of bands on my chart to watch the interaction just as I have always done with the classic indicators.

Volatility is not a constant in financial markets -- it is constantly oscillating between 'balance' (vol contraction) and 'out of balance' (price auctioning to a new point of balance). Every mathematical model must make an assumption about future volatility. This method Jerry has showed in the video uses straight historical data. If using this to trade, just be aware that this is effectively making the assumption that the 'future is like past' -- that is, expecting historical volatility to continue into the future. This assumption is loaded with risk -- known as 'model risk'. I think using both sets of bands (yours and mine) gives a more complete picture of the nature of financial markets. That is, you see the historical volatility alongside the expanding/contracting volatility-bands and this helps to show the way the market 'breathes'. See chapter 19 of the book 'Street Smarts' (Rashke) for more on this topic -- it is highly fascinating.

[Nick1984]

	Quote:	jperl
	Good observation Nick. There is a solution to this which we will eventually get to in later threads having to do with how you incorporate previous days, weeks, months VWAPs and their SD into todays price action. At this point in time, our NEWBIE trader waits for the distribution to develop and also waits for the price action to touch the VWAP. But, coming up in the next thread, we will introduce a paradigm shift in NEWBIE's thinking. Stay tuned. Perhaps you might want to expand on this, to give us an idea of what you are thinking about here.

Hey guys. Basically what I'm thinking of here with using regression with time as a variable is to see if its possible to get an idea of how price as the dependent variable moves in relation to volume and time i.e: for a one unit increase in time (being measured in whatever scale you prefer, 1 min, 5 min etc...) we may expect price to move between two points with 90% confindence (i'm thinking of a standard statistical 90% confidence interval here). This is obviously going to be dependent on volume as well.

If we have sufficient data to test how significant time as a variable is then we might be able to use time in a model to help get more precise entry points earlier on in the day.

My logic here is that over an average period of time data, price will tend to behave differently in various intervals over a trading time i.e: more active in the morning session compared to lunch session.

Then again I have no idea how to represent this graphically, and perhaps its impossible, but if you can somehow get the data to stream into some sort of logarith which computes price as a function of volume and time you can get a constantly changing estimated price based on the market stats being generated in real time.

Price = Beta0 + B1(Volume) + B2(Time) would be the statistical equation used i guess with your null hypotheses H0: B2 = 0 i.e time has no effect on price.

Once you've tested to see if the model is free from any statistical multi-colinearity etc, then you could use your estimated price based on your equation to gague whether the current market price is at "fair value".

I'm really sorry if that sounds very conveluted but its kind of late here and I've been up all day. If you have any questions ask and I'll try to clarify my thinking further.

I have no idea how to code stuff, and I'm hopeless with computers, so if anyone does understand what I'm on about and thinks its possible let me know and we can try to test it.

[Nick1984]

	Quote:	Dogpile
	<<I have posted my code in Tradestation forum for calculating cumulative