Trading with Market Statistics VIII. Counter Trend Trades in Symmetric Distributions

[jperl]

	Quote:	dbntina
	One more question.... The VWAP/PVP relationship determines the type of distribution (skewed/symmetric). Ignoring for the moment that as price moves around and volume trades that the VWAP and PVP change and the distribution will probably change. Let's just say we were pretty sure that the type of distribution was not going to change for the next hour. The point I am trying to clear up is that price in relationship to VWAP determines trend. I don't think it does. I think the distribution function itself tells you the trend bias and strength based on the strength of the skew. Who cares where prices are in relationship to the VWAP. I am thinking in terms of statistics and occurences in relationship to the disribution function.

You are quite correct for static distribution functions. Note the emphasis on the word static. If you knew the distribution function in advance, you would know exactly how to trade it.
As you correctly point out, if the distribution had positive skew, you would go long every time the price action dropped below the PVP and vice versa for negative skew. Similarly for no skew distribution, always trade toward the VWAP. This is the classical reversion to the mean theory.
The problem as you realize, is that in a real market, the distribution function is dynamic. Note the emphasis on the word dynamic. Reversion to the mean, does not necessarily occur. Price action itself renormalizes the distribution as more prices are added to the time series. As a consequence, the relation of the price action to the VWAP and PVP becomes critical in choosing trade direction. Trading AWAY from the VWAP then becomes the more likely scenario. Trading TOWARD the VWAP is then relegated to symmetric distributions, which is the classical case.
Trading away from the VWAP then defines the trend (if in fact you want or need a definition)as UP when the price action is above the VWAP and DOWN when the price action is below the VWAP. This definition breaks down of course when the distribution passes through the symmetric state. At that point there is no trend.

[dbntina]

Jerry,

Thanks a bunch...makes sense that the fact that it is dynamic it changes how you have to approach the distribution.

Very interesting and it is starting to come together.

Thanks for sharing,

dbntina

[emmster]

Thanks to jperl for his generous insights, and dbntna and BlowFish for their work on theTradeStation code to see what is being discussed on the chart (received a copy from a trader on this board).

Sorry for being dense on this, but I am a bit confused on the trade direction when viewing the Price vis-a-vis the VWAP and the PVP.

My understanding is that the skew is in the direction of the VWAP (so VWAP > PVP means skew is up).

When VWAP > PVP but current price < VWAP, is there a directional bias?

Thanks for the threads and any clarification.

[jperl]

	Quote:	emmster
	My understanding is that the skew is in the direction of the VWAP (so VWAP > PVP means skew is up). When VWAP > PVP but current price < VWAP, is there a directional bias?

The answer to your question emmster is a definite maybe.
If the price action stays below the VWAP and breaks out below the 1st SD, then the bias would be down. But until that happens, all you will get are oscillations between the VWAP and the 1st SD below it. See the breakout thread Part VII for further clarification.

[BlowFish]

Emmster despite following carefully at home I discovered something that I had missed. Skew and Trend can be quite different. I am not sure Jerry introduced the term 'bias' I think he has always talked about skew until others (like me ) mentioned it. Early on I chose to think of skew as a 'bias'. On reflection maybe that was not such a good idea.

Might be easier to just think of skew (more volume above or below the PvP) i.e the distribution is skewed. And trend i.e. price is trading above or below the VWAP.

In the basic trades skew and trend match things become trickier when they don't.

Caveat - I'm still learning this stuff too so I hope I haven't given duff info. I am sure Jerry will correct me if I am mistaken.

[emmster]

Thank you for the responses Jerry and BlowFish. I can see that I must be careful in the terminology I use because I may be introducing my own mental "bias" by using the wrong wording in my head when reviewing the posts.

In ES Thurs, I see where skew and trend match, producing very nice outcomes.

The idea of a breakout against the skew is also clearer on review and Jerry's comment above to dbntina - "the distribution function is dynamic. Note the emphasis on the word dynamic. Reversion to the mean, does not necessarily occur."

(Lightbulb starting to flicker on....)

[steve46]

1.) Why not look to a longer time frame when trying to determine how long a distribution will remain symetric. I am thinking that following the skew on a daily and/or weekly basis might add value to the process.

2.) I have started to look for, and catalog "HUP's". Based on Jerry's characterization of them as places where price is held up while the market decides direction, I am keeping a list not only of the price but of the time (time and duration) to see if there is some utility to be had from that.

[jperl]

	Quote:	steve46
	1.) Why not look to a longer time frame when trying to determine how long a distribution will remain symetric. I am thinking that following the skew on a daily and/or weekly basis might add value to the process. 2.) I have started to look for, and catalog "HUP's". Based on Jerry's characterization of them as places where price is held up while the market decides direction, I am keeping a list not only of the price but of the time (time and duration) to see if there is some utility to be had from that.

Good points Steve. These two issues are related. I was going to do a thread on HUP but discovered that the presentation was too complicated. I'm still searching for a simple way to start it.

[Dogpile]

jerry,

was curious your thoughts on yesterdays volume distribution? it looked like a near perfect symmetrical distribution for much of the day then ended up building a fat lower tail. In situation like this, where you would expect to go long low in a symmetrical distribution (as it looked with an hour or two to go) -- but the distribution did not END UP symmetrical. curious your take?

thx

[jperl]

	Quote:	Dogpile
	jerry, was curious your thoughts on yesterdays volume distribution? it looked like a near perfect symmetrical distribution for much of the day then ended up building a fat lower tail. In situation like this, where you would expect to go long low in a symmetrical distribution (as it looked with an hour or two to go) -- but the distribution did not END UP symmetrical. curious your take? thx

Sorry for the delay in getting back to you Dogpile. Just got back from China.

I took a look at the Sept 20 data for ES. Here is my take on it:

In the first chart you can see that the skew is positive until 14:20 PM. So I would have been biased for breakout trades to the upside. There were quite a number of these that would have been good trades to the 1st and 2nd SD. Breakout trades to the downside would have all failed until 12:28 PM when the market broke out to the downside.
It was only at 14:20 that the market became symmetric (PVP=VWAP at the last bar shown in the the first chart).
As a trader you then had to decide whether to take an upside move back to the VWAP. Here is where the Shapiro Effect comes in handy.
In the second chart which shows the whole day, there was no Shapiro effect to the upside after 14:20. In fact all the Shapiro effects occurred to the downside with good trade entries at 14:44, 14:58, 15:30 and 15:56. By 14:44 the skew went negative, so I would have only looked for short entries anyway.
Too bad I had already left for China. It would have been a great trading day.
JERRY