New Robot model ?

[Rembert]

While a currency or commodity based robot would be nice. I think Pascal has previously mentioned that it is more difficult (if not impossible) to build a robot on an ETF that doesn't have an underlying basket of stocks. Correct me if I'm wrong on this Pascal.

I'd like to see a robot with a low correlation to the existing robots. If correlation is very high then one might as well just increase position size in existing robots in my oppinion. GDX was a nice complement to IWM in that regard.

Here are some ETF's and their correlation coefficients relative to the existing robot ETF's.
The closer to zero the lower the correlation.

REMX (Market Vectors Rare Earth/Strategic Metals)
0.57 IWM
0.40 GDX
0.48 Average

TAN (Guggenheim Global Solar Energy)
0.43 IWM
0.10 GDX
0.26 Average

XLE (Select Sector SPDR Fund - Energy Select Sector)
0.67 IWM
0.54 GDX
0.60 Average

So based on these numbers I'd like to see a TAN robot.

The correlation info comes from : http://www.etfscreen.com/corrsym.php?s=TAN
You can change the ETF ticker in the url to see the correcation numbers relative to that ETF.

[Pascal]

While a currency or commodity based robot would be nice. I think Pascal has previously mentioned that it is more difficult (if not impossible) to build a robot on an ETF that doesn't have an underlying basket of stocks. Correct me if I'm wrong on this Pascal.

I'd like to see a robot with a low correlation to the existing robots. If correlation is very high then one might as well just increase position size in existing robots in my oppinion. GDX was a nice complement to IWM in that regard.

Here are some ETF's and their correlation coefficients relative to the existing robot ETF's.
The closer to zero the lower the correlation.

REMX (Market Vectors Rare Earth/Strategic Metals)
0.57 IWM
0.40 GDX
0.48 Average

TAN (Guggenheim Global Solar Energy)
0.43 IWM
0.10 GDX
0.26 Average

XLE (Select Sector SPDR Fund - Energy Select Sector)
0.67 IWM
0.54 GDX
0.60 Average

So based on these numbers I'd like to see a TAN robot.

The correlation info comes from : http://www.etfscreen.com/corrsym.php?s=TAN
You can change the ETF ticker in the url to see the correcation numbers relative to that ETF.

Yo are entirely right. I need to work on the underlying and see if the EV patterns can help us get an edge. So for example, I would not know how to work on a VXX or treasuries model. I'd need to start from scratch.

Also for TAN, how many people in our group are really trading TAN?
GDX/IWM are traded by many and have a low correlation.


Pascal

[Rembert]

>>Also for TAN, how many people in our group are really trading TAN?<<

Probably not many. I've never traded TAN myself. But I would start doing so if there'd be a profitable robot for it. I'm also not saying it should be TAN, but my preference if there is to be a new robot is for something that has a low correlation to the existing ETF's. I think it will be difficult to find an ETF that is currently traded by many members and has low correlation to IWM/GDX.

After all, diversification of low correlated but individually profitable systems is the only free lunch there is on wallstreet.

[Adriano]

Yo are entirely right. I need to work on the underlying and see if the EV patterns can help us get an edge. So for example, I would not know how to work on a VXX or treasuries model. I'd need to start from scratch.

OK, sorry for not understanding this.

Also for TAN, how many people in our group are really trading TAN?
GDX/IWM are traded by many and have a low correlation.


Pascal

Then in an ideal world I would like not correlated, with liquid 2x / 3x ETFs, popular. I would also care where a stop loss is going to be, not too wide.

[asomani]

Here are two return tables (with stats included, but no pivots).
The LT signal is for both linked to the 20DMF.

Attachment 8594

Attachment 8595

SPY data is until May 20 (when I did that work). IWM is until yesterday.
Both start in August 2007

Pascal

Thanks Pascal. If the return on each trade was adjusted for the difference in volatility between SPY and IWM (multiply the return of each SPY trade by 1.18 for example), I wonder what the comparison would look like.

Regardless, I can see why you may not see a SPY Robot as worth doing. If someone wants a lower beta version of IWM, they can simply reduce their position size.

Another idea: junk bonds (eg. JNK, HYG, etc.). They seem to trend quite well and could potentially offer the highest volatility-adjusted returns - and the correlation with equities is relatively low.

I would be interested to see a robot that takes trades in the area of commodities offering the most potential reward-to-risk - so, for example, it would trade XLE, GDX, etc. depending on which one is offering the highest potential reward-to-risk as determined by the robot's calculations. To me, this is preferable over simply a robot that trades one area of the commodities sector.

[manucastle]

For general US market sectors these are the correlations to IWM :-

XLF 0.74
XLV 0.70
XLB 0.84
XLY 0.82
XLK 0.84
XLE 0.67
XLI 0.85
XLP 0.57
XLU 0.54

So if we are looking for the least correlated sectors for a new Robot - Utilities, Consumer Staples and Energy may fit the bill at this present time.

Trev

[Riskslayer]

Here are two return tables (with stats included, but no pivots).
The LT signal is for both linked to the 20DMF.

Attachment 8594

Attachment 8595

SPY data is until May 20 (when I did that work). IWM is until yesterday.
Both start in August 2007

Pascal

Hi Pascal,

Given that SPY and IWM have different Multi-Time Frame Pivots; thus giving different entry levels that may get triggered on a particular day for one of the ETF's but not for the other ETF; e.g. I believe SPY is currently trading about 2% below its yearly pivot, while IWM is more like 4% below its yearly pivot - It seems possible that SPY could trigger a short sale more quickly in the short term than IWM (and perhaps more interestingly, IWM may not trigger a short at all and miss a trade that a SPY Robot would take). Moreover, an IWM and SPY Robot would on average miss a similar amount of trades that other might have taken and thus the average gains of each would reflect this. Also, I assume this effect would be more pronounced/relevant when the robot is trading in the mean reversion regime (i.e. affecting 40% of each's gain).

My point is... is that an account trading both a SPY Robot and IWM Robot would probably be able to get more trades in a given period than just one or the other; and given each's positive expectancy, this would lead to increased gains overall and probably most during the mean reversion trading periods. Perhaps even adding a QQQ Robot would help further with the goal of getting more trades in a given period.

I believe the pivot methods and money flow techniques both have their strongest edges when applied to large indexed-based ETF's vs single stock (or ETF's with few underlying stocks - maybe the GDX Robot refutes this point, but I'm not following that). If so, then we should keep to the new Robots focused on their areas of greatest strength.

My first reaction to the Q of which Robot next was to think about non-correlated ETF's - I think this would help in the trend following regimes. I think you would want to stick with the ETF's composed of large numbers of underlying stocks; among the non-correlated ETF's mentioned, I guess energy probably best fits that.. which is where you started this thread.. So, you are once again ahead of us :)

Thoughts?

Thanks,

Shawn

















I was wondering if a

[grems8544]

One of the powerful aspects of the IWM/GDX robot is that their correlations to the markets are relatively low, compared to other alternatives. This permits one robot to be long while the other is in cash, or short, as the markets dictate.

I think the consideration of any other robot has to start with correlation to the other existing robots. To me, and I know this is against the grain here, creation of a SPY robot will duplicate much of the in/out market behavior of the IWM robot, so I fail to see the point (completely ignoring that Pascal has already done this through May 2011 and the results underperformed the IWM robot).

To throw my 2 cents into the pool, I submit the following chart:



This is a correlation matrix of the various indexes with IWM, GDX, and various ETFs from the Oil & Gas group. This is a correlation back over the past 2 years of data.

Here are the symbols, in case you are not familiar with Yahoo!:

^DJI: DJ 30 Industrials
^GSPC: S&P500
^RUT: Russell 2000
^IXIC: Nasdaq 100

Two equities are said to be correlated when they have a value of "1.0". There is no correlation when they have a value of "0.0", and there is inverse correlation when they have a value of "-1.0".

As you can see, over the long haul, IWM and GDX have a correlation value of 0.4 to each other, e.g., they are loosely correlated. This gives them the ability to have different behaviors, which is why we are long GDX at the present time but looking for a short entry into IWM.

You can read across the IWM row and see that IWM is tightly correlated with the general markets (0.87, 0.92, 1.00, 0.94). Contrasting, GDX is not (0.39, 0.41, 0.41, 0.35).

I think that this is a desireable behavior of a robot, and I think that if time and effort are going to be spent, we need to find another ETF that is not well correlated with IWM or GDX.

Enter XOP.

XOP is the Oil and Gas Exploration & Production ETF, and it has solid volume. It's correlation with IWM is 0.72 and with GDX only 0.56, so again, it's a loosely-correlated candidate. Further, it's correlation with the major markets is loosely correlated (0.67, 0.73, 0.73, 0.67).

Another attractive aspect of XOP is that it has two levered ETFs, DIG and DUG. The correlations with XOP are very, very close.

I've included OIL, another ETN (Note, not Fund), only as a reference. ETNs have tax consequences, so while it can provide further orthogonality, the practical nature of OIL in terms of taxes and subsequent K1's may preclude it from consideration.

Correlations change with time -- I can model this using a weighted methodology developed by the RiskMetrics Group. If we consider approximately on the last 100 days of data (or so), here's the correlation matrix:



Note how there is a tighter correlation across the board with the various components, as well as with the major indices.

Taking this to the extreme, the correlation across the last 30 days of data (or so) produces the following:



The implications are subtle -- sometimes market conditions rotate so that previously orthogonal ETFs are out of sync, as desired, and sometimes the markets rotate so that they become more in sync. We could find ourselves completely out of the market with a IWM, GDX, XOP robot triad, or we could find ourselves in the markets.

Again, my 3 cents (inflation)...

Regards,

pgd

[EB]

Another benefit of XOP is it is well diversified. XLE is 30% (approx) Exxon and Chevron alone, so it is likely to be more correlated with large cap indexes.

[Rembert]

XOP seems to be a good overall candidate indeed.
http://www.etfscreen.com/corrsym.php?s=xop

XOP (SPDR Oil & Gas Expl & Prod)
0.65 IWM
0.50 GDX
0.57 Average