Measuring your correlation to professional futures traders

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
BuyHigh SellLow
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Wed Apr 27, 2011 12:46 pm
Location: U.S.

Post by BuyHigh SellLow » Wed Sep 28, 2011 3:33 pm

Chris67 wrote:SURELY 2 TF's with correlation below 50% are not TF 's ?
Seems carazy to expect 2 firms who both follow trends in global futures markets to have a correlation of less than 0.5 ? if tehy do something maybe wrong
Once again Institutional investors looking for what probably doesnt exist ? looking for the holy grail perhaps ?
Of course many TF's have lowly correlated performance over the short term but as we all know that counts for diddly

IncidentallyJAZ in the World of TF - Correlation of 0.75 would get me very excited ? 0.88 gets me going over the long run hehe
Best
C
I think finding CTAs with low correlation to each other is easy. Throw in a fundamentals/meteorologist CTA; option writer CTA; spread trader CTA; pattern recognition CTA and a trend following CTA. Whether these strategies outside of trend following are worthwhile is another discussion.

But combining TREND FOLLOWING CTAs will only go so far. What if the only diversification you're getting there among the trend followers is because they're not PURE trend followers, i.e., they're employing other strategies in addition to trend following?

drm7
Roundtable Fellow
Roundtable Fellow
Posts: 96
Joined: Sun Apr 20, 2003 9:02 pm
Location: Richmond, VA

Post by drm7 » Wed Sep 28, 2011 4:09 pm

Chris67 wrote:SURELY 2 TF's with correlation below 50% are not TF 's ?
Seems carazy to expect 2 firms who both follow trends in global futures markets to have a correlation of less than 0.5 ? if tehy do something maybe wrong
I think by the broadest definition of "trendfollowing" (a medium to long term system trading a globally diversified portfolio) you're probably right. However, market selection can create differences. An all-currency program may perform very differently than an all-metals or all-grain program. (Or even an all non-US diversified vs. an all-US diversified.)

System tempo may also make a difference - a 20 day breakout system could have a much different looking equity curve than a long-term triple MA system.

Demon

Post by Demon » Thu Sep 29, 2011 4:32 am

As an allocator I would think if you buy a TF CTA you want it to behave like a TF CTA, so overall correlation of 0.75 I would think gives you that. However I would think the perfect scenario is where correlation for up months / down months is below circa 0.5. So you know you're going to get the LT performance of a TF CTA but would hope to see a reduction in your month on month volatility by adding this particular programme / fund to your portfolio.

jas-105
Roundtable Knight
Roundtable Knight
Posts: 130
Joined: Sat Aug 02, 2008 2:32 am
Location: London, England.

Post by jas-105 » Thu Sep 29, 2011 5:50 am

Demon

Thanks for the clarification. In terms of importance, how does the monthly up/down correlation rank when you are looking at allocating to a fund, CTA or otherwise ?

Demon

Post by Demon » Thu Sep 29, 2011 6:35 am

Sorry Jas - when I said 'As an allocator' I didn't mean I was, I was just trying to put myself in there shoes so to speak.

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Thu Sep 29, 2011 7:01 am

I Disagree on the CTA phrase again - either a fund is a trend follower or its a CTA (which means nothing) or it employs some other sort of futures strategy
Getting a basket of lowly correlated TF's together (long term) is not really possible as they should all have high correls in the long term - which they do
Howver that being said we need to define high correlation and that goes back to the "what sort of systems should I mix in my suite " argument
I keep an index of 30 L/T TF's which Ive offered to send to anyone who wants it - going backl to 1977
What do you think happens if you combine 30 guys who are fairly high;y correlated ? This is what institutions dont understand
If you have 2 guys who do it right over the long term correlation = 0.9 you get pleasantly surprising results

SimJimons
Senior Member
Senior Member
Posts: 31
Joined: Mon Jun 20, 2011 9:59 am

Post by SimJimons » Thu Sep 29, 2011 7:17 am

Chris67, can you please elaborate on your last two sentences, since I don't really understand what you mean? Thx...

jas-105
Roundtable Knight
Roundtable Knight
Posts: 130
Joined: Sat Aug 02, 2008 2:32 am
Location: London, England.

Post by jas-105 » Thu Sep 29, 2011 7:30 am

But there's no getting away from the fact that investors will still look at monthly returns even on an investment that is deemed to only really "work" over 2+ years at the very least. So, if you have monthly numbers that zig when others zag that has to be a good thing (although not necessarily important in the long term) ?

Demon

Post by Demon » Thu Sep 29, 2011 9:10 am

By TF CTA I mean a Trend Following CTA as opposed to a Mean Reversion CTA or a Discretionary CTA for example. If every TF or TF CTA out there was correlated at 0.80+ to every other TF CTA in up markets, down markets and all markets then what point would there be in trying to build a portfolio of CTAs or TF CTAs or whatever you want to call them? Would there be much point in launching a new fund or programme?

I don't believe that you have to be correlated at 0.80+ in every which way to the Newedge CTA Index or every other TF CTA out there to be able to call yourself a TF CTA, I do believe you can go from A to B over a long period of time like most TF CTAs but via a different path. In order to do so I believe you need to be different from the herd, so a single strategy on a specific portfolio may be more succesful than loads of 'different' systems on 100+ markets. Ultimately it all depends on what you are trying to achieve and from what type of investor you are seeking an allocation.

SimJimons
Senior Member
Senior Member
Posts: 31
Joined: Mon Jun 20, 2011 9:59 am

Post by SimJimons » Thu Sep 29, 2011 11:02 am

Demon, while correlation is interesting, what it eventually comes down to is drift. You can have a correlation of 1.0 to competition and still manage to sell your product, as long as you deliver higher risk-adjusted return :wink:

Demon

Post by Demon » Fri Sep 30, 2011 2:50 am

Sim, absolutely! after all who wants to be in a fund with low correlation if its's consistently losing money!

Chris67
Roundtable Knight
Roundtable Knight
Posts: 1046
Joined: Tue Dec 16, 2003 2:12 pm
Location: London

Post by Chris67 » Fri Sep 30, 2011 3:14 am

Mr Jimmons

What I mean is simply this
If had 90 Million to Invest as an institutional investor I would give $3 Million each to the 30 TF's in my index - some of these TF's are correlated at 0.95+ and some are correlated as low as 0.75
If you study the outcome of this you get a very smooth equity curve - not necessarily what you expect since by common definition a lot of these funds are "highly correlated"
I would rather have 30 Highly correlated TF's anyday than a few TF's and a a few other Futures trading strategies - the outcome is vastly superior over the long term - if you stick 1000 investments in a basket each with a correlation of 0.9 you dont see the results you expect - in fact others should do thier own research to investigate this further - it sort of blows a hole in modern portfolio investment theory and shows it for tje utter dross that it is
JAS - re your point about Investors - you are very correct - but then ask these investors who focus on the last 2 years what their L/T investment performance looks like - Bet I could do better with cash in the bank

babelproofreader
Roundtable Knight
Roundtable Knight
Posts: 138
Joined: Wed Nov 10, 2004 4:36 pm

Post by babelproofreader » Sat Oct 01, 2011 10:31 pm

I have written an extremely basic Octave script (available here) to investigate the issue of correlation, the output of which is shown below.

I created 3 simple return streams of consecutive up and down returns

return stream a (black) is up 2, down 1
return stream b (blue) is up 3, down 1.5
return c (cyan) is the same as b, but offset by 1

and then plotted the return streams' equity curves individually and combined (a+b being red and a+c being green). Obviously the correlation of both the equity curves and return streams of a and b is 1, whilst the correlation of the equity curves of a and c is 0.6314 but the correlation of the return streams of a and c is -1. This is an important difference. In the hunt for the ideal equity curve (green, with no draw down, compared to red, which suffers 50% retracements of previous draw ups) it is important to look for return streams that are as uncorrelated as possible in their returns' departures from their respective average return with, of course, the average return being positive.

I think almost all traders would prefer the monotonically increasing green equity curve over the volatile red one, even though they both have the same ending equity. The danger in searching for systems/strategies that have highly correlated, individual equity curves is that you end up with the red equity curve and not the green. Concentrate instead on searching for uncorrelated departures from average return.
Attachments
correlation.png
correlation.png (17.18 KiB) Viewed 5448 times

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Sun Oct 02, 2011 4:56 am

babelproofreader wrote: I think almost all traders would prefer the monotonically increasing green equity curve over the volatile red one
Yes, some kindly, cuddly old dude in the States spotted that a few years back. What was his name? Bernie somebody or other? :D

babelproofreader
Roundtable Knight
Roundtable Knight
Posts: 138
Joined: Wed Nov 10, 2004 4:36 pm

Post by babelproofreader » Sun Oct 02, 2011 3:42 pm

@AJF Garner

I would agree that if some money manager offers a green equity curve as their historical record one would have to be very suspicious of a "Bernie somebody or other" scheme and exercise extreme due diligence. However if, as Chris67 has suggested, one has to place money with a number of TFs or CTAs (or you are combining your own systems) each presenting different variations of equity curves such as a, b and c the way to combine them in your own portfolio is on the basis of uncorrelated departures from average return.

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Mon Oct 03, 2011 5:12 am

Yes, agreed of course. It is just that in many respects back testing is such an elusive and misleading experience. It has to be done and it certainly has great value but the future just does not turn out like the past as we all know. With CTA programs, as people have pointed out as nauseam, you have the additional problems of survivorship (or lack thereof) and monthly data points.

Zig when others zag, yes, yes, yes the whole gamut but in practice I suspect you are unlikely to approach the green curve. Which is not to say don't aim for it, just to say that you will only be certain of achieving it once you HAVE achieved it. Or not, as the case may be.

I always like the seasoned, wise and honest approach of the sage of Incline Village on such matters. When people want to give him money to duplicate that legendary equity curve of his quoted in Wizards, he says they should give him the money in 1970 and then he can do it.

jas-105
Roundtable Knight
Roundtable Knight
Posts: 130
Joined: Sat Aug 02, 2008 2:32 am
Location: London, England.

Post by jas-105 » Thu Oct 06, 2011 4:58 am

There is certainly a lot of zigging and zagging going on among the big trendfollowers as they post their Sep numbers.

Post Reply