CTA performance data

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.
Ag Trader
Contributor
Contributor
Posts: 3
Joined: Mon Jun 27, 2011 12:30 am

CTA performance data

Post by Ag Trader » Fri Jul 15, 2011 3:34 am

Hello All,

I am recent newcomer to the Blox community. Any good resources online that display CTA / managed futures performance? Anything that lists CAGR, MAR, etc. would be great I'm testing systems but really have no idea what are some realistic figures. MAR of .5? 1.5? etc. Thanks for the help.

- Ag Trader

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

Post by Chris67 » Fri Jul 15, 2011 4:42 am

www.iasg.com
www.autumngold.com


Testing you will probably find results are achiveable that push teh MAR way above what has "actually" been achived out there in the last 20/30 years
An MAR of 0.75 in the real world is good - some are worse - down to 0.5 for some of the L/T trend followers

Reason your backtesting will out perform reality is due I suppose - and please someone correct me if im wrong - to the fact that running a 20 year back test leaves you with so much money that you could obvioulsy never "re-invest" due to liquidity constraints etc - but it is the best way to back test to get a general feel that the overall philosophy of what you intend to do is correct !
Also your biggest draw down is always ahead !
C

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

Post by babelproofreader » Fri Jul 15, 2011 1:09 pm

Any good resources online that display CTA / managed futures performance?
I recently posted this link on another blog as part of an attempt to contribute to that particular discussion, and which frankly was a waste of time. All I got were links to other parts of the blog which did nothing but pan one of the authors of the linked paper without attempting to address the issue at hand. I offer it up here to answer your request for online sources. Make of it what you will; I offer no comment.

ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt » Fri Jul 15, 2011 6:11 pm

MAR ratios of reporting CTA's.
Attachments
Mar_Ratio_ScatterPlot.jpg
Mar_Ratio_ScatterPlot.jpg (45 KiB) Viewed 8646 times
Mar_Ratio_Distribution.jpg
Mar_Ratio_Distribution.jpg (55.36 KiB) Viewed 8647 times

DPH
Roundtable Knight
Roundtable Knight
Posts: 113
Joined: Tue Nov 17, 2009 1:07 am
Location: Central Pennsylvania
Contact:

Post by DPH » Fri Jul 15, 2011 6:21 pm

I’ve recently (last few weeks) started putting together a blog with exactly this sort of data. There is a table with performance filtered from over 1000 managers here.

http://www.tradersconnection.net/conten ... nce_1.html

You can copy and paste this all to Excel and sort it to your heart’s content. In the weeks to come I will be adding more detailed information and content.
Last edited by DPH on Fri Jul 22, 2011 12:47 pm, edited 2 times in total.

Ag Trader
Contributor
Contributor
Posts: 3
Joined: Mon Jun 27, 2011 12:30 am

Post by Ag Trader » Sat Jul 16, 2011 7:48 am

Great stuff! Cheers

fjpenney
Contributing Member
Contributing Member
Posts: 9
Joined: Sat Mar 19, 2011 1:45 pm
Location: Canada
Contact:

Post by fjpenney » Sat Jul 16, 2011 11:53 am

Dean:

Thanks for providing the info on your site. I have an issue though with providing the worst drawdown, for example, when the timeframe is different for each advisor. What is the point of providing a worst drawdown for a one year old fund and a fifteen year old fund without noting the different timeframes over which the worst drawdown may have occurred. Would it not be more useful to provide the worst drawdown over the past 36 months?

I would be interested to hear your reaction

duplicate removed by Moderator

DPH
Roundtable Knight
Roundtable Knight
Posts: 113
Joined: Tue Nov 17, 2009 1:07 am
Location: Central Pennsylvania
Contact:

Post by DPH » Sat Jul 16, 2011 11:04 pm

I agree that it is essential to compare apples-to-apples when evaluating the CTA universe. Comparing a new CTA who has not yet gone through a rough time (his maximum drawdown) verses a veteran CTA is simply useless. There are countless variations on this theme.

It is for this reason that I suggest anybody who is serious about the process, invest in a sophisticated database program and data feed service such as Barclay Hedge offers. This way you can set all your various parameters and search criteria. I have created a semi self serving video showing an example of how this can work for you here:

http://www.screencast.com/t/lQQHSIOqlSOL

This service is not cheap. Barclay Hedge offers a package (with data and analytics software) that is about $6,000 a year (and that is one of the cheaper ones). In my case, as a CTA and money allocator it is worth it.

If you are not up for spending that much money on the service, then I suggest working with somebody who has access to this sort of resource and is highly experienced.

I also updated the previous performance link to include hyperlinks to a free version of the Barclays database. You can see that here:

http://www.managedfutures.ws/ManagedFutures.html

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

Post by Chris67 » Sun Jul 17, 2011 1:40 am

any hedge fund / fund that survives 20 years plus will have a big d/d at some point - this is the bit the investment community fail to see
I launched my new fund last year and had an MAR of about 10,000 for a few days
now a year down the line, funnily enough, its not quite so good .. and if im lucky enough to be alive , still in business in 20 years time I'd take 0.6/0.7 all day long

DPH
Roundtable Knight
Roundtable Knight
Posts: 113
Joined: Tue Nov 17, 2009 1:07 am
Location: Central Pennsylvania
Contact:

Post by DPH » Fri Jul 22, 2011 12:55 pm

fjpenney,

I have gone ahead and standardized the listings to some degree...

In the following link I restricted showing performance to the following criteria.

1. Program must have 1 million or more under management

2. Program must have minimum account sizes between $50,000 and $1,000,000

3. Program must have at least 36 months of performance data

4. Time window of performance is from 4/1/2007 to 5/1/2011

Number 4 really levels the playing field. This way your not seeing somebody's 20 year worst drawdown verses somebody's 20 month worst drawdown.

The link loads a big spreadsheet, so it takes a few seconds to load:

http://www.tradersconnection.net/conten ... nce_1.html

fjpenney
Contributing Member
Contributing Member
Posts: 9
Joined: Sat Mar 19, 2011 1:45 pm
Location: Canada
Contact:

Post by fjpenney » Fri Jul 22, 2011 9:01 pm

Thanks for doing that Dean. Sorting by CAGR/WDD (in other words the MAR ratio) I see that there is a small number of funds that have a ratio greater than 2.0. There aren't many funds that achieve that kind of performance.

I am very close to having one year of published trades on Collective2 and my MAR ratio is just over 2. Personally I look at a fund's MAR rather than CAGR in isolation.

Fred

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

Post by Chris67 » Sat Jul 23, 2011 3:27 am

But looking at any fund with a track record of less than mulltiple years is fine - but dont expect the MAR to stay above 2 for long ?
Mar dcreases astime increases very simple weld held up correlation
Many investors believe a new fund < 3 years with an MAR of say 2 pursuing a similar strategy to a fund thats been around 20 years witha MAR of 0.7 , is somehow doing things better = very dangerous belief

fjpenney
Contributing Member
Contributing Member
Posts: 9
Joined: Sat Mar 19, 2011 1:45 pm
Location: Canada
Contact:

Post by fjpenney » Sat Jul 23, 2011 5:29 am

I think the main point here is that funds are being compared over the same time period. That's a starting point. Of course, if you were considering allocating money to a fund you would look at the fund's longer term performance to see how it fared through all the cycles. Prior to the table being organized to provide fund performance measures over the same time period, comparing the stats of two funds in the table was meaningless.

stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps » Fri Sep 02, 2011 10:34 pm

ecritt wrote:MAR ratios of reporting CTA's.
See attached image which graphs Compound Returns and Max DD against year of inception. My conclusions

The distribution early on looks like a truncated levy distribution (like a normal distribution but with fatter tails). So it looks like there is a lot of survivorship bias in the returns as you would expect.

For older hedge funds the very high flyers disappear - killed perhaps by excessive risk, or reversion to the mean?

The center of the distribution seems to be higher the older the funds are. This makes it look at though it's getting harder to make money these days. Consistent with this I observe (by looking into their figures at IASG) that the older funds have achieved lower returns over the last 10 years than previously.

Visually the middle of the distribution seems to be around 10% annualized return these days. Not very exciting. (You cannot just compute the average due to survivorship bias which means we have a truncated distribution).

Max Drawdown increases as the funds get older, which you would expect.

Some funds exist for no obvious reason eg one old fund with 3% return, Sharpe ratio 0.37.

The long term survivors have positive skew and positive Kurtosis. Cutting losses and hanging on for big winners.

I am looking for a way to deduce the underlying distribution, by somehow allowing for the survivorship bias. How?
Attachments
iasg_performance.png
Compound return and Max DD
iasg_performance.png (99.15 KiB) Viewed 7804 times

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Sat Sep 03, 2011 12:18 pm

stopsareforwimps wrote:The long term survivors have positive skew and positive Kurtosis. Cutting losses and hanging on for big winners.
Don't confuse (Positive Skew of Trade Outcomes) with (Positive Skew of Monthly Returns) * .

If a profitable trader rides winners and cuts losers, indeed she will have positive skew of trade outcomes.

On the other hand, if a profitable trader has a positive skew of monthly returns, it doesn't necessarily mean the trader rides winners and cuts losers. I present a counterexample below. It only took 2 hours to create and run it; I made tiny modifications to downloaded code from the Blox Marketplace REF1, REF2, and ran it on a previously described portfolio REF3. Quick and simple.

This system cuts winners short (!!). It exits at a profit target of 0.75 Rmultiples. Its profit target is closer than its stop (!). It certainly does NOT obey Seykota's "Essentials of Trend Trading". And yet it is profitable, the equity curve is in Figure 1.

Here we have a system whose returns (as reported by IASG etc) show positive skew and positive kurtosis, and yet this system does not "Ride Winners and Cut Losers".

The distribution of monthly returns (which is the only data you can get from a CTA tracking website like IASG etc) is shown in Figure 2. Yes indeed the distribution has positive Skew and positive Kurtosis. But when we look at the distribution of trade outcomes (Figure 3 linear scale, Figure 4 log scale) we see that the Skew and Kurtosis are both negative.

Figures 3 and 4 clearly show that this trader does NOT ride winners and cut losers. Unfortunately, you can't get this kind of data from CTA tracking sites.

Figure 5 show a couple of outlier trades. I'm including them to remind everybody that you can lose more than your risk-to-stop and so you can have a trade with a worse outcome than -1.0 Rmultiples. The system got long in Cattle and then price gapped beyond the stop (with Limit Locked moves to boot!). Final result: -4.2 Rmultiples.

Happily it is also possible to make profits greater than your profit target. In Brent Crude, the profit target was exceeded on Tuesday and so the system exits on Wednesday's open. Boom! Prices gap ENORMOUSLY and Wednesday's open is fantastically higher than the profit target, so the trade outcome is +5.4 Rmultiples, even though the profit target was +0.75 Rmultiples. Buy a round of drinks and be grateful.

* Sometimes people throw around the word "skew" without actually knowing its definition (LINK). They've overheard discussions of "positive skew" being an attribute of trendfollowing, but haven't actually looked it up in a dictionary. These folks incorrectly imagine that "positive skew" might mean "positive expectation" and/or "very profitable" and/or "average winning trade bigger than average losing trade" and/or "average duration of winners is longer than average duration of losers". Um, no. It's a property of a distribution , in this particular case the distribution of trade outcomes. And so is kurtosis, which wikipedia also discusses.

EDITs - fix broken hyperlinks
Attachments
Outlier_Trades.png
Fig 5
Outlier_Trades.png (21.96 KiB) Viewed 7745 times
TradeOutcomes2.png
Fig 4 (log scale)
TradeOutcomes2.png (35.25 KiB) Viewed 7747 times
TradeOutcomes1.png
Fig 3 (linear scale)
TradeOutcomes1.png (29.64 KiB) Viewed 7747 times
Monthly_Returns.png
Fig 2
Monthly_Returns.png (27.99 KiB) Viewed 7747 times
The_System.png
Fig 1
The_System.png (35.64 KiB) Viewed 7747 times
Last edited by sluggo on Mon Sep 05, 2011 8:15 am, edited 4 times in total.

stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps » Sat Sep 03, 2011 7:05 pm

sluggo wrote:On the other hand, if a profitable trader has a positive skew of monthly returns, it doesn't necessarily mean the trader rides winners and cuts losers.
Sluggo I am impressed that you were able to come up with this example.

Questions

a) Is this a pathological case or is it representative of some class of strategies with have high Kurtosis and Skew on a monthly basis but the reverse on a per-trade basis?

Maybe this statistical picture is true of Taleb distributions generally before they blow up.

b) What if any conclusions can we draw about trade strategies from real life monthly performance statistics?

Trend Rider
Full Member
Full Member
Posts: 11
Joined: Sun Aug 28, 2011 4:00 am
Location: A Hot and Dusty Place

Post by Trend Rider » Sat Sep 03, 2011 9:37 pm

Sluggo -

Very informative and a excellent example of what reality is behind the words....

Thank you for sharing this with us. As both a Swing Trader (long and short) and a Trend Follower (long and short) I often hear that I "must not be successful" as a Swing Trader. I used to debate them but not anymore. In the end it is the results that matters and not the debate.

My methodology is to Swing Trade (long or short) good set-ups within my longer term Trend Following Trades as a way to balance things by adding to overall profits while minimizing losses. The key is to carefully manage slippage and broker costs as well as losses in the shorter term Swing Trades so as not to erode the longer term Trend Following Trades.

Again thank you for posting this informative visual depiction.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Sun Sep 04, 2011 11:26 am

It is a simple but surprising fact
  • A trading system (and/or a trading manager) has more than one "skew" and more than one "kurtosis"
Now you possess another tool to detect phonies and blowhards: if they talk about "positive skew" or "positive kurtosis" without mentioning skew OF WHAT (or kurtosis OF WHAT) -- they're a phony and a blowhard.

You also know they're a phony if they state that positive (skew of monthly returns) is only possible if a trader rides winners and cuts losers.

Please feel free to play around with this system and these ideas, in your own research. Download the code if you wish, duplicate the portfolio if you're so inclined. Fiddle around with the parameters, run some tests, and see what you can learn. It will be fun!

Here's a brief little homework problem, a finger exercise, that's surprisingly illuminating: what setting of the Profit Target parameter, makes the skew of trade outcomes exactly zero? What Profit Target makes the kurtosis of trade outcomes exactly zero? Are these two Profit Target parameter settings equal to one another? Why or why not?

Protip: MS Excel has built in functions =SKEW() and =KURT(). You can use these for your calculations, or as a gold standard reference for debugging and cross-validating your own custom implementation. Excel's documentation of these functions, along with Wikipedia's, ought to get you up and running very quickly.
Attachments
Summarize.png
Summarize.png (5.53 KiB) Viewed 7622 times

stopsareforwimps
Roundtable Knight
Roundtable Knight
Posts: 199
Joined: Sun Oct 10, 2010 1:47 am
Location: Melbourne Australia

Post by stopsareforwimps » Mon Sep 05, 2011 6:12 pm

sluggo wrote:... phonies and blowhards...

Please feel free to play around with this system and these ideas, in your own research. Download the code if you wish, duplicate the portfolio if you're so inclined. Fiddle around with the parameters, run some tests, and see what you can learn. It will be fun!
It is fun, so far. I will report back in due course.

jankiraly
Roundtable Fellow
Roundtable Fellow
Posts: 52
Joined: Tue Feb 03, 2004 7:23 pm
Location: San Diego

Post by jankiraly » Mon Sep 05, 2011 11:01 pm

Hello stopsareforwimps, I think I have found a photograph. Is it right? http://tinyurl.com/3dexs2n

Post Reply