Question for Trading Blox users

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.
Post Reply
rhc
Roundtable Knight
Roundtable Knight
Posts: 464
Joined: Tue Nov 18, 2008 8:46 pm
Location: Oz

Question for Trading Blox users

Post by rhc » Sun Jun 14, 2009 12:41 am

I am not, as yet, a Trading Blox customer, so I will post this question here rather than at the customer support forum. (which of course I’m unable to access)

o.k. here goes;
With Blox, I can back test a given portfolio with a given system parameter set over a given time frame (say 1985-2005) and get the results for this period of time. O.k. fine.
Is there a facility within the Blox software that then allows for extracting the ‘rolling’ 5 year results.?
What do I mean by this?
Well, as I say above we already have the results from 1985-2005. Is it possible to then break this down to results for 1985-1989, 1986-1990, 1987-1991, 1988-1992 etc etc.
In other (clearer?) words, what would have happened to my account if I began trading at any particular 5 year period?

Of course, one could simply run multiple 5 year tests (i.e. 21 individual runs in the above example) and get the individual results, but I’m more interested in running the one test and getting the results of all the 5 year tests on the one page.

Is this a doable thing in Blox? (via a custom programme perhaps?)

Thanks in advance for any advice

kingami
Full Member
Full Member
Posts: 15
Joined: Mon May 04, 2009 12:59 am
Location: Atlanta

Post by kingami » Sun Jun 14, 2009 1:46 am

You can use "Start Date Stepping" which steps through different start dates (in your case five year interval) and compares the results.

Stephen Newton
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Dec 31, 1969 7:00 pm
Location: London, England and Cape Town, South Africa

Daily 5 Year CAGR vs Drawdown - Does System timing work?

Post by Stephen Newton » Mon Jun 15, 2009 9:29 am

rhc

As kingami suggests this is certainly possible using the start date increment parameter and the length of test parameter.

I have recently been through this exercise when talking to potential investors and can share some of my experiences here for your benefit. I would also appreciate others comments on the notion that the 'market timing' thinking of people like Sy Harding are quite analogous to a 'system timing' philosophy.

The chart attached shows the results of 5808 tests for a system starting on 1/1/82 using single day increments and running the system for 1320 days (22 days per month * 12 months * 5 years). I have plotted this against the drawdown of the system, trading from from 1/1/82 (uninterrupted till the data runs out).

I certainly see that outsize returns occur when drawdowns are greater than 50%.

This posses a dilemma when risking your own funds or convincing investors that this is the time to invest. At 50% DD making an argument to investors that the last 2 years performance is NOT what counts but rather that it presents the real opportunity / another edge - can be a little challenging unless you have a set of very sophisticated investors. cf referred to this as Recency bias in his book and Sy Harding quotes Dunbar research showing that investors have made 2.6% when managed funds made 9.4% and the S&P made 12.? % (numbers from memory so might be slightly out) in the period 84-05. He uses these stats to suggest that investor bias to invest in the best fund last year is a poor strategy. Could systems traders bias to avoid systems in Drawdown be equally poor? Do system traders care? I would be interested to hear from this group whether they consider this factor at all or assume that over the long term (greater than 5 years a good system will have an almost flat red line in my chart?)

I am playing with adding capital in DD to see the effects on trading but wondered if others had thoughts or ideas.

On the process to produce this information - it was not trivial. TB/Windows/Mac all seem to struggle with the output of 5808 tests. I therefore had to break each test into 3 chunks of 1936. I ran this test for year 1, year 2, year 3 and year 4 as well as the year 5 shown here. This took about a week to pull together as I struggled through various blue screens etc. Achievable and certainly worth the time for me.

Stephen
Attachments
5 Year return vs DD.png
5 Year return vs DD.png (130.55 KiB) Viewed 6764 times

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Mon Jun 15, 2009 12:29 pm

Adding capital in a DD assumes you have a way of identifying situations in advance, after which the systems return is exceptional. I don't think you are able to do that, any more than you are able to do that for any market.

Look at your picture. Assume that the biggest DD, the initial one, would have happened just to the right of the picture (in the immediate future). You look at the rest and come to your conclusion: every time the system reached 50% DD, it went back up with improved return, so I'm going to add capital at each 50% DD. You start doing that today and the big one happens: you add at 50% DD and go way down to 90%.

My point is this: at the point of 50% DD you are not able to tell whether it's going up tomorrow, or whether another 50% DD (to a total of 75%) will happen.

------------------------------------
Unrelated: for my taste these wide differences in return and DD are too large, too choppy. I would try to work further along to get a more steady picture. Mathematically speaking: you need to reduce the standard deviation of the 5-year-return (and of the DDs as well).

Stephen Newton
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Dec 31, 1969 7:00 pm
Location: London, England and Cape Town, South Africa

Post by Stephen Newton » Mon Jun 15, 2009 1:04 pm

Thanks for the reply Asamat but I think you might be misunderstanding the chart and hypothesis slightly.

I am not interested if the system will show returns the next day or the next month or is going down into a deeper DD the next day or next month (or 6 months for that matter).

The red line shows the CAGR (Compound Growth AFTER 5 Years). I think you can reasonable say that an investment made when the DD curve was below 50% (5 YEARS previously) almost always produced a 5 YEAR CAGR greater than investing when DD was less than 50%.

You can certainly act on your DD today to influence an outcome in 5 Years time?

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Mon Jun 15, 2009 1:53 pm

Stephen Newton wrote:You can certainly act on your DD today to influence an outcome in 5 Years time?
I'm sorry, but I disagree. You think you have discovered a correlation, while I think this correlation does not exist.

The correlation I think you think you have discovered is the correlation between the blue and the red curve. That is, between todays DD and high return over the next 5 years. I rather think the red curve is a transformation of the blue curve. To describe the red curve in words: "hugh returns come after big DDs" is the same as describing the blue curve: "after every DD comes a new equity high". Since in order to recover from a 90% DD you need to have several years of near 100% returns, this is kind of a tautology.

Since in constructing systems you select with hindsight only systems, which survive (and therefore recover from their DDs), I think it is neither a surprise that the blue line recovers, nor that it does so with high returns after high DDs (the red line). On the contrary: I would predict that every system which recoveres from high DDs shows this behaviour.

Essentially what you are suggesting is a Martingale money management strategy, applied to the equity curve instead of trades. Martingale strategies seem to work fine, until the one time when they fail. In the long run in trading these strategies do not work.

Stephen Newton
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Dec 31, 1969 7:00 pm
Location: London, England and Cape Town, South Africa

Post by Stephen Newton » Mon Jun 15, 2009 2:28 pm

Asamat

I still think there may be some confusion on what the red curve is.
To describe the red curve in words: "hugh returns come after big DDs" is the same as describing the blue curve: "after every DD comes a new equity high"
I think this description is an interpretation of the lines - not a description of what the lines actually are. I may not disagree with the interpretation (still testing and getting ideas from others here) but I think any interpretation must be off the the correct facts.

To clarify the red curve is ... the CAGR at the end of 5 Years of trading this system STARTING on the day it is plotted against on the x axis.


Plotting these points on a chart in day increments could create the misunderstanding that the next point is the previous point with one trading day added. It is not. It is still a 5 year CAGR just starting one day later (ie on the day of the next point on the x axis). So the red line is not an equity curve as you see it on all of TBs output. It is a line of 5 year CAGR's with one day increments to the start but still for a fixed time of 5 years.

The blue line is a standard DD curve as produced by TB. ie day 5's DD is the result of one more days trading than day 4.

In my interpretation of this understanding, this graph shows that for THIS system any investment made at the point of time when the DD was below 50% displayed superior returns 5 years later. This would suggest that making an investment when the system was at 50% DD would be better than making it when at 0% DD.

So can I exploit this when timing my investment once I have concluded that this is a system whose risk and other factors I can tolerate?
Essentially what you are suggesting is a Martingale money management strategy, applied to the equity curve instead of trades. Martingale strategies seem to work fine, until the one time when they fail. In the long run in trading these strategies do not work.
I dont think I am suggesting Martingdale using the equity curve at all. For one the Red line is not the equity curve of the system as traditionally viewed (explained above). I am also talking about timing an investment in the system - not about averaging down or up (as a Martingdale system does).

I am interested in the comment that 'these' strategies do not work? What strategies? Martingdale, Timing, Systems trading, systems that experience drawdowns?

Stephen

Asamat
Roundtable Knight
Roundtable Knight
Posts: 175
Joined: Fri Jun 03, 2005 7:50 am
Location: Walldorf, Germany

Post by Asamat » Mon Jun 15, 2009 2:50 pm

I did not misunderstand what the red curve is, you made it clear in the first post. My opinion about it is stated above, basically I think the red curve is something like a derivation on the blue one. It is kind of the difference between the blue curve 5 years ahaed and the blue curve at the point today, but expressed in CAGR instead of absolute terms.
I am also talking about timing an investment in the system - not about averaging down or up (as a Martingdale system does).
I'm not sure what difference you see in "averaging down" and "timing an investment after major losses".
I am interested in the comment that 'these' strategies do not work? What strategies? Martingdale, ...?
Yes, I meant Martingale type money management does not work in trading.

Anyway, my opinion is given, let's see what others have to add.

borderline
Roundtable Fellow
Roundtable Fellow
Posts: 65
Joined: Thu Mar 26, 2009 12:19 am
Location: Austin

Post by borderline » Mon Jun 15, 2009 4:46 pm

Would I put money into a fund after a drawdown? Well that "depends" but it would really only matter if it was a signficant drawdown and here are my thoughts on investing in funds with significant drawdowns:

Would I put money into a fund that has had a 50% drawdown? Only if I'm looking for 50% or greater returns, as my risk tolerance should be correlated in a nearly linear fashion with my reward for non-aribtrage strategeis (arb in literal sense - which btw I expect to be fleeting in the Information Age).

So if I'm not targeting 50%+ returns then I simply don't have the taste for a 50% drawdown and would personally consider that system death for the types of systems that fit *me*. My guess is you wont' find many "sophisticated" investors who find that it fits them either. When you already have a lot of money you are just as concerned with a return of your money as you are a return on your money. Compounding at 10-20% will do just fine.

Incidentally I believe an old trader/fund manager here in Chicago put together a published analysis of this martingale-like idea awhile back. He seemed enthused but on a cursory look I didn't see why - I don't recall the numbers being that much better.

It would seem that by trying to "invest" on the "pull back" you are trying to predict, and therefore live in, the non-existent future.

Best,
Borderline

Stephen Newton
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Dec 31, 1969 7:00 pm
Location: London, England and Cape Town, South Africa

Post by Stephen Newton » Mon Jun 15, 2009 7:25 pm

Asamat and Borderline

Thanks for your comments and thoughts. It is interesting to me that the observations you make relate to the absolute value of the DD vs the returns in the chart displayed. While these points are very interesting and of course very relevant to ones personal risk profile (and to potential investors) what I was seeking to stimulate some debate on was timing an investment in a system. To draw attention away from the numbers on the chart what would your thoughts on timing be if you took a zero off the % axis - ie Change the chart to have a circa max 9% DD and circa max 10% return as opposed to the circa 90% DD and circa 100%+ returns.

With regard to attempting to predict by using a measure today (in this case todays absolute DD as a %), I ask myself the philosophical question.... How is this different from looking at the Moving Average of Price Data today? It is a known variable that due to its actions over x periods suggests a buy. DD is the output of a system using this method. I know what it is today (like price) and if I know that returns are better (and certainly no worse than starting trading at a random starting point) should I not take this into account?

It seems to me that inserting a step to look at the 'best' possible time to start trading might be sensible. This is the hypothesis I am testing here...

You can adopt one of two approaches:

1) That timing is irrelevant and invest when your risk, return needs are met by the system OR
2) Attempt to determine if you can improve your odds by timing your investment based on known information (todays DD) - and its relationship to past performance.

My mechanism for trying to do 2 was to produce this analysis. For this system an investment made when the system was in a 50% DD (change to 5% if that works) was better than one made when the system was in 0% DD. That suggests to me that option 2 has some legs for this system.

Is anyone doing this kind of TIMING when entering new systems?

Once again thanks for the thoughts and ideas.

Stephen

DeanoT
Roundtable Fellow
Roundtable Fellow
Posts: 52
Joined: Sat Apr 15, 2006 7:37 am
Location: Melbourne, Australia

Post by DeanoT » Mon Jun 15, 2009 11:56 pm

By definition, any system which has stood the test of time, and is currently (or recently) making new equity highs, will appear to be an attractive candidate for timing the start of your trading the system, or adding new funds, at the time of historic drawdowns. But what about all those other systems which have not stood the test of time, and which have failed to recover past gains, and post new equity highs ? The theory of timing entry or addition of funds may not hold true for those systems.

The suggestion of timing your entry or addition of funds to a system to co-incide with historic drawdowns is, by its nature, inherently predictive, which in my experience, most successful traders do their best to avoid.

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

Post by babelproofreader » Tue Jun 16, 2009 7:47 am

As I understand it the hypothesis is that the current system draw down has predictive value for the 5 year CAGR figure 5 years in the future. It would seem to me that Monte Carlo testing is ideal for testing the statistical soundness of this type of statement. It is insufficient to rely solely on looking at a chart such as that given.

Stephen Newton
Roundtable Knight
Roundtable Knight
Posts: 127
Joined: Wed Dec 31, 1969 7:00 pm
Location: London, England and Cape Town, South Africa

Post by Stephen Newton » Wed Jun 17, 2009 10:56 am

babelproofreader

I totally agree with you. As a compliment to TBs built in Monte Carlo simulation, I do find this helpful.

Stephen

Gaia X
Contributor
Contributor
Posts: 2
Joined: Sat Jun 20, 2009 3:10 am
Location: Indonesia

Trading Blox Question

Post by Gaia X » Sat Jun 20, 2009 9:07 am

I has several question that i want to ask to Trading Blox user :
1. how much is different percentage of profit between simulation and real trading by using standard system or your own system in Trading Blox ?
2. Has any Trading Blox for stock market using other than in the list (like in other country) and where did your get the data ?

Thanks & Regards,

AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter » Mon Jul 06, 2009 1:08 pm

With regards to the idea that returns are higher after a large DD, Faith touches on this in WOTT when he talks about people proclaiming "The Death of Trend Following", which then tends to correlate with subsequent periods of better returns for trend following.

It does seem logical that after a large DD there would be less money chasing that particular (or related) strategy - the drawdown itself might be compounded by withdrawal from funds using this strategy, etc.

It's an interesting hypothesis, but as asamat points out it's quite tough to test without getting misled by biases..... hmmmmmmn
:?: :?: :?: :?:

Post Reply