IWM Robot Fail-Safe Freeze

[Billy]

Since the 20 DMF Market Direction Model missed the buy signal by a tiny porosity hair in December and stayed neutral, the IWM robot ST/LT settings have looked and will continuously look for shorting the uptrend based on a “neutral” 20 DMF MDM signal.

That’s because the IWM robot backtests have always found edges in fading strength for reversion-to-the mean trades when in a “neutral” mode. It would have been long if the buy signal triggered back in December. It is an unfortunate repeat in reverse of what happened last August when the robot continuously tried to go long against the downtrend which was also qualified as a “neutral” market by the 20 DMF MDM.

The current uptrend is further fueled by abundant available liquidity and defiance of sovereign bonds.

As we wrote on the last short signal that was stopped out today, we are now objectively forced to trigger a fail-safe stop rule for the whole IWM robot system. It is programmed to only go short under present “ neutral” 20 DMF settings and it could become a one-way road to hell. So, until we have a non-neutral signal from the 20 DMF, we will freeze the IWM robot and let it in cash whatever the ST/LT settings say.

I agree in advance with all your reactions : The market is sometimes a cruel, cruel world for mechanical systems with rigid “lines in the sand” rules. At times, it will save you from a catastrophe, but at other times it can become stubbornly stupid because of a microscopic indicator miss. We encouraged you to avoid leverage, but we cannot be confident that the IWM robot will help you improve your performance in the very near term and feel compelled to responsibly make this drastic decision.

The GDX robot is in no way concerned by this event.
Billy

[nickola.pazderic]

Billy,

Thanks for the update. It is very helpful.

[TraderD]

Since the 20 DMF Market Direction Model missed the buy signal by a tiny porosity hair in December and stayed neutral, the IWM robot ST/LT settings have looked and will continuously look for shorting the uptrend based on a “neutral” 20 DMF MDM signal.

That’s because the IWM robot backtests have always found edges in fading strength for reversion-to-the mean trades when in a “neutral” mode. It would have been long if the buy signal triggered back in December. It is an unfortunate repeat in reverse of what happened last August when the robot continuously tried to go long against the downtrend which was also qualified as a “neutral” market by the 20 DMF MDM.
Billy

One thing to consider here is the long-range effect of the OB/OS oscillator since it latched (or missed) its most recent turning point. The Dec 20 near-miss of the 20DMF long posture (and staying in neutral instead) was about 6 weeks ago. Theoretically, this condition could proceed uninterrupted (ie "stuck on stupid") for a very long time.
AFAIK this potentially unlimited time span for a model state is uniquely different from the other EV-based elements of the system, where historical/cumulative measurements are 20-days long (or roughly twice that with 20D EMA smoothing applied). In other words, the rest of the system keeps refreshing itself and getting "unstuck" continuously. I would think re-evaluating the construction and use of the OB/OS oscillator in the IWM robot is warranted.

Another point I wanted to make is that while following the IWM price action following attempted short entries, I sense that there is merit in a very short-term mean-reverting trade setup using the advocated pivot-based price points. These setups seem (not formally tested as it stands) to materialize within a day or they should be scratched.

Trader D

[xr-3609]

Since the 20 DMF Market Direction Model missed the buy signal by a tiny porosity hair in December and stayed neutral, the IWM robot ST/LT settings have looked and will continuously look for shorting the uptrend based on a “neutral” 20 DMF MDM signal.

That’s because the IWM robot backtests have always found edges in fading strength for reversion-to-the mean trades when in a “neutral” mode. It would have been long if the buy signal triggered back in December. It is an unfortunate repeat in reverse of what happened last August when the robot continuously tried to go long against the downtrend which was also qualified as a “neutral” market by the 20 DMF MDM.

The current uptrend is further fueled by abundant available liquidity and defiance of sovereign bonds.

As we wrote on the last short signal that was stopped out today, we are now objectively forced to trigger a fail-safe stop rule for the whole IWM robot system. It is programmed to only go short under present “ neutral” 20 DMF settings and it could become a one-way road to hell. So, until we have a non-neutral signal from the 20 DMF, we will freeze the IWM robot and let it in cash whatever the ST/LT settings say.

I agree in advance with all your reactions : The market is sometimes a cruel, cruel world for mechanical systems with rigid “lines in the sand” rules. At times, it will save you from a catastrophe, but at other times it can become stubbornly stupid because of a microscopic indicator miss. We encouraged you to avoid leverage, but we cannot be confident that the IWM robot will help you improve your performance in the very near term and feel compelled to responsibly make this drastic decision.

The GDX robot is in no way concerned by this event.
Billy

Billy & Pascal,

As you are looking at these issues; I have a question regarding the "robustness" added by the inverse ETF confirmations regarding short 20DMF signals. In 20DMF evolution: November 14, 2010 , (if I understand it correctly): the 20DMF true signal passes thru zero line without SDS confirmation = 321.42% cum return (Table D) and the 20DMF with ETF's confirmation = 347.40% (Table E), is the 26% difference in cumulative return statistically significant given that the data only goes back to 2007? Also - did the 20DMF rules model fail to issue a short signal in late July 2011(?) and was it due to the failed confirmation of the inverse ETFs? If this is correct; have you updated the cumulative return back test data for the efficacy of the inverse ETF confirmation after that sell off? I understand the logic of the ETF confirmation, such that institutional investors will hedge instead of selling - but at certain times does fear cause them to just dump stocks?

Greg

[davidallison@gmail.com]

Billy / Pascal,

Would it be possible to keep the IWM robot active for information purposes only? Perhaps just put a disclaimer on the page.

Dave

[Billy]

Billy / Pascal,

Would it be possible to keep the IWM robot active for information purposes only? Perhaps just put a disclaimer on the page.

Dave

Dave,
Yes, all statistics and the multi-pivot spreadsheet will normally be automatically updated as usual on the robot page. But the message will be “no advised position”.

Greg and Trader D, both of your questions are more in Pascal’s field of expertise than mine and I will let him answer.

Billy

[Pascal]

Billy & Pascal,

As you are looking at these issues; I have a question regarding the "robustness" added by the inverse ETF confirmations regarding short 20DMF signals. In 20DMF evolution: November 14, 2010 , (if I understand it correctly): the 20DMF true signal passes thru zero line without SDS confirmation = 321.42% cum return (Table D) and the 20DMF with ETF's confirmation = 347.40% (Table E), is the 26% difference in cumulative return statistically significant given that the data only goes back to 2007? Also - did the 20DMF rules model fail to issue a short signal in late July 2011(?) and was it due to the failed confirmation of the inverse ETFs? If this is correct; have you updated the cumulative return back test data for the efficacy of the inverse ETF confirmation after that sell off? I understand the logic of the ETF confirmation, such that institutional investors will hedge instead of selling - but at certain times does fear cause them to just dump stocks?


Greg

Greg,


if you continue reading the document, there is more about the short rules that we are using with or without inversed ETFs confirmation. The list of trades is also added at the bottom of the file. I add a screen short for every one to access the document.


Pascal

12608

[xr-3609]

Greg,


if you continue reading the document, there is more about the short rules that we are using with or without inversed ETFs confirmation. The list of trades is also added at the bottom of the file. I add a screen short for every one to access the document.


Pascal

12608

Pascal,

My question is did you look at the data given the premise that the "primary confirmed short" trade need not be "confirmed" by the inverse ETFs and all other rules being the same (I know it complicates the fake bounce rule)? It may be in the data and I do understand it. There may be other obvious reasons (which I do not see) for not considering this modification.

Greg
Greg

[Pascal]

Pascal,

My question is did you look at the data given the premise that the "primary confirmed short" trade need not be "confirmed" by the inverse ETFs and all other rules being the same (I know it complicates the fake bounce rule)? It may be in the data and I do understand it. There may be other obvious reasons (which I do not see) for not considering this modification.

Greg
Greg

Greg

I do not understand the question.
Please send me a private mail and we will discuss it out of the board.


Pascal

[TraderD]

Dave,
Yes, all statistics and the multi-pivot spreadsheet will normally be automatically updated as usual on the robot page. But the message will be “no advised position”.

Greg and Trader D, both of your questions are more in Pascal’s field of expertise than mine and I will let him answer.

Billy

Suggest to color code the robot message in RED: "Please note that the IWM Robot short positions should be avoided until the 20DMF is out of its neutral stance. The trades stated below are for information only."

[adam ali]

Pascal,

Could you elaborate more on what the thinking was behind the IWM Robot "following" the last 20DMF non-cash signal?

If I understand the Model correctly, the Robot is trained to lean in the direction of the last 20DMF buy/sell signal when looking for a trade. I'm wondering why - if the 20DMF itself is neutral - the Robot doesn't approach the question of long/short in the same fashion, i.e., unbiased, when looking for new trades...until the 20DMF gives a new non-neutral signal.