Better OddBall SystemBy
The original OddBall System no longer performs since end of year 2001 (see OddBall System – An Update). Here is an attempt to improve the system.
The original Oddball system no longer performs due to the change in the behaviour in the NYSE advance issues has changed.
The change can be described in 2 different areas. First the range in which the advance issues data covered before end of 2001 is significantly smaller in comparison to the range we see in the past few years. Second, the extreme levels recorded in advance issues are significantly higher/lower now than the levels recorded before 2002.
Thus the original rate of change method than detects market turns in the early stage becomes more noisy and less reliable.
To improve the system, we can either use a finer resolution to obtain more details from the data, or, we can use some form of filtering to remove signals they are deemed to be less reliable. I am going to apply both techniques to see if we can improve the results.
A New Setup
1. S&P or emini S&P 30-Minute Chart, 9:30 AM to 4:00 PM ET
2. NYSE Advance Issues 30-Minute Chart
3. Use full trading day data only
4. Go long when rate of change on advance issues exceeded the prescribed long level
5. Go short when rate of change on advance issues exceeded the prescribed short level
6. Do not take the long signal in 4 when daily Stochastics exceeded the predefined filter level
7. Do not take the short signal in 5 when daily Stochastics dropped below the predefined filter level
How It Looks
Here is a chart with this system applied,
The overall performance of this new variation of OddBall looks impressive, especially if we compare that to the significant drawdown in the original OddBall system.
Although this new system does not suffer the huge drawdown as in the original version, nonetheless, its performance has dropped significantly comparing now to the period before 2002. One of the contributing fact is the contraction in daily trading range of the S&P. In order to further improve the system, we will need to take into account this factor.
To those who is interested in optimization, an optimized version of this new oddball system can do 30% better if I use long specific levels and short specific levels. The setting I am using now is symmetric for both long and short side.
The data for the symbol
QC:ADVN.NY are available directly for download if you switch to use No Feed in NeoTicker.