March 9, 2010

This week started off with the S&P breaking out above its previous 52 week high at 1177 and went on to continue its recovery rally ending this week’s session at a new high: 1193 in the S&P June futures: 1194 in the CASH Index.

Like everything in nature Markets change and evolve over time. As technology has come to play a greater role in market, computerized order execution has become more dominate.

This week price action appeared to be result of computerized order execution occurring in narrower range structures and for longer period of time than at any point in the S&P recent history. The result of this latest technological evolution has given rise to the most limited opportunity for individual S&P eMini traders than I have witnessed before.

The purpose of a market is to facilitate trade. In order to profit from participating in a market a trading range is required.

Periods of narrow range high density volume execution do not favor activity participation. The firms who executed trades using computer algorithms dominate the price action. The risk rewards ratio offers no opportunity to the individual participants.

In order to illustrate the difficulty in discriminating the data series during periods of high volume density computer generated order flow execution, I will begin with a description of Monday’s price action.

The machine code computerize algorithm execution began on Monday, as price traded up to a new high. The break-out above the previous 52 week high at 1177 was still within the parameters of normal market development. Then, in C-period price traded up to 1184 and encountered statistical high volume in the form of responsive selling. At that point the order flow structure executed trades within a narrow 3 point range between 1184 and 1181until the close of Monday’s session.

Monday’s data series provides us with examples statistical high volume density order flow execution wherein the BID ASK dynamic, i.e. the interaction between initiated buying and responsive selling, and or initiated selling and responsive buying offer no clear opportunity.

In other words, at the close of Monday session the S&P was neutral at the new high, in an extremely narrow range .The price action was so unusually that I noted in Monday’s blog, it appeared as if market markers were trading among themselves.

Tuesday: The S&P traded below Monday’s settlement, auction down to 1178, at or near the previous high. Buying interest was present at the open, the A-period low. Price traded back up into Monday’s narrow range cluster and from C-period thru I- period the same machine code computerize algorithm execution started again. In J-period price broke-out above the narrow range high and traded up to 1188. There was a minor pull-back to the point where the initiated buying first entered the order flow and price close the session at 1186.

Wednesday: The S&P sold below Tuesday settlement during the overnight Globex session. The opening range was between 1185 and 1181. Minor range extension occurred during C-period and price auction down to 1178 and found support at/or near Tuesday’s low. Price traded up to 1183 and went FLAT until H-period when price traded up to Tuesday’s settlement at 1186.

The lack of buying interest was evident and selling pressure entered the order flow driving price down to re-test the open range low at 1178. Selling pressure increased during L-period which witnessed a break-down below the open range low. Price traded down to 1173 and found support at the March 1^st settlement. A subsequent up tick off 1173 witness previous sellers cover their positions and price closed at 1178.

The development through Wednesday would lead an observer to interpret the price action at the Monday and Tuesday’s high as a form of commercial capping, i.e. establishing at upper limit.

However, Bayesian logic would argue that Wednesday’s pull-back was aloes the first test minor support at the March 1^st low, the starting point of the recent development that resulted in the newly made 52 week high. Therefore, if a subsequent probe back to Wednesday’s low 1173 found support conditional probability would favor a potential re-test of the high. The logic being: a pull-back to support following a new high off would represent the first opportunity for buyers to get aboard the developing trend.

Indeed, Thursday’s open re-tested the March 1^st low. There was minor selling pressure down to 1171 that meet with responsive buying.

All S&P sectors ended their initial decline after the 1^st hour of Thursday’s session. Buying interest increased and price rallied off the open range low and price traded back up to 1185.

During Friday’s session price traded up to Tuesday’s high during B-period. The opening range was 5 point: 1183 the low, 1185 the high. At that point, the same computerize algorithmic order execution witnessed earlier in the week once again dominated the auction.

The machine was back in control. The pattern of the order flow execution occurred within a 5 point distribution. The value area with 3 points and the order flow rotation was limited to 2 points.

The narrow trade range oscillated between initiated buying and responsive selling entering at high volume density. At the end of Friday’s session initiated buying dominated and price broke out above the upper limit during M-period and the S&P June futures contract closed the week at a new 52 Week high.

High volume density order execution, within a narrow range makes it difficult to visualize to the outcome of BID / ASK dynamic, i.e. the interaction between buyers and sellers. The observer’s ability to provide intuitive interpretation as to the directional outcome is limited.

High volume density order execution transacted within the limited parameters of a 5 point distribution limits the distance above and below the mean is 2.5 points. The rotations within 30 minute samples occurred within 2.5 point ranges.

The machines algorithms executed with precisions. Order flow shift were visually perceptible but manual execution against the machines algorithms was difficult. The risk reward ratios were limited.

The limited reward and lack of movement is the proto-typical definition of poor trade facilitation.

Description of this type of machine code has appeared in academic literature under the broad topic of Kernel Development. Advanced software application like Math-Lab and Mathamatica contain code for such equations in their library functions.

http://mathworld.wolfram.com/QuadraticEquation.html

http://mathworld.wolfram.com/KernelPolynomial.html
http://www.mathworks.com/access/helpdesk/help/toolbox/stats/f4218.html

In order to model the results of nonlinear movement within the order flow and to determine the structure of nonlinear kernel development a new visualizations method must be developed.

Articles on time series prediction algorithm based on least squares support vector machine languages have been published in the Journal of Central South University of Technology.

Many times you will come across equations that are not easy to factor or solve. In those cases, there is a special formula called the quadratic formulas that are used to solve such problems.

The price series that we have observed would be described as follows:

A series of many numbers: a polynomial
A complex data structures where the localize mean, the weighted function within the data points is difficult to estimate.

In theory, a kernel function could be used to represent the core density of the supply demand cluster:

Kernel functions play an increasingly important role in Machine Learning, Pattern Recognition and their applications.

http://www.springerlink.com/content/f07m327r45151535/

http://www.amazon.com/Pattern-Recognition-Learning-Information-Statistics/dp/0387310738

http://demonstrations.wolfram.com/EstimatingTheLocalMeanFunction/

http://mathworld.wolfram.com/K-MeansClusteringAlgorithm.html

http://mathworld.wolfram.com/NonparametricEstimation.html

A method for determining the forces of supply and demand curve is called an optimizing predictive controller.

Thus the code exits to construct a polynomial kernel function approximator that could be used for optimizing predictive accuracy of the shifts in order flow bias: BID to Offer/ Offer to BID supporting the micro structure.

Vector machine languages supports programming a polynomial kernel function approximator based nonlinear model. A support vector machine based predictive model of the high volume density order flow execution could be established.

This is obviously not the kind of thing anyone is going to program in Trade Station.

Even if such a model could be constructed, the question remains what would you do with it? For example; in the context of Monday’s price action the standard frequency distribution indicated a bias to the buy side. Indeed, the S&P closed the session at the high. Still once the machine algorithm was turned off and the afterhours Globex session started, price sold off from its high.

 During Friday’s session visual observation was sufficient to intuitively infer the late in the day breakout to the new high.

Never the less, in the light of the limited intraday opportunities current available in the S&P as a result of the high volume narrow range order execution algorithm, I have discussed with several members the benefits of looking at alternative markets.
I have been observing the intraday price action in the CME Euro [6E M0] over the past few weeks. The price action in the EURO appears to be a well defined trading range wherein market developments conform to profile analysis.

I have started to create a chart book of the EURO contract. My initial reach will include the standard drill down from the long term perspective to the current market condition. Soon I will start posting charts on the website. After the workshops I will host an afterhour’s open forum to discuss the research with IOAMT members.

Technical Reference Points for June S&P Futures Contract

Potential Next Up-side Trading Range

1221, September 23, 2008 High

1200, Monday’s projected high

Resistance

1193 June Futures, April 9^th new 52 Week High [1194 CASH]

Support

1173, March 7^th Low, minor

1165, March 31^st Close

1155, near term support March 26 Daily Low

1148, Major Support March 22nd low, the FOMC Low

1136, March 15th Daily Low

1134, March 9th initiated buying M-period, March 10th Daily Low, March 11th Daily Low

1125, the upper parameter of the 3-day distribution, March 2nd thru March 4th

1116, the lower parameter of the 3-day distribution, March 2nd thru March 4th

1105, minor, once tested, the February 26th settlement, overnight low

1086, the February 25th daily low

1078, February 16th low and the upper limit of weeklong consolidation range

1044 February 5^th Correction Low [Cash Index]

If the ideas and concepts of auction market theory appeal to you and you would like more information, you are invited to visit our website at www.IAOMT .com.