XLX signals

[senco]

Pascal, in the last few days XLE and XLF switch signals every day. Could you provide some insight on how the signals are determined? (I presume this is due to whipsawing around the oversold level, but some deeper understanding will be helpful).
Also - at this point, is the algorithm for the various XLX ETFs identical to the algorithm for the GDX money flow?

[Pascal]

Pascal, in the last few days XLE and XLF switch signals every day. Could you provide some insight on how the signals are determined? (I presume this is due to whipsawing around the oversold level, but some deeper understanding will be helpful).
Also - at this point, is the algorithm for the various XLX ETFs identical to the algorithm for the GDX money flow?

I'll check that. Both should maybe be in Bought Oversold mode.


Pascal

[Pascal]

Pascal, in the last few days XLE and XLF switch signals every day. Could you provide some insight on how the signals are determined? (I presume this is due to whipsawing around the oversold level, but some deeper understanding will be helpful).
Also - at this point, is the algorithm for the various XLX ETFs identical to the algorithm for the GDX money flow?

I indeed saw an error in the status from Bought Overbought, when the signal was just below the Average, it swiched to a short protected as if we had been down from the average.

Thanks for pointing this out, as it obviously generated artificial whipsaws. The error has been corrected.


Pascal

[pdp-brugge]

Pascal,

I presume that this slight error is in all the XLs models.
Now that you have corrected it, could you regenerate the signals for all the sectors going back starting from July 2007?
I am very busy doing backtests on these signals to create a model for trading then with balanced position sizing rules.
If this also effects the PM MF, could you regenerate those signals too please?

PdP

[Pascal]

Pascal,

I presume that this slight error is in all the XLs models.
Now that you have corrected it, could you regenerate the signals for all the sectors going back starting from July 2007?
I am very busy doing backtests on these signals to create a model for trading then with balanced position sizing rules.
If this also effects the PM MF, could you regenerate those signals too please?

PdP

All these signals are regenerated automatically every day from 2007. It just takes a few seconds.

If you went through all the XLX signals, you will have noticed that the "Shorted Overbought" signals are rarely profitable. This is probably due to the low volatility of these big S&P500 ETFs that are prone to "trend following." I therefore tested the XLX ETFs using a model without the Shorted Overbought signals (meaning that when we are long, we stay long through an overbought situation and sell when a short signal is issued.) This strategy leads to better results for most XLX ETFs, but to worst results for GDX.

This means that we could need to run models with or without the "Shorted Overbought" signals.

Another information: The RT systems for all the XLX are ready, but since we had quite a few whipsaws on the GDX RT, we decided to put this off because the situation could become uncontrollable with potentially many Buy/Sell signals being issued intraday... Without even knowing whether they should be followed since no RT back-test has been completed.

Another solution could be to publish the RT figures for all the XLX ETFs, but without the messages related to the trading signals. That way users could follow the signal in RT and take appropriate decisions when they feel necessary.



Pascal

[senco]

If you went through all the XLX signals, you will have noticed that the "Shorted Overbought" signals are rarely profitable. This is probably due to the low volatility of these big S&P500 ETFs that are prone to "trend following." I therefore tested the XLX ETFs using a model without the Shorted Overbought signals (meaning that when we are long, we stay long through an overbought situation and sell when a short signal is issued.) This strategy leads to better results for most XLX ETFs, but to worst results for GDX.
l

Pascal - if the results of the shorted overbought are mixed - maybe moving to cash on overbought would yield better risk adjusted return than continuing to hold long (lower downward deviation, higher Sortino and Sharpe, even if CAGR is a bit lower).


Regarding the RT system:
I did not have the opportunity yet to backtest the signals with next day open trades and compare to the theoretical EOD. If the EOD is better, there might be value in providing a real time signal 5-10 minutes before the close. It might be the expected EOD signal; another option to consider - declaring whatever the real time data says 5 minutes before the close as the EOD signal (ignoring rare possible last minute reversals for the benefit of keeping everything in sync... but maybe there are no means to backtest that).



Another solution could be to publish the RT figures for all the XLX ETFs, but without the messages related to the trading signals. That way users could follow the signal in RT and take appropriate decisions when they feel necessary.l
For myself a real time system would be usable only if I receive automatic alerts, or have means to create those (...difficult to constantly monitor RT inputs that produce a signal each once in two weeks or so). I presume a data feed into a charting application would be difficult, but maybe a small applet to create alerts at one's desktop and/or mobile might be realistic. Having said that - I will probably not use the information anyways until we have reasonably sufficient backtest :-)

senco

[Pascal]

Pascal - if the results of the shorted overbought are mixed - maybe moving to cash on overbought would yield better risk adjusted return than continuing to hold long (lower downward deviation, higher Sortino and Sharpe, even if CAGR is a bit lower).


Regarding the RT system:
I did not have the opportunity yet to backtest the signals with next day open trades and compare to the theoretical EOD. If the EOD is better, there might be value in providing a real time signal 5-10 minutes before the close. It might be the expected EOD signal; another option to consider - declaring whatever the real time data says 5 minutes before the close as the EOD signal (ignoring rare possible last minute reversals for the benefit of keeping everything in sync... but maybe there are no means to backtest that).



For myself a real time system would be usable only if I receive automatic alerts, or have means to create those (...difficult to constantly monitor RT inputs that produce a signal each once in two weeks or so). I presume a data feed into a charting application would be difficult, but maybe a small applet to create alerts at one's desktop and/or mobile might be realistic. Having said that - I will probably not use the information anyways until we have reasonably sufficient backtest :-)

senco

Regarding the XLX returns in 2012, basically most of the Buy signals have worked well, while short signals disappointed as much as they did on the 20DMF. As we can see in the analysis of past returns, most of the sectors ETFs have worked well when a "Buy Oversold" signal was issued. This simply means "buy the dip" strategy, while keeping an eye on AAPL's EV pattern.

For the RT, indeed, a back-test is first necessary.


Pascal