Triple moving average...

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.
Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Triple moving average...

Post by Mike Cautillo » Tue Apr 17, 2012 3:49 pm

Hi Guys and Gals,

I have been recently testing a few systems and fairly new to back testing. Was wondering if anyone would be so kind to share with me some input, what are some of the most important numbers you should look at right off the get go when the results come back.

One of the systems I am testing is a great system but I find the trades take a long time to develop and exit...sometimes 2 years....how will I judge if I have sufficient capital to trade this system.

Lastly, looking at the total risk chart, it at times gets up near 60%..does this mean how much of my total equity is in trades at once?

Thanks,
Mike

illuminati
Full Member
Full Member
Posts: 24
Joined: Sun Apr 15, 2012 12:52 am

Post by illuminati » Tue Apr 17, 2012 5:24 pm

A few important metrics to look at would be MAR ratio which gives a look at the risk reward of your system.

the equation is simply CAGR/Max drawdown

Sharpe ratio is the next one, but I prefer sortino. You can google the equations online. These are static metrics, but they are good enough to start off with.

When you mention about triple moving average, i assume you are in the mindset of developing a trend following system. These systems in general depend on the winners that take time to develop.

I guess if you want to find something shorter in time frame, you should go look into other system ideas other than lttf.

hope this helps,
il

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Triple moving average...

Post by Mike Cautillo » Tue Apr 17, 2012 6:40 pm

Thanks for the help illuminati....I take it the higher the MAR the better...although I did reead in a trend follwoing article that hall of fame traders will achieve a Mar of .5-1.0....is this true. Thus far the beest I can muster up is anywher from .75-1.

I will look into the Sortino and I am definately a trend follower.

How a the total equity risk chart...the lower the better I presume??

What is the trade off...highe CAGR automatically results in a higher DD?

Again thanks.....cheers!!

illuminati
Full Member
Full Member
Posts: 24
Joined: Sun Apr 15, 2012 12:52 am

Post by illuminati » Tue Apr 17, 2012 7:42 pm

If you are referring to the total risk chart inside tblox, i am assuming that its a total risk you have but displayed in a rolling basis. Total risk in my definition is aggregating each open positions price to stop loss as a percent of total equity. So if all open positions were to be stopped out, thats the % of money you will loose.

Theres a whole debate on whats the best MAR. I suggest not going down that road. Its better to really combine multiple metrics to gauge a systems performance. Personally, I also new to systematic trend following. Been at it for a year and 2 months and still learning, but then from reading around, I find that a lot of system is rated based on static metrics which doesn't take in to account a lot consistency.

Back to your question about higher MAR, I think a tradable system would be around > 1.5. Anything above 2.5 in my opinion suffers from some sort of curve fit, but thats single system. If multiple system with aggregate with such high MAR, then I think you got a pretty good thing going.

The trade off is that you are looking to compensate yourself with a given level of risk. The high the DD, the higher you want your CAGR to compensate. As percent wise, it takes 100% to recover from 50% DD.

Don't take anything from me 100%. I am no more experienced than any other people new comer. But hey, everyone starts somewhere eh?

il

fab1usa1
Roundtable Knight
Roundtable Knight
Posts: 383
Joined: Sat May 21, 2011 8:30 pm
Location: New York

Post by fab1usa1 » Tue Apr 17, 2012 9:31 pm

Also consider that no one system performs well in all market types (ie bull/bear, volatile/quiet). For example, I know that my long-only stock system that uses 3MA LTTF works best during low volatility bull runs. It is performing exceptionally well since about December. I know that the good times will come to an end. That is when I will shut down this system and start up another that does well in volatile markets. I do rely on Market Type indicators to tell me when to make the switch.

I could be wrong but I believe that the pros do it differently. They have many systems and have access to more capital than we have. I believe that they run all systems concurrently, and do not attempt to time the market. But like I said I could be wrong.

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Triple moving average...

Post by Mike Cautillo » Tue Apr 17, 2012 9:38 pm

No worries Ill...aren't we all learning everyday regardless of our skill level. I have been learning the trend following method for about 6 months...I have been following Covel..read one of his books and bought his training.

Regarding the Mar....the crazy thing is the results you get with a MAR of even lets say .90 are very good when you look at expectation , profit factor and win %....I think that is the most challenging part is the balance between how much risk you are willing to take for what reward, but then again isn't that alway the case.

Any suggestions you have to obtain a 1.5 MAR would be great....thus far I have yet to see a mar that attractive....unless you ar looking for a low CAGR%...if you shoot for 20% and over and can obtain a DD of less than 20%..please show me.

Thanks for all of ur help.

illuminati
Full Member
Full Member
Posts: 24
Joined: Sun Apr 15, 2012 12:52 am

Post by illuminati » Tue Apr 17, 2012 9:57 pm

Switching between systems has been an idea for me for a long time. I've always considered Dalios idea invaluable. The true "holy grail" in achieving consistent return is through combining uncorrelated return streams. From my interpretation, its the idea of combining uncorrelated systems. A member of this forum did a piece of experiment through combining uncorrelated systems and you will see that as he introduces different uncorrelated system, the MAR goes up. (to give you (mike) some idea)

The big boys, like you said, have loads of cash to deploy. They do it in the most creative ways possible. The meat is in MM.

I read and crawl around the web a lot and if you look closely to all the ideas on this forum and some other talented individuals blogs, you'll definitely find a lot of uncorrelated strategy themes. Just to name one, momentum seem to pair with futures trend following quite well. If you follow the return on momentum and compare to standard CTAs like dunn, you'll see that when TF zags, stock momentum zigs. Example would be early this year. General CTAs have been doing not as well compared to indexes, but momentum has been doing quite good. Also, if you are like me and have ran too many backtests to count, you'll find that the period between 2004(3)-2006 standard TF systems did poorly, but it was splendid for momentum.

just my 2 cents,
il

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

Post by AFJ Garner » Wed Apr 18, 2012 3:51 am

The total risk chart in TB must be looked at in terms of the type of system and trade length you are considering. In a short term system you may find you want to restrict overall risk to (by way of example only) 10% or 15%. In a longer term system you may want to take risk up to 25% or 30%. In a short term system the damage can be done by a swift series of losing trades coming one after the other so the fact that risk is restricted at any one point in time to 15% does not mean that you can only lose 15%. In a longer term system losing trades come less swiftly on top of one another but of course you are standing there in the market and will take a big hit when markets move against you. It is therefore misleading to judge a system's risk by the the aggregate close to stop level alone. Nothing is that simple in this game.

As to MAR ratio it is an interesting pain to gain ratio but you should look closely at what MAR CTAs have achieved over the long term. Forum member DPH published some statistics on this forum showing that on average CTAs have achieved an MAR of 0.40. And that is without taking into account survivorship bias - in other words if you take into account figures from CTAs who have gone out of business and whose figures no longer figure in the statistics you would probably find the MAR much lower than that.

Put that in perspective. Take the desired return and divide it by 0.40 and it just might give you an idea of the maximum DD you may suffer in the future for a given level of return. Of course this is only an average; no doubt some methods of trading would achieve better MARs and some worse. Ditto better and worse skill levels.

A 10% return may be expected to produce a 25% drawdown
A 20% return may be expected to produce a 50% drawdown.

rabidric
Roundtable Knight
Roundtable Knight
Posts: 243
Joined: Mon Jan 09, 2006 7:45 am

Post by rabidric » Wed Apr 18, 2012 4:26 am

and just to add to AFG's words, MAR is quite sensitive to leverage, stray much above say 0.5Kelly and it is liable to catastrophic collapse(along with equity), but it is easy to overlook that running too low, and you will never benefit from the "risk ignition" sweetspot where the numerator of the MAR is appropriately boosted(more than the denominator) by compounded geometric returns.

Personally , in other threads, I have made the case that you need to adjust MAR for it's arithmetic bias, i.e. (1+CAGR)*(1-MaxDD) makes a lot more sense than straight CAGR/MAXDD. If you optimise leverage for my modified MAR, then you end up much nearer to the ideal backtest Kelly Fraction for that strategy(which you then moderate appropriately for walkforward deployment ofc) than you would with arithmetic MAR.

Evidently it is not all about MAR as has been stated already , but those are some considerations.
Don't even get me started on Sharpe!......

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Momentum indicator...

Post by Mike Cautillo » Wed Apr 18, 2012 8:09 am

Thanks to all for your posts...great feedback. Ill...yes this is true about running multiple systems simultaneously in an uncorrelated portfolio reduces risk but I think you can also manage this through a well diversified uncorrelated portfolio.

I think for me anyway, I need to perfect one system before I can run a multiple of them. So much data to interpret....sometimes don't you think too much??? I doubt Jesse Livermore had the abundance of data we had. You definately need to be able to filter a lot of it for what it's worth.

Good piece below on the Mar...it now gives me perspective on to what ranges Mar can vary in. As I said I did read- hall of fame traders range anywhere from .5-1 MAR.

As for the total risk chart...I tested a system for 20 years LTTF method, al the numbers look pretty good except for the total equity risk chart which did run to 60% at times but I think this has to do with the fact that winning trades are open for a long time-all the while others are being placed and losing. Any feedback on this.

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

Re: Momentum indicator...

Post by AFJ Garner » Wed Apr 18, 2012 9:17 am

Mike Cautillo wrote: but I think this has to do with the fact that winning trades are open for a long time-all the while others are being placed and losing. Any feedback on this.
Depending on your system you may find that your stop is very rarely hit. Risk is determined by TB on the stop. Say that you start with a 5 ATR stop - on a successful trade in a long term system you will find yourself way away from this stop and hence "risk" on that trade for betsizing purposes may "artificially" inflate. You may want to look at an additional wider trailing stop and assess risk on the greater of the initial stop or the trailing stop.

Bear in mind one other point on CTA MAR which I forgot to mention earlier. When looking at an individual CTA, you need to look at different time periods to get an accurate gauge of return versus drawdown. Take one simple example: Winton. Winton's max DD is 25% but that dates from the days when Harding was shooting for 20% returns. He is now shooting for returns of closer to 8% (or somesuch) and therefore when looking at the MAR you probably ought to divided Winton results into pre and post this decrease in volatility/return target. Ditto many other long running CTA who have inevitably made huge changes to their programs over the years. You can't just look at MAR for the entire period of their existence and draw any meaningful conclusion.

View back testing in the same way if your system alters over the years of the back test. Do you bet higher when capital is lower and reduce betsize later in the test? Your portfolio - is there data for every instrument in the portfolio from the very beginning of the test? If not, then you may need to consider the effect of an increasing portfolio as data comes on stream from new instruments.

The fun is endless and the conundrum can not be solved by science, statistics or maths alone. Judgement, skill, forecasting and plain common sense are all required in equal measure.

Beware the drawdown. They are not amusing. Do not over estimate your tolerance.

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

AFG on Total equity risk (TER)...

Post by Mike Cautillo » Wed Apr 18, 2012 10:40 am

So regarding increasing the stop to obtain a better perspective on the total equity risk....I did 5ATR and we get 60% TER max. & 17% CAGR and 26% DD...I then bumped it up to a 20ATR stop and got 30% TER max- 6.6% DD and 6% CAGR...I then put a 1ATR stop and got a 80% TER max-a 54% DD and a 22.8% CAGR.

I am assuming when the stop is lower it actually closes more trades and as a result will increase draw down....if you could further explain would be great.

One more point is that the lower the stop...the lower the win percentage....this strikes me as being obvious...is it so??

fab1usa1
Roundtable Knight
Roundtable Knight
Posts: 383
Joined: Sat May 21, 2011 8:30 pm
Location: New York

Post by fab1usa1 » Wed Apr 18, 2012 10:55 am

Mike, try adding a blox called Total Risk Limiter. Run a simulation with steps from 5% to 50% in increments of 5%. Afterwards look at the shape of the MAR and CAGR curves. You may discover that there is a "sweet spot" where the MAR and CAGR are maximized as well as stable. For my long-only stock system I run with a max risk of 30% with an initial stop of 5ATR.

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

Post by AFJ Garner » Wed Apr 18, 2012 11:09 am

All I was trying to say is that if you have a wide stop in a long term system which is set at initiation and held but does not move, then "risk" as per the TB risk chart is probably exaggerated.

illuminati
Full Member
Full Member
Posts: 24
Joined: Sun Apr 15, 2012 12:52 am

Post by illuminati » Wed Apr 18, 2012 11:16 am

AFJ, MIke and all, apologies for my misleading points on TB risk. I don't use it as my main testing package so am a bit rusty on it's features.

Mike in terms of limiting risk, Fab1usa1 is correct about introducing a total risk limiter. It's where you cap total open risk and avoid additional positions when you are at the limit. I in terms disagree about finding the sweet spot as it's sounds rather curve fitting. Rather, you should lay out all the values and find the one that is providing the most consistent return over the years and choose that one.

MAR ratios and all the statistics are just there to give you a glimpse of risk reward. I agree that nothing ever will reflect your testing when you start to trade real life. But nevertheless I believe that a system should be at least be good on historic data before it should be traded given you haven't curve fit. I don't think that one should avoid developing good systems with the knowledge that their good systems risk reward will ultimately converge to the 0.4 level.


Il

rabidric
Roundtable Knight
Roundtable Knight
Posts: 243
Joined: Mon Jan 09, 2006 7:45 am

Post by rabidric » Wed Apr 18, 2012 12:06 pm

yes, what fabusa wrote. there is a good thread somewhere around here dedicated to that discussion...
illuminati wrote: I in terms disagree about finding the sweet spot as it's sounds rather curve fitting.
The sweetspot can be pretty wide, so not much of a spot, but more of a large thoroughfare, with the river on one side where you could get swept away as a weak swimmer, and a minefield on the other, where you may get blown up. :P
illuminati wrote:But nevertheless I believe that a system should be at least be good on historic data before it should be traded given you haven't curve fit.
And therein lies the crux of the matter. That sentence encapsulates the whole paradox of Systematic trading. 8)

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

Post by AFJ Garner » Wed Apr 18, 2012 12:14 pm

rabidric wrote:
illuminati wrote:But nevertheless I believe that a system should be at least be good on historic data before it should be traded given you haven't curve fit.
And therein lies the crux of the matter. That sentence encapsulates the whole paradox of Systematic trading. 8)
Yes indeed. Never a truer word................

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Post by Mike Cautillo » Wed Apr 18, 2012 1:16 pm

I will try this fab....this is a very good idea....as I do want to evaluate the effects of total risk limits on my portfolio...I imagine this will have a noticeable impact.

Just want to throw this out there....how are you all going about calculating the required capital to start trading your system??

Once again ....thanks for everyone's help.

Allow me to share this with you, I think it suits aspiring traders....

“Character cannot be developed in ease and quiet. Only through experience of trial and suffering can the soul be strengthened, ambition inspired, and success achieved.â€

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

Post by AFJ Garner » Wed Apr 18, 2012 1:26 pm

[quote="Mike Cautillo"]I will try this fab....this is a very good idea....as I do want to evaluate the effects of total risk limits on my portfolio...I imagine this will have a noticeable impact.â€

Mike Cautillo
Senior Member
Senior Member
Posts: 27
Joined: Thu Mar 15, 2012 2:59 pm

Post by Mike Cautillo » Wed Apr 18, 2012 1:40 pm

AFJ...are you saying that you will need to establish how and what markets are traded when the the total equity risk has been hit....if so would it not be on a first come first serve basis....random??

Post Reply