Trading with Market Statistics. IV Standard Deviation

[Head2k 10]

  metalhead said:

I'm afraid I don't know how to get VMP on my charts

I am afraid you cannot get it there. Or you can, but you cannot base a computation on it. The term Volume Weighted Price I used only to show the relation to Volume Weighted Average Price. The VWP is the Volume Distribution Function, as Jerry calls it. Somebody calls it Volume At Price chart or Volume Profile. I believe Sierra has such a chart but I doubt you can somehow "read" the length of the line representing the volume at each price. That's why you probably cannot base a computation on it. I am afraid you will have to code it. But again, I don't know Sierra at all.

[darthtrader2.0 10]

  jperl said:

We have so far analyzed the distribution in terms of two properties, a)the peak volume price ( PVP ) and b) the volume weighted average price ( VWAP ), which is the mean for the distribution. Both of these are dynamically updated throughout the trading day as the volume distribution function dyanmically changes. In [thread=2008]Part III[/thread], we showed how the relationship between the VWAP and the PVP could be used for an entry technique in a simple newbie VWAP trading strategy.

 

While I will probly forever respect this thread on here for motivating me to be non igorant of statistics and probability, from what I've learned in the past year most these intial posts are pure frequentists nonsense with holes you can drive a truck through.

Essentially, this entire series puts forth a very scientism view that between the high and the low of any given session, one can use a probability density function based off how volume is distributed dependant on price,an idea that already assumes a normal distribution to the way that fundamentally, we already know that each trade is a stochastic process, and the sum of that stochastic process has nothing to do with a normal distribution as far as price or number of trades at a certain price goes. Let alone as far as the stochastic process that each of the participants information that they base their decisions on as far as a single unit of trade, just more NULL here.

What I would suspect is that this system works for you absolutely because of the unknown variable of your understanding of price action and that all this nonsense, IS your subjective bayesian filter for your OWN understanding of "price action", it certainly has nothing to do with the bunk science of frequentist statistics who can not predict shit.

 

Touché

Edited September 5, 2008 by darthtrader2.0

[BlowFish 10]

  darthtrader2.0 said:

 

What I would suspect is that this system works for you absolutely because of the unknown variable of your understanding of price action and that all this nonsense, IS your subjective bayesian filter for your OWN understanding of "price action",

 

This is an interesting idea. I have always suspected that people who trade from 'lines' be they had drawn S/R MP fib that success or failure is determine not by the lines but by the traders ability to read PA and how they manage trades. Mind you Jerry has covered both of these topics in his threads.

 

 

Cheers.

[Head2k 10]

  BlowFish said:

I have always suspected that people who trade from 'lines' be they had drawn S/R MP fib that success or failure is determine not by the lines but by the traders ability to read PA and how they manage trades.

I came to this suspicion too.

[jperl 11]

  BlowFish said:

This is an interesting idea. I have always suspected that people who trade from 'lines' be they had drawn S/R MP fib that success or failure is determine not by the lines but by the traders ability to read PA and how they manage trades. Mind you Jerry has covered both of these topics in his threads.

 

 

Cheers.

 

This is essentially correct Blowfish. Your success as a trader depends more on how you manage a trade then on what methods you use to enter a trade.

 

Darths comments however are too extreme. I will comment below

[jperl 11]

  darthtrader2.0 said:

... I will probly forever respect this thread on here for motivating me to be non igorant of statistics and probability

 

Darth, I'm glad that you were motivated to learn something about statistics. It's a difficult subject.

 

 

  darthtrader2.0 said:

from what I've learned in the past year most these intial posts are pure frequentists nonsense with holes you can drive a truck through.

I know you can drive something through, I don't think a truck. But perhaps you would like to tell us what you learned in the past year that gives you this feeling.

  darthtrader2.0 said:

Essentially, this entire series puts forth a very scientism view that between the high and the low of any given session, one can use a probability density function based off how volume is distributed dependant on price,an idea that already assumes a normal distribution to the way that fundamentally, we already know that each trade is a stochastic process, and the sum of that stochastic process has nothing to do with a normal distribution as far as price or number of trades at a certain price goes.

It's tough to comment on so much in one discombobulated sentence. But you are forgiven considering you posted it at 3 am in the morning.

But here is my take on it. THERE IS NOTHING THAT I PRESENTED THAT ASSUMES A NORMAL DISTRIBUTION. One does look at a finite sample population and that sample population has a mean, a standard deviation, a skew and a kurtosis. Those are the first four moments for the distribution.

This implies nothing about the distribution being Gaussian(ie normal). Every finite distribution has these properties. These moments can and do change with additional data added to the population. I think I have shown this numerous times in the videos.

How you might use this information depends on your trading style. I presented some simple ways as example, but there are many others.

The question you might raise is: Is this statistics the best way to view the data or are there other ways. At the moment this is the best I know. There are other alternatives such as fractal analysis a la Mandelbrot or perhaps some kind of Bayesian analysis.

 

  darthtrader2.0 said:

What I would suspect is that this system works for you absolutely because of the unknown variable of your understanding of price action

That is perfectly correct. But I couldn't understand price action without knowing what the volatility looks like, how the skew is changing and where the mean is at the moment.

 

  darthtrader2.0 said:

and that all this nonsense, IS your subjective bayesian filter for your OWN understanding of "price action", it certainly has nothing to do with the bunk science of frequentist statistics who can not predict shit.

I won't comment on this since it requires a whole new thread for discussion. The debate between the frequentists and the bayesians has been around for ever. Your comment doesn't add anything to the debate.

[Head2k 10]

  jperl said:

THERE IS NOTHING THAT I PRESENTED THAT ASSUMES A NORMAL DISTRIBUTION.

IMO there is something. Using skew to detrmine direction of a trade you actually assume that the skewed distribution should converge to the normal one.

 

I came to the same conclusion as Blowfish when I papertraded your system. IMHO skew alone is not a good way to determine direction. And I wasnt able to properly judge price action, so I left the paper trading and now I am studying price action.

Yet I definitely wouldnt be as harsh as Darth, these threads learned me a lot of useful things and I appreciate the effort you put into them. But I came to a conclusion that the only system that really works for somebody is the system that the person develops on his/her own.

But I believe that these threads can serve as a base for development of a system, better base than most.

[jperl 11]

  Head2k said:

IMO there is something. Using skew to detrmine direction of a trade you actually assume that the skewed distribution should converge to the normal one.

Sorry Head, but this is incorrect. A skewed distribution MAY return to a SYMMETRIC distribution, which is not the normal one. Notice my emphasis on MAY and SYMMETRIC. Using the exact definition of skew which you presented, a skew of zero does not imply anything about the distribution being normal, only symmetric. Example would be a double peaked distribution which is far from normal. There are many others. Moreover, there are many days when the skew never returns to zero.

 

  Head2k said:

I came to the same conclusion as Blowfish when I papertraded your system. IMHO skew alone is not a good way to determine direction. And I wasnt able to properly judge price action, so I left the paper trading and now I am studying price action.

I learned something new from you HEAD when you presented a way of looking at the skew using the exact definition of the third moment. There is more there than I think you realize, certainly more than I originally realized using the approximate definition. One of the most difficult problems traders have commented on to me using market statisitics is how to decide which direction the market will go once it touches a standard deviation point. Knowledge of skew action can help in this regard. Perhaps I will start a new thread on this if I can find the time.

 

 

  Head2k said:

Yet I definitely wouldnt be as harsh as Darth, these threads learned me a lot of useful things and I appreciate the effort you put into them. But I came to a conclusion that the only system that really works for somebody is the system that the person develops on his/her own

But I believe that these threads can serve as a base for development of a system, better base than most.

 

I agree with this. I hope I haven't mislead anyone into thinking that what I presented is a "system" for trading. It is rather a way of looking at the data from which you have to develop your own trading style. I presented some simple ways that you might use market statistics to trade, but in the final analysis every trader must develop his/her own style.

[Head2k 10]

  jperl said:

Sorry Head, but this is incorrect ... Using the exact definition of skew which you presented, a skew of zero does not imply anything about the distribution being normal, only symmetric.

Yes, my fault. You are correct.

 

And I am glad I was able to offer you some new insight, though I probably dont understand its full meaning

Edited September 6, 2008 by Head2k

[BlowFish 10]

  Head2k said:

 

I came to the same conclusion as Blowfish when I papertraded your system. IMHO skew alone is not a good way to determine direction. And I wasnt able to properly judge price action, so I left the paper trading and now I am studying price action.

 

Always nice to see some 'action' on these threads. I think I might have given the wrong impression. One of the issues facing all traders (regardless of methodology) is not so much determining direction but determining 'break out' or 'retrace'. Different methodologies have different terms for this type of action. One way Jerry deals with this is wait for the event to occur and then use the shapiro effect (price action) to trigger a trade.

 

Since we last spoke I have paper traded again for a few weeks this was a couple of months ago. I had great results percentage wise however I did have rather too many stops where I bought 1 at SD1 another at VWAP and stoped at pvp. Coupled with the fact that I am prone to close early (if the price action 'shakes me') I found those a little hard to deal with. Thats another story.

 

I do wonder if the PVP is the best place for a stop? Of the maybe 3 out of 10 of these trades that hit the PVP I would guess at least half just prod it by a tick or 3. There are two alternatives I can see. Either trade with a wider stop around the opposite SD band or trade with a tighter stop behind the shapiro bar. Probably the only time to take an SD1 trade (with scale at VWAP) is when there is a 'clearly' established trend.

 

As for Pearsons approximation for skew being a good proxy for the real one. I am unsure of that. There are certain circumstances (which I talked about early in the series and Jerry dealt with in the break out section) where they are clearly not. And that leads me back to where I started ascertaining if we are breaking out or returning to balance/accumulating/rotating/range bound.

 

Cheers.

[jperl 11]

  BlowFish said:

 

Since we last spoke I have paper traded again for a few weeks this was a couple of months ago. I had great results percentage wise however I did have rather too many stops where I bought 1 at SD1 another at VWAP and stoped at pvp. Coupled with the fact that I am prone to close early (if the price action 'shakes me') I found those a little hard to deal with. Thats another story.

 

I do wonder if the PVP is the best place for a stop? Of the maybe 3 out of 10 of these trades that hit the PVP I would guess at least half just prod it by a tick or 3. There are two alternatives I can see. Either trade with a wider stop around the opposite SD band or trade with a tighter stop behind the shapiro bar. Probably the only time to take an SD1 trade (with scale at VWAP) is when there is a 'clearly' established trend.

If you scaled in at VWAP and there was no retrace back toward break even, it suggests the market is strongly moving against you. You might want to condsider a reversal trade provided you are still within your risk tolerance for the day.

 

  BlowFish said:

As for Pearsons approximation for skew being a good proxy for the real one. I am unsure of that. There are certain circumstances (which I talked about early in the series and Jerry dealt with in the break out section) where they are clearly not. And that leads me back to where I started ascertaining if we are breaking out or returning to balance/accumulating/rotating/range bound.

 

I understand your dilemma. I introduced the approximate skew because it was easy to visualize, and required no further computation on the part of the trader. However since Head did the complete computation of the third moment, I took a closer look and it seems like that may be a better alternative.

 

[darthtrader2.0 10]

  jperl said:

It's tough to comment on so much in one discombobulated sentence. But you are forgiven considering you posted it at 3 am in the morning.

But here is my take on it. THERE IS NOTHING THAT I PRESENTED THAT ASSUMES A NORMAL DISTRIBUTION.

 

Well its 2:53am and I'm drunk...maybe I can be clearer this time.

I take issue that you are not assuming a gaussian distribution...what distribution in your code is the code deviating from in this thread if not gaussian?

When it gets down to it there is a far more fundamental problem here that has to do with a fundamental and not computable moving nash equilibrium price intraday and the probability distributions above/below that equilibrium in order to quantify a bet. You can quantify all day on a random variable, however if you are not quantifying distributions on the "correct" variable, I don't see how that differs from straight out single agent/brain/heuristic "guessing".

Edited September 12, 2008 by darthtrader2.0

[Head2k 10]

  darthtrader2.0 said:

Well its 2:53am and I'm drunk...maybe I can be clearer this time.

I take issue that you are not assuming a gaussian distribution...what distribution in your code is the code deviating from in this thread if not gaussian?

When it gets down to it there is a far more fundamental problem here that has to do with a fundamental and not computable moving nash equilibrium price intraday and the probability distributions above/below that equilibrium in order to quantify a bet. You can quantify all day on a random variable, however if you are not quantifying distributions on the "correct" variable, I don't see how that differs from straight out single agent/brain/heuristic "guessing".

It is 10:38 am here and I am sober. Taking trades in direction of skew (at certain area) assumes that skewed distribution tends to apprach a symetric state.

Now to quantifying a day. What VWAP and SD's do is not particulary quantifying a day, but rather presenting some structure to the daily data. I guess that if you want to trade you need to organize the data you trade so it shows some structure. You need to define a trend, where you are in that trend and what is potential support and resistance. There are more ways to do that. Drawing trend channels and S/R lines is a way. Plotting VWAP and SD's is another. Wathing the Volume Distribution tells you also something about the inner dynamics of that day.

If nothing else, VWAP and its SD's can serve as an objective reference or confirmation of the daily trend and to where we are in that trend.

 

My 0.02

[darthtrader2.0 10]

  Head2k said:

If nothing else, VWAP and its SD's can serve as an objective reference or confirmation of the daily trend and to where we are in that trend.

 

My 0.02

 

As are fibonacci nonsense or a random number generator "predicting" support and resistance. Who cares what you call it if the end result is no better than random. Noise is a pretty good term.

If you want to take statistical assumptions off the table there is really a more fundamental problem with assumptions of the VWAP of futures here:

Who is hedging other bets risk wise vs who is taking outright positions to add to risk is unknowable??

 

You can throw all the statistics you want at the unknown, yet even if I can find a mathematical relationship that defines that I can see a dog in the clouds, no matter how abstract and probabilistically informed it is, does not mean an actual dog is in the clouds. Betting on what a fake dog in the clouds is going to do is a clear logical absurdity.

You can not seperate the two variables there so we are basically both talking about horseshit, that has little to do with what we are actually trying to argue about.

The difference is, I at least understand this perspective, while you do not.

Maybe this perspective is wrong, maybe it is not.

[jperl 11]

  darthtrader2.0 said:

I take issue that you are not assuming a gaussian distribution...what distribution in your code is the code deviating from in this thread if not gaussian?

 

The deviation is from a symmetric distribution which can have any shape.

You seem to have a fixation that averages only have statistical significance if the distribution is gaussian. Nothing could be further from the truth. You can always define an average for a finite sampling of data. The question that you should be asking is, is the average stable or does it change if you take a different sampling of data?

  darthtrader2.0 said:

If you want to take statistical assumptions off the table there is really a more fundamental problem with assumptions of the VWAP of futures here:

Who is hedging other bets risk wise vs who is taking outright positions to add to risk is unknowable??

Yes, and that's why you want to use statistical data. There are too many variables to ever possibly know anything deterministic about a single trading event. If you want to argue that the price data is not the proper variable to use in the statistical analysis, then you are a believer in the existence of hidden variables. Until you define what those variables are however, you don't have a valid argument against using statistical analysis.

[steve46 12]

Jerry, what would your response be to the suggestion that any distribution other than normal or gaussian is ultimately unstable? Referring only to stock market price data of course.

[jperl 11]

  steve46 said:

Jerry, what would your response be to the suggestion that any distribution other than normal or gaussian is ultimately unstable? Referring only to stock market price data of course.

 

Based on the history of all financial markets, they are all unstable. This includes normal distributions which don't remain that way for very long. This does not mean however that you as a trader should not use the evolving statistics as a framework for deciding both trade direction and trade management. On the contrary, without the statistical information you are depriving yourself of what the market's price action might be like. 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.

[steve46 12]

Is your use of Skew and VWAP the only solution (that you know of) to having to deal with unstable distributions?

Edited September 22, 2008 by steve46

[jperl 11]

  steve46 said:

Is your use of Skew and VWAP the only solution (that you know of) to having to deal with unstable distributions?

 

I don't know. There is considerable research going on in this field associated with unstable time series evolution. Examples would be fractal analysis and other non-linear dynamic analysis (chaos theory and complexity theory). I'm in the process of looking at these in some detail, but haven't reached any conclusions yet in so far as there use for traders.

[darthtrader2.0 10]

  jperl said:

I don't know. There is considerable research going on in this field associated with unstable time series evolution. Examples would be fractal analysis and other non-linear dynamic analysis (chaos theory and complexity theory). I'm in the process of looking at these in some detail, but haven't reached any conclusions yet in so far as there use for traders.

 

Mandelbrot found his patterns in the clouds in futures markets yet no one has been able to make pragmatic sense of his ideas since.

If you really want to get into modern reseach your best bet is to include either the words "stochastic" and/or "bayesian" in your search, better yet both.

"game theory" is also a good key phrase but we are not even close to computing power needed to game markets, but its still an interesting idea.

[Pepperdog 10]

I've been reading and studying these threads with great interest. Thank you Jerry, and everyone else for new perspectives I have on price action. I think I understand most of the trade setups. I've also noticed some very interesting things lately by watching how price acts around the VWAP and +/- standard deviation lines, independently of PVP.

 

One thing I've seen lately, that I am not clear on, is that price can be trading below (or above) both the VWAP and the PVP while the skew is opposite of the direction price is moving. Of course, over time the VWAP should eventually "catch up" as the price continues to move away, causing the skew to "confirm" the direction of the market. I guess this situation would be considered a break-out, the more advanced trades? From my somewhat basic understanding of the trade setups, I would anticipate waiting for the skew to "catch up" with the direction price is moving for the "newbie" trades but often it seems that is after *much* of the move has already happened -- it takes a while for my VWAP/skew to "catch up" (assuming that my calculation of VWAP and PVP is correct which I am pretty certain of.)

 

I'll see if I can come up with some screenshot examples, but I've seen this happen quite a lot over the last few trading days on the ES with the massive selloff we've been seeing. So maybe these recent days aren't really great representative samples!

 

Can anyone comment or clarify?

 

Thanks,

J

[jperl 11]

  Pepperdog said:

One thing I've seen lately, that I am not clear on, is that price can be trading below (or above) both the VWAP and the PVP while the skew is opposite of the direction price is moving. Of course, over time the VWAP should eventually "catch up" as the price continues to move away, causing the skew to "confirm" the direction of the market. I guess this situation would be considered a break-out, the more advanced trades?

You have this correct Pepperdog. A newbie trader would just have to sit and watch, while the market moves against the skew. After you feel comfortable trading in the skew direction, you can become a more advanced trader and trade against the skew as well. This is discussed in some detail in the later threads on Market Statistics. The skew then no longer becomes a relevant factor in your trading. However be careful here. You need to know that you are doing this and what it implies. Trading with the skew is the teaching mode. Trading against the skew is where the real action lies. You can't do the latter unless you're comfortable with the former.

[Pepperdog 10]

Thanks Jerry, I'll go back and review the later threads a 3rd, even 4th time!

 

Up until the last few months, my trading has been pretty newbie-ish, and I really appreciate the time you've spent on these threads. With these new perspectives and tools I am very certain I will be taking my newbie self to the next level.

 

One thing I am seeing over and over, regardless of the skew, is the vwap and SD lines will repeatedly act as major or minor (failed) support and resistance on virtually every instrument I've been following. So it would seem to me that these levels serve as "decision points" for market forces beyond my comprehension...can anyone provide some insight?

 

Thanks,

J

[Head2k 10]

  Pepperdog said:

One thing I am seeing over and over, regardless of the skew, is the vwap and SD lines will repeatedly act as major or minor (failed) support and resistance on virtually every instrument I've been following. So it would seem to me that these levels serve as "decision points" for market forces beyond my comprehension...can anyone provide some insight?

 

Thanks,

J

 

VWAP is Volume Weighted Average Price. So what does it represent? It represents the average price of all contracts traded since the beginning of VWAP computation. For traders who entered a position since then it represents the average entry price or, in another words, an average breakeven. 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. Now lets say the price gets to the VWAP. Both buyer and seller are now at breakeven. Seller had paper profit before that, so it is likely that he is not going to cover at breakeven, because he had a chance to cover in profit and he didnt, so he believes even in more profit. And even if he doesnt believe in more profit, he wants at least that profit he could have had before.

Buyer sees it from different perspective. He was in a loss and now he has a chance to exit at breakeven. So in this case VWAP is a point where seller wants to buy the least and buyer wants to sell the most. Hence the price bounces back down.

Of course there is another scenario possible, and that is such a change of conditions or confidence that seller and buyer reassess their positions. Then VWAP is the last point where seller is not at loss so he will be eager to cover there, or if he is not confident he will place his stop there. On the other hand, if buyer is confident he will not exit at breakeven but will aim for some profits.

 

I am not quite sure how to interpret SD's in a similar manner though. When taking SD's in account you dont consider one composite trader anymore but different levels of "majority of traders", or better to say contracts.

[jperl 11]

  Pepperdog said:

One thing I am seeing over and over, regardless of the skew, is the vwap and SD lines will repeatedly act as major or minor (failed) support and resistance on virtually every instrument I've been following. So it would seem to me that these levels serve as "decision points" for market forces beyond my comprehension...can anyone provide some insight?

 

Head2k has given us a "simple" explanations of what is going on at the VWAP. I have heard these types of explanations before. Whether they will help you as a trader is another matter. Or perhaps they now will make you more comfortable in thinking about VWAP and the associated SD as you have described them, as "decision points for market forces" what I call hold up prices or dynamic HUP as described in thread XI of the Market Statistics threads. Once you draw the HUP lines in for the trading day, you have a powerful tool for understanding and dealing with the price action.