XLB

[pdp-brugge]

Pascal,

The XLB MF is in "Shorted Overbought" state since the close of July 6. Trading this signal has not been productive. The statistics of Shorted Overbought for this model show indeed a serious dip after day 8. This gives me some hope. What is needed for the XLB MF to change it's state?

In brief, could you summarize the rules for the signal changes of the 9 S&P sub-components?

PdP

[Pascal]

Pascal,

The XLB MF is in "Shorted Overbought" state since the close of July 6. Trading this signal has not been productive. The statistics of Shorted Overbought for this model show indeed a serious dip after day 8. This gives me some hope. What is needed for the XLB MF to change it's state?

In brief, could you summarize the rules for the signal changes of the 9 S&P sub-components?

PdP

PDP,


The XLX models are all identical. Their only differences lie in the specifics of sector volatility and hence, OB/OS levels and porosity factors. Going through all the rules is somewhat long, but basically, the whole idea is that the model is a finite state machine that is based on a MF based 20D rolling window. This machine tries to capture the main moves of a trend in one sector.

It is easy to see below that the sates move from overbought to oversold and start issuing signals in between. There are also Buy/Sell signals issued when crossing over the 0 level and with reference to the MF 20D Average.

15190

The idea of having nine finite states machines and their related stats for the four types of trades is that there is a degree of sectors rotation within the S&P500 and hence, sometimes it is better to switch from one sector to the other. However, when the whole market sharply moves up or down then even a specific sector will be pulled or dragged by the whole market. This is why usually, I would trade probably three sectors in different states.

I would however favor states that show positive looking statistics, where the potential profits are higher than the risk. For example, for XLB, the potential profit is about similar to the potential drawdown.

15192

Finally, even if the model is a finite state machine that works with EOD data, you have the RT signal on XLB that shows you where you are. We are still working at adding the OB/OS lines there.

When on Shorted Overbought, the model will switch back to cash if it hits the Overbought level again or it will turn to a "short protected" mode if it crosses below the 0 level. It did cross below the 0 level on July 11 in RT, just before the bounce and then turned positive ("BUY" in RT). This means that since July 11, the RT XLB model - which we do not track right now - is in "BUY" mode. The EOD model stays short because it did not catch the move below 0.

15191

I understand that it is difficult to follow both an EOD and a RT model. This requests some flexibility. We are still working on the RT system so that we will be able to better show what is happening and to have alerts automatically issued, but we ran into some administrative delays.

On another note, you might have noted that if you add all the returns of the XLX models you will note everyday that they are almost always positive. This means that the model used among nine sectors correctly catches the sector rotation.

The main weakness of such model is when the markets goes into a very strong up or down-trend: when the trend lasts longer than the analysis window (20 days) without a reversal, then you can get stuck in the wrong state for quite a long time.

Here also, some flexibility is requested from users to understand what the model is doing.

Thank you for your patience.




Pascal

[pdp-brugge]

Pascal,

My gut feeling tells me that the RT models should be a lot better then the EOD models. The results look very promising. However, I will only trade a system that I have back tested. I think I am not alone with this attitude.

Once realistic back tests are available for the RT models, then a lot of members from this site shall be very interested in a (paying) RT robot subscription, I presume.

If I could help you with doing back tests with RT data, just give me a sign...

PdP

[Pascal]

Pascal,

My gut feeling tells me that the RT models should be a lot better then the EOD models. The results look very promising. However, I will only trade a system that I have back tested. I think I am not alone with this attitude.

Once realistic back tests are available for the RT models, then a lot of members from this site shall be very interested in a (paying) RT robot subscription, I presume.

If I could help you with doing back tests with RT data, just give me a sign...

PdP

Thank you for your offer.

It is OK to wait for a full back-test, which is in the plan, but since you have access to RT data, I advise that you'd be "flexible" enough with your positions and trade them "as you see fit," taking into account the RT data.

I am doing it all the time now. This at least helps to better find entries/exits during the day.

Markets are now reacting pretty fast and I believe that everybody needs to adapt... or take a much longer view point, which implies maybe larger "temporary" drawdowns and no leverage.



Pascal