[RobGiles]

OK thanks Len,

For my S&P500 system which has 16 positions, 20 out to 50 had no meaningful affect using the code in this thread.

[RobGiles]

Len Zir wrote:

Rob,
I didn’t follow any rules. I just tested.

thanks Len….did you get any uplift in results?

[RobGiles]

Len Zir wrote:

Daniel,
Used same code above plus:
MinRank = Param(“Min Rank”,20,1,500,1);
SetOption(“WorstRankHeld”,MinRank);

I used 5 positions for Nasdaq aggressive.

Len, Others, is there a rule of thumb here on how many positions you have in your portfolio and how far down the ranking list the cut-off has to be? i.e. the example given in Nick’s presentation was a system with 10 positions, will still hold them if they are in the top 20. So if your’re running 15 positions, top 30? 5 positions / top 10?

FWIW I tested my USMOM system (S&P500) by optimizing the MinRank variable and it made no real difference.

[RobGiles]

Great thanks Nick.

Am delving into the ASX 100 system more today.

[RobGiles]

Thanks for sharing Said,

Do you mind giving us an indication of what % of your overall investment capital you have in MOC systems?

[RobGiles]

Thanks Nick. I’ve looked at the USMOM through many different time periods, including the 2007 – 2010 and was happy with the results. I guess what Scott & I are asking is should we be concerned that a system’s performance is acceptable on one market (S&P500) but when applied to another market (ASX100 with exactly the same parameters and parameter values) the results deteriorate markedly?

I would have thought that stock markets in different countries, although correlated, would still exhibit very different personalities due to exchange rate influences, different mix of industries, political influences etc?

USMOM1 Backtest result for the 1/1/07 to 1/12/11 period:
CAR 7.29%
MaxxDD -30.88%
that was using a long Index MA. If I shortned the MA by 80 days it gave much better results:
CAR 11.77%
MaxDD -20.81%

but of course you wouldn’t have known this at the time, although maybe your once a year reoptimization of things like Index MA filters may have got you there?

[RobGiles]

thanks Scott….you raise some good points. I’d value your insights please Nick? Does the fact that the results are poor on the ASX for my USMOM1 system render that system useless or does it downgrade the robustness score?

I’d also be interested how your MOM systems designed for the US markets perform on the ASX (and / or vice versa)….were they similar or like mine? Anyone else tried this out?

[RobGiles]

Scott McNab wrote:

Rotational systems need more time to give a feel for performance…5 year windows can vary much more dramatically than MRV or MOC as only getting a few data points each month…is the USMOM system you trade any good on ASX Rob ?

Hi Scott

Yes I have noticed, and expected, the variance in returns based on which 5 year window was being analysed. So what window would you use for testing a MOM system….10?

If I use the exact same rules and parameter values that my S&P500 system has on the ASX100 I get the following over the last 10 years:
CAR 2.35%
MaxDD -31.40%
MAR Ratio 0.07

vs the S&P500 system with the following metrics:
CAR 14.81%
MaxDD -20.84%
MAR Ratio 0.71

Note that the parameter values were optimised for the S&P500 system, but not for the ASX100 system yet. Are these results an indication of lack of robustness (given that they are materially different across the 2 markets) or would one expect the value of the metrics to be materially different in different markets?

[RobGiles]

Thanks Trent

Well I thought I’d choose a random 5 year period as my ‘In Sample’, try to get something that works then test it ‘Out of Sample’ to check robustness. Happy to take suggestions on a better methodology. How would you test it / what time frame would you use?

[RobGiles]

Murphy’s Law!

[RobGiles]

August ’18 performance:

USMOM1 +13.28% for the month.
Big gains in a lot of stocks like AMD +37.3%
+7.4% since inception (adjusted for 2 tranches of fresh capital invested May & July 2018).

USMOC1 -0.33% for the month
– 19.4% since inception.
Have traded the system using the CBT & ROC as a ranking approach.

ASX Growth +0.86% for the month.
+39.6% since inception.

System development:
– ASX100 Rotational. Experimenting with linear regression & ROC weighted over 2 periods. Test period Oct 2012 to Oct 2017. Still a bit of work to do to get desired results.

– NDX concentrated rotational.

[RobGiles]

Re other markets to trade – thanks for sharing Julian,

Would it be worth spending the same energy & time simply developing a system for a single instrument that is highly liquid in order to add diversification to your overall portfolio, like some of the major US futures contracts?

[RobGiles]

Nick, I’ve only just read this, so just want to confirm my understanding.

If I have Amibroker set to > File > Database settings > Configure > “Capital reconstructions and special distributions” I will not be seeing any returns pertaining to dividends in my backtest results?

I’m assuming that these can be added manually or one can make an educated guess as to the additional return relating to divi’s depending on the universe you’re testing?

[RobGiles]

Agree with Daniel, If you’re building a momentum system watch Nick’s CMT presentation. It’s given me more clarity as to what I want to investigate next.

[RobGiles]

Thanks Nick, that’s exactly what I thought was going on. These guys that bang on about winning competitions in order to sell courses are really just relying on attracting people that know bugger all about survivor-ship bias.

[Author]

Posts