Monday, June 28, 2010

Identifying market regime shifts with CSSA

Structural breaks and non-stationarities are prevalent features of markets. The questions which occur most often when considering them is how to distinguish regimes and whether they can give rise to trading opportunities?

Apart education, this post is to draw your attention to the CSSA-Coupling Index. This indicator in the CSSA package uses the variance in price accounted by two individual cycles in a sliding window to return a degree of coupling between them.


In the example below, we show the CSSA-Coupling Index applied to the Dow Jones Industrial Average EOD (DJI: ^DJI).

The top panel shows price. The bottom panel shows the index. We have marked the tops of the index that occurred above the 0.2 level with red vertical lines.

The regimes in price could arise in a number of ways:
- Change in trend,
- Change in volatility…

Note that the marks delineate some of these regimes quite well. They mostly show early signs that the trend is about to change; peaks might even develop before a local change occurs. This is very interesting not so much because peaks signal that something is amiss with the current trend but because changes in volatility tend to be persistent in between meaning that markets can be seen as a series of regimes each of which is homogeneous.

This should appeal to traders who can use these early signals to reconsider their trading strategy after confirmation that a structural is about to occur.


Chart Example: Dow Jones Industrial Average EOD (DJI: ^DJI):
Top panel: Price, OHLC
Bottom panel: CSSA-Coupling Index (Close)
m-histories = 40 bars
Component#1 = 5
Component#2 = 2
Window = 100 bars

Wednesday, March 17, 2010

Synthesizing technical indicators with Noxa CSSA


Technical analysis has become ubiquitous in financial markets. Various technical indicators were developed over decades with supporting rationale, theory and mindset. In this post, we show that some indicators can be approximated by simply using band-pass filters.

As seen in the previous post, a Singular Spectrum Analysis (SSA) readily extracts periodic components that behave as though they are band-pass filters. So, when configured correctly, CSSA of price can be used to synthesize long-established indicators.

As a way to illustrate this statement, we show MetaStock 10.0 chart examples. The top panel shows the indicator we want to emulate with CSSA of Close. The middle panel shows the emulated version of the indicator. The bottom panel shows price.


In the following two examples we construct the MACD indicator when applied to DIA and XOM. Note that the setting in CSSA (GroupStart = 2 and GroupDepth = 7) is identical to both instruments.


Chart Example #1: SPDR Dow Jones Industrial Average (NYSEArca: DIA)
Top chart: MACD (9) indicator
Middle chart: CSSA Cycles(Close)
CSSA Cycles parameters:
m-histories = 100 bars
GroupStart = 2
GroupDepth = 7




Chart Example #2: Exxon Mobil Corporation Common (NYSE: XOM)
Top chart: MACD (9) indicator
Middle chart: CSSA Cycles(Close)
CSSA Cycles parameters:
m-histories = 100 bars
GroupStart = 2
GroupDepth = 7





In the remaining charts we demonstrate the construction of the TRIX, RSI and CMO indicators with CSSA.

Chart Example #3: Research In Motion Ltd. (RIMM)
Top chart: TRIX (12) indicator
Middle chart: CSSA Cycles(Close)
CSSA Cycles parameters:
m-histories = 100 bars
GroupStart = 2
GroupDepth = 3




Chart Example #4: Research In Motion Ltd. (RIMM)
Top chart: RSI (50) indicator
Middle chart: CSSA Cycles(Close)
CSSA Cycles parameters:
m-histories = 100 bars
GroupStart = 2
GroupDepth = 25




Chart Example #5: Avid Technology, Inc. (NasdaqGS: AVID)
Top chart: RMO indicator
Middle chart: CSSA Cycles(Close)
CSSA Cycles parameters:
m-histories = 100 bars
GroupStart = 2
GroupDepth = 19





As you can see, the GroupStart parameter is invariably set to 2 in all charts meaning that the emulated indicators are in fact low-pass filters. The GroupDepth parameter is set so as to match the width of these filters to the frequency content of the indicators.


Noxa Blogger
www.noxapredict.com


Tuesday, March 9, 2010

Smoothing without lag penalty

As you probably are already aware, “smoothing” is of critical importance when dealing with the stock market, and particularly so when trying to figure out the turning points.

At its simplest form, CSSA provides a set of bandpass filters that can be used for smoothing. The extent of smoothing can be adjusted by tuning the GroupStart and GroupDepth parameters. It turns out that traditional indicators can be smoothed and enhanced with no lag penalty using CSSA making it attractive as an alternative to traditional digital methods.

Below are MetaStock Professional 10.0 chart examples to show you how this is done. The top panel shows the indicator we want to smooth using CSSA. Here we chose the Chante Momentum Oscillator and the TRIX indicator. The middle panel shows the indicator smoothed by taking the CSSA Cycles of the indicator. The bottom panel shows price.


Chart Example #1: Research In Motion Ltd. (RIMM)
Top chart: Chande Momentum Oscillator
Middle chart: CSSA Cycles(CMO)
CSSA Cycles parameters:
m-histories = 50 bars
GroupStart = 2
GroupDepth = 8


Chart Example #2: Research In Motion Ltd. (RIMM)
Top chart: TRIX indicator
Middle chart: CSSA Cycles(TRIX)
CSSA Cycles parameters:
m-histories = 50 bars
GroupStart = 2
GroupDepth = 10



As you can see the smoothed version of the indicators show small or no lag at all! This is a radical departure from traditional digital methods which lag by construction.

Note that CSSA does not use hindsight or forward-reference making it suitable for trading; it means that the CSSA cycles are not changed as new data comes in.