Trading with Market Statistics. IV Standard Deviation

[Head2k 10]

If I want to use something I need to understand it. If I think of VWAPs this way it is easier for me to think of implications of where is price in relation to VWAP, and I can assingn a meaning to bounce or breakout and yes, it makes me more comfortable with VWAPs.

I read the thread you gave me a link to, concerning MIDAS method. This is basically a method of placement the beginning of computation into some significant places of shift of psychology and I really like it.

Whether it will help me as a trader I dont know... I am a newbie and I still have to find my own way. But VWAPs definitely contributed to my market understanding and to organising data in a logical structure. Sure they will not make me a successful trader but they are a powerfull tool IMHO.

[ramora 10]

Jerry, thanks for the information. When starting to calculate VWAP at the open it seems to be less helpful as the hours pass. Have you thought about starting a new VWAP every hour or at every PVP? As the spread between vwap and price increases later in the day, do you change your view of its value?

 

Thanks again

[jperl 11]

When starting to calculate VWAP at the open it seems to be less helpful as the hours pass. Have you thought about starting a new VWAP every hour or at every PVP? As the spread between vwap and price increases later in the day, do you change your view of its value?

 

Thanks again

 

No Ramora, I don't start a new VWAP every hour. You lose two important pieces of information by doing so: a) The present trend, determined by price relative to the VWAP and b)the day's volatility as measured by the standard deviation. The present trend determines my bias and thus whether I will favor longs or shorts. The SD tells me how much I should expect to pull out of the market when I trade. I don't want to lose either of those two pieces of info by restarting the VWAP computation.

[BlowFish 10]

Unless scalping of course (by your definition of scalping) In which case you might want to know the present hours trend and present hours volatility.

 

Edit: One of the things I played with was having two sets of bands that reset every hour, one set starting on the hour one on the half hour. You use the 'mature' set to trade from. This has the advantage of having some statistical significance (the set you trade have had at least 30 minutes to form) plus having 'scalp' characteristics. A 'continuous' set might be better but never got round to programming them.

[jperl 11]

Unless scalping of course (by your definition of scalping) In which case you might want to know the present hours trend and present hours volatility.

Yes of course. If you scalp you could start the VWAP anytime you want to have a quicky. This works well in non trending markets.

[Head2k 10]

I think it is much better to start VWAP at specific logical moments rather than to use random starts. A beginning of a session is a logical place to start one. Then you can start some at significant highs, lows or breakouts. Simply places of changes in psychology / behavior. Then you can assingn better meaning to bounces, breakouts and price action in general. See attached chart.

[BlowFish 10]

Head2k, that looks remarkably like an implementation of Midas. Do you have the indicator? What platform is it? Something I have been trying to locate for a while.

[Head2k 10]

Head2k, that looks remarkably like an implementation of Midas. Do you have the indicator? What platform is it? Something I have been trying to locate for a while.

As you might remember I use AmiBroker. I coded the indicator myself. It is a VWAP + 1st SDs with custom starts. Unfortunately AmiBroker is the only platform I can programe in, as I simply bought it and read the manual

[BlowFish 10]

<doh> Slaps head. Of course you do!! Senility is a terrible thing, worse if I haven't had coffee yet!

[darthtrader2.0 10]

You can call the sum of those traders a composite trader and you can distinguish composite buyer and composite seller. Now lets say price is below VWAP. That means that the composite buyer is losing and composite seller winning.

 

Well again, that assumes that the VWAP is the exact average in a normal distribution of price/volume and this is what actually happens in markets.

The entire idea of a "composite" seller/buyer is bogus...it negates the entire idea of a "trend day"...

[darthtrader2.0 10]

For example, if I enter a trade, how much should I expect the market to move.

The answer is simple if you are following the market statistics. It's one standard deviation.

 

which you should add is one std dev of a normal distribution, which is assuming a normal distribution of price/volume, which you admit yourself is wrong..

First, I'm not argueing that you are a baddass trader. you obviously are.. I highly suspect though your "baddassness" has much more to do with years of experience with price action and order flow and your discretionary ability to negate trades than this setup you put forth has to confirm them.

Maybe I'm right, maybe I'm wrong..but I dont' take anything in this game at face value just because someone says its so.

Maybe we are both right and the assumption of a normal distribution is good enough but some other probability distribution is slightly better...

why not try to find that function since we know a normal distribution is not optimal, but maybe that is foolish..

I do know your stuff doesn't backtest, so something else is going on..

[jperl 11]

which you should add is one std dev of a normal distribution, which is assuming a normal distribution of price/volume, which you admit yourself is wrong..

No darth, you are incorrect. The one std dev is for whatever the distribution is when you make the trade. It has nothing to do with the normal distribution. Again you still seem to be hung up on the idea that std deviations are defined in terms of the normal distribution only.

I highly suspect though your "baddassness" has much more to do with years of experience with price action and order flow and your discretionary ability to negate trades than this setup you put forth has to confirm them.

You are partially correct here. Years of experience do make a difference in how well you trade. The best analogy is playing a musical instrument. I can teach you all the techniques for playing, but it takes years of practice to develop your own style and to be good at it.

My point of the market statistics threads was to teach you a technique for analyzing the data and how you might use it. But it will still take you years of experience to use it successfully and to develop your own style

[BlowFish 10]

Darth, from your last couple of posts it seems you have missed a couple of the key premises that Jerry's stuff is built on. There is no assumption about the distribution being normal, quite the contrary. In fact Jerry relies on scewedness (asymmetry in the distribution) to define trend. Theres a couple of really neat ideas presented quite elgantly, it's certainly worth a review.

[davla 10]

could you pls provide me with a .txt file so I could code into my system?

thanks

[darthtrader2.0 10]

No darth, you are incorrect. The one std dev is for whatever the distribution is when you make the trade.

 

No, both you and blowfish are missing a very naunced point to this whole thread of ideas. That problem is in the std dev calculation being put forth here, because...the above statement acts as if the distribution has no effect on variance...comeon.

Well if you break it down, if you "deviate" in a "standard" way, then you are calculating a variance away from a central mean.

I agree, VWAP is probly a great way to model the central tendancy of an unknowable distribution.

If your deviations though are semetric from that mean?

What else could that possibly mean other than that your backtrading a normal distribution? Which anyone with half a brain knows is bogus with the markets.

This strikes me as stats 101 vs probability theory...I'm self taught though , maybe I'm way off.

[jperl 11]

the above statement acts as if the distribution has no effect on variance

 

Not sure what you mean here. Of course the distribution has an effect on the variance. The variance is computed from it. As the distribution changes with time, so does the variance.

 

I agree, VWAP is probly a great way to model the central tendancy of an unknowable distribution.

 

Whew!, I was getting worried that you wouldn't at least agree that we can define an average for a finite data set. Keep in mind though that this average is dynamic and constantly changing as more data is added. (It may in fact be totally undefined for the infinite population, but that is another story)

 

 

If your deviations though are semetric from that mean?

What else could that possibly mean other than that your backtrading a normal distribution?

 

If I interpret what you are implying here, it is that since I compute the standard deviation about the mean symmetrically, that it suggests that the distribution is symmetric about the mean.

This is not the case. Keep in mind that the variance is a positive number by definition of the second moment, but that standard deviation is either positive or negative since it is the square root of the variance. The fact that the standard deviation can have either sign implies nothing about the shape of the distribution. Perhaps it's the word standard that you don't like, but that is the general usage for any distribution normal or not.

[BlowFish 10]

It is probably worth making a couple of points. A distribution does not have to be Gaussian to have a valid variance. The squares (for the variance calculation) are weighted in the same fashion (by volume) as the VWAP.

[steve46 12]

I wonder folks, if you (any of you) have an understanding of the practical context of VWAP?

 

I think I understand (now) the basis for the use of it in Jerry's system, however the way "the VWAP" affects trading could perhaps be important depending on what you are trying to do (your time frame, your market bias long or short, and your position sizing).

 

For those who may not know, VWAP is used as a bogey for institutions who generally speaking are trading size (blocks from 25k on up). These institutions execute both in the market (you see them on your DOM) and out of market (in the liquidity pools, and premarket in response to "indications of interest" communicated between principals). Depending on the bias long or short, the party executing the transaction is trying to get it done either above or below the average for that time frame. In most all cases the firm executing is a third party who has to report whether they had success hitting that bogey (target) or not. This is important to them, because often the transaction is only part of a larger piece of business that needs to be done over an extended time frame. If the executing firm has success hitting that target (or bettering it) they may often "win the order" to execute the rest of the position. If for example they are not successful in executing at the VWAP, they probably won't see any more business from that order.

 

So, on a given day, what matters is whether institutions and size players are trying to get long or short (trying to execute at, above or below the VWAP). As with most things in life, it isn't black or white, so what matters is what the majority of players are doing, by chance or by design.

 

So what I am thinking is this....on a given day, if a majority of players are SELLING inventory, they are going to be trying to execute at or above the VWAP, and your systematic approach should probably be "tuned" to that reality

 

If on the other hand, the majority of players are BUYING inventory, people trading this particular systematic approach are going to want to tweak it to the other side a bit.

 

Now, I do this myself, and it doesn't require that you trade using Jerry's exact approach. It simply means that you keep in mind that there is likely to be a tendency for size to transact on one side or the other of the VWAP, and the reality is that programmed execution is going to be coming in on a timed basis, trying to execute size without moving the market (executing in pieces). If your market is an open outcry market, they are looking for evidence that players are coming back with more to business to do. Once they see that, well thats part of what causes markets to trend. if you have a way to determine this early in the session, you have a significant edge.

 

Hope this helps a little bit

 

Steve

[BlowFish 10]

Good point Steve. I think it was made some where on one of the threads but it is worth remembering:)

 

I also believe it can be used as a measure of trader performance working a large order? I guess you would compare the VWAP of that traders trades against that of the general market? The bigger the difference the better the trader providing he is the right side of course!

 

Glad you 'got' Jerrys system. Personally I think its both novel (very rare) and elegant, great stuff.

[darthtrader3.0beta 10]

Not sure what you mean here. Of course the distribution has an effect on the variance. The variance is computed from it. As the distribution changes with time, so does the variance.

 

Whew!, I was getting worried that you wouldn't at least agree that we can define an average for a finite data set. Keep in mind though that this average is dynamic and constantly changing as more data is added. (It may in fact be totally undefined for the infinite population, but that is another story)

 

Well I want to say again, I know I'm knitpicking here and not saying you are not a master trader. I am though highly influenced by the concepts in Evidence Based Technical Analysis, so I start from the idea that whatever I hear does not work at face value and move on from there. From my experience this system probly works for you as a way to filter out trades, and the real "meat" is in your understanding of price action...thats besides the point though. My main point is this is not the optimal way to calculate things...

" Of course the distribution has an effect on the variance. The variance is computed from it."

Of course the variance that is computing your std dev bands is based on the distribution that your assuming in your variance calculations...

How can you make that calculation without assuming a distribution? My point is YOU CANT. So your assuming a normal distribution to make your variance calculations, which we fundamentally know is far from optimal. The fact that its good enough for you to trade on does no mean its optimal.

 

"we can define an average for a finite data set. "

 

of course we can define a mean for a finite data set..the problem is though you have to define a distribution to calculate the variance from that mean...hence you have to quantify a distribution in order to calculate std dev bands from the VWAP...you currently define those bands assuming a normal distribution...which we know is wrong if we know anything. Personally, I would think at least a poison distribution would be a better assumption, but its interesting to think about what is far more optimal.

 

Also, steve46 point has no point in this discussion. Of course the VWAP is traditionally used an institutional bench mark and/or execution algorithm, thats where the liquidity is if your "big"....I imagine the VWAP is far more meaningless with futures directly as futures are a way to hedge those VWAP taken equity positions by the big guys...This is a discussion of mean and variance..

Thats why I said it could be interesting to compute a underlie vwap with something like neoticker for the SPX then throw it over ES.

[darthtrader3.0beta 10]

So what I am thinking is this....on a given day, if a majority of players are SELLING inventory, they are going to be trying to execute at or above the VWAP, and your systematic approach should probably be "tuned" to that reality

 

The problem with this is in the NYSE breakdown of volume statistics...50% "programs", 40% "institutional", 10% or less "retail"...since no one is really running "programs" at the "retail" level the volume is 90% "the big guys" vs the 10% retail...

If 90% of volume were truely trying to execute mean reversion wise at the VWAP then we would not see the trends we currently see.The harsh reality is that the market is made up of the big guys trying to front run the big guys trading mean reversion and catch them at the wrong time...who on the other side of the trade are trying to front run "the big guys"...Thats why to me if you try to use the VWAP outside of being a tick precise "mean" then you are missing the point of its value...

This is exactly why I don't post in the "VSA" threads, its just bogus...a misunderstanding of the "enemy" at hand...

[darthtrader3.0beta 10]

Glad you 'got' Jerrys system. Personally I think its both novel (very rare) and elegant, great stuff.

 

All I know is this series of threads led me to this book, and this book made me question this entire thread...there is no reason to not follow this path...

 

Probability Theory: The Logic Of Science

 

An interesting sidenote you will find in that book is that Jaynes believes the entire idea of a "stochastic process" is bogus...interesting when you look at the current problems and how much was based on financial engineering of treating markets as "stochastic process"...

[sevensa 10]

All I know is this series of threads led me to this book, and this book made me question this entire thread...there is no reason to not follow this path...

 

While you are questioning other methods based on your believes and books you are reading, people who are using these methods are making money.

 

I rather let my account tell me my method is profitable instead of a book telling me I should not be. If you rather read books and nit-pick about how things really should be calculated...for each their own.

[BlowFish 10]

From my experience this system probly works for you as a way to filter out trades, and the real "meat" is in your understanding of price action..

 

This can be probably be said about any trading method, I remember having an aha moment many years ago along those lines. Having said that there are clearly 'better' areas to look for trades than others.

 

 

 

 

Of course the variance that is computing your std dev bands is based on the distribution that your assuming in your variance calculations...

How can you make that calculation without assuming a distribution? My point is YOU CANT. So your assuming a normal distribution to make your variance calculations, which we fundamentally know is far from optimal. The fact that its good enough for you to trade on does no mean its optimal.

 

"we can define an average for a finite data set. "

 

of course we can define a mean for a finite data set..the problem is though you have to define a distribution to calculate the variance from that mean...hence you have to quantify a distribution in order to calculate std dev bands from the VWAP...you currently define those bands assuming a normal distribution...which we know is wrong if we know anything. Personally, I would think at least a poison distribution would be a better assumption, but its interesting to think about what is far more optimal.

 

My knowledge of stats isn't great (i have Jerry to thank for diving a bit deeper) but no no no no. You absolutely CAN calculate the variance without knowing the distribution. The whole point of calculating variance is to tell you something about the distribution of the sample! (the amount of dispersion).

 

I think you might be getting confused, because of some of the often quoted stuff about normal distributions. Specifically if you do have a normal distribution then 68% of the samples will fall inside 1 standard deviation. Therefore people sometimes cheat and draw bands at 68 or 70%. I suppose I should say 'use a heuristic method' rather than cheat. Jerry makes no such assumption. He calculates the sum of the weighted (thats the really clever bit) squares and this tells him about how the data is dispersed.

 

The only assumption that is made is that the data set is one that you can actually calculate the variance for which it clearly is. (e.g. some data sets diverge from the mean and/or tend towards infinity).

 

I am sure Jerry (or someone) will correct me if I am wrong.

[Head2k 10]

Darth, you don't have to assume any distribution. You simply know it. You know the data so you know the distribution. And statistical quantities such as mean, mode or standard deviation are defined and can be calculated for any distribution, without any assumptions about its character.

I quite studied Jerry's threads and I came to the following conclusions:

Organising data with VWAP and SDs is simply a way to find potential S/R and to define trend. Hence it is a tool which imposes structure on price/volume data. It is not a method (took me a while to realize it). A method needs to be built on understanding PA within this structure.