Triple Moving Average vs. All The Others

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.
TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Triple Moving Average vs. All The Others

Post by TrendMonkey »

I have been fiddling around with the VT TMA system for quite a while now and it seems that no matter how much I mangle the various parameters, the numbers look "Okay". Meaning, it's tough to produce a negative CAGR with this system and the MAR is more often than not decent.

On the other hand, for the other 5 VT systems, while I can get some great CAGR/MAR numbers, they are usually immediately surrounded (in terms of parameter steps or portfolio selections) by ruinous ones. Furthermore, the other 5 almost always generate lots of negative CAGR's whereas as I said above TMA hardly ever does.

So I am edging towards actually putting on some trades using TMA. Just wondering if I am missing some key piece of the puzzle here.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

If you're thinking about trading real money, you may wish to think about testing with real data. For example, if your experiments have been with the Veritrader-presupplied Forex data, you may wish to become a customer of a Forex data seller and purchase the Forex data that you'll actually be trading with, including real time updates of today's prices.

Maybe you'll find that (your data vendor's data) gives different results than (the presupplied Veritrader data). If so, you'll probably wish you had discovered this important fact before you start trading rather than after. Maybe you'll even find that your enthusiasm for one trading system vs. another, is modulated by your choice of data.

I don't mean to criticize the free data that Veritrader generously supplies. Rather, I am merely reminding you that you won't actually be trading based on that data. You'll be trading based on data you buy from Worden Brothers or Pinnacle or Reuters or CSI or their competitors. Maybe you'll trade based on no-cost data from Yahoo Finance or MSN Money. But in any case, trading is not 100.0 percent identical with horsing-around-on-Veritrader-straight-out-of-the-box. Obviously, you already know this. Consider it a gentle and friendly reminder.

Choice of data source is just one way, among many, that trading differs from playing with Veritrader straight out of the box. Commissions, slippage, and interest are three more. Once you begin trading you'll discover (painfully) a few others.
TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey »

Its funny you should mention that because that (testing with more data than comes free with VT) is the last thing on my list of stuff to go through. I am actually most concerned about the effects of different backdating algorithms, of which I think UA has a handful. Next week's project.

Thanks for the feedback & have a great weekend.
JAM
Senior Member
Senior Member
Posts: 41
Joined: Wed Nov 17, 2004 4:25 pm
Location: Toronto

Post by JAM »

I actually found the same thing when I was testing with Veritrader straight out of the box, although I also found Dual Moving Average systems also worked well.

I do think moving average systems can be good systems, if you have the stomache for the drawdowns...ie a CAGR of 60% is fantastic, but often comes with a similar drawdown for these systems (I haven't seen many of these with a MAR much greater than 1.0). When its actually your money on the line, your stomache isn't as strong as you think is in testing, in my experience.
stancramer
Roundtable Fellow
Roundtable Fellow
Posts: 60
Joined: Sat Feb 07, 2004 10:34 am
Location: Washington DC

Post by stancramer »

Using the All Liquid portfolio of data that comes "out of the box" with Veritrader, I get MAR's 1.5 to 1.75, not 1.0. An example using the Dual Moving Average system produced the stats here. Standard Veritrader commissions and slippage as recommended by DanG were used, including zero interest.

MAR was 1.60 and max drawdown was 37%. Hunt around, good parameter settings are Out There.
Attachments
twoma.png
twoma.png (21.87 KiB) Viewed 22588 times
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Both the triple and double MA work as well with a wide variety of "real" data as on the pre-supplied data over a very wide range of parameters. At least in my tests. At least in my opinion. It is most instructive to test individual contracts on a single contract basis in a non portfolio setting, without money management, before moving on to portfolio testing. Not only do many combinations of both double and triple MA trade a much wider range of instruments successfully than , say, the BB Breakout, but the curve (even on single instrument tests) is often a) a lot smoother and b) performs as well or better in more recent history as it does in the distant past.

I mean at some stage to quantify the "goodness" of such slopes by calculating std deviations etc and coming up with a very large list to put the proposition to myself in a more statistical and less off the cuff basis. To more accurately judge the differences between the systems.

Not so with manay parameters of the BB Breakout although this can sometimes be alleviated by widening out the stops and going for longer term averages. Reasonably long term MA systems also seem far, far better able to cope with some of the notoriously difficult commodities (as opposed to financials and energies). The ability to trade a wider range of contracts successfully seems a pretty important asset. There are of course contracts where the reverse it true: the curve looks rather better with a BB Breakout at 80/2 on Crude than it does on many moving average combinations. But I have found the reverse in many if not most situations.

Portfolio testing also seems to show, over many portfolio combinations, an accelerating equity curve rather than a decelerating one over the past decade for many longer term MA combinations. Not so with the BB breakout.

Adding money management does of course greatly complicate matters. To name but a few issues you need to use some measure external to the MA crossover for betsising. ATR stops for instance: if you set that stop wide enough (15 for instance) on a reasonably long term MA crossover the stop will rarely get hit but you will need a lot of startup capital. If you go too small (say 2 on a very long term system) you will get some horrendous hits in terms of R Multiple and often go bust.

The MA systems in Trading Blox differ from VT in that the stops are not used as exits. In VT , with use ATR stops set to "true", the stops were executed but the next day, if the MAs were still stacked in the right way, you got right back in, albeit usually with less contracts. Thus the ATR stops acted as a sort of "re-sising " mechanism.

It is important to test other methods with MAs: for instance using MAs as real stops (as in VT) not just sizing mechanisms and not getting back in until a new crossover etc etc.

I would also point out for those who have not seen it the simplicity of the trend following algorith used in the calulation of the Mount Lucas Index www.mtlucas.com . Price crossing a 265 day moving average, monthly dealing and re-balancing. And it has produced a very decent equity curve and return even at the ungeared level. The company manages a large proportion of its assets by replicating this ultra simple mechanism.

So, to cut a long story short, yes, in my opinion also simple MA systems have a very great deal going for them. But then, as we all know, the future has an uncomfortable way of messing up the most meticulous of research based on data from the past.
TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey »

StanCramer, my VT-in-a-box didnt include an 'All Liquid' portfolio. Do you have a different version or something?

I have fiddled around with a variety of portfolios and although for sure this can have a big impact, I have found that any reasonably-diversified portfolio works OK. Of the four in-a-box portfolios, I found C&F was significantly worse (but still "OK") and the other three were better and roughly similiar.

Anyhow, thanks for the posts, one and all.
TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey »

AFJ, may I ask: when you test individual contracts and so on, are you doing so by hand? Or are you using multiple testing products (eg. not just VT?)
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Triple MA

Post by AFJ Garner »

I am using Trading Blox. I also use Trading Recipes.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

SC, the All Liquid portfolio (filename All Liquid.set) was prepackaged with the beta software, not the full release Veritrader. Fortunately it draws upon the same market data that is shipped with regular non-beta software, so anyone can experiment with it if they like. The composition of the All Liquid portfolio is below. Just click Edit -> Edit Portfolios -> New Portfolios and then select these markets:

AD BP C CC CD CL CT ED EM EU FC GC HG HO HU JY KC LC LH MP NG S SB SF SI TY US W
Tim Arnold
Site Admin
Site Admin
Posts: 9015
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold »

The All Liquid futures portfolio is included with Trading Blox Builder. I have enclosed the .set file here. You can put this in your Futures Sets folder if you don't have it.
Attachments
All Liquid.set
(109 Bytes) Downloaded 1262 times
TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey »

Thanks a lot Tim! (and SC, AFJ etc.)
pakua42
Full Member
Full Member
Posts: 23
Joined: Thu Dec 01, 2005 12:35 am
Location: Florida

Post by pakua42 »

Can someone tell me what the trading blox triple moving average system has done this past 2005? Also, can anyone tell me which instruments would work better on 4 hour or 2 hour time frames or anything other than daily.

thanx,
marc
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

Recommendations for user pakua42 are found at this hyperlink
pakua42
Full Member
Full Member
Posts: 23
Joined: Thu Dec 01, 2005 12:35 am
Location: Florida

Post by pakua42 »

thanks sluggo. However I used omni trader with great fascination. Trade station was too complicated and unenjoyable. i have managed to teach myself how to trade equities very well without learning excel and coding. If i buy trading blox and i feel it will be worth learning code, i will do it. but i just want to win at the futures game, not tweak a system for a year to get an extra 5%.

All I want are return % numbers for the trading blox system for the default settings for turtle and triple moving average. Why is this a big secret? I appreciate all feedback and will keep asking questions because I am hungry. -)

marc
JAM
Senior Member
Senior Member
Posts: 41
Joined: Wed Nov 17, 2004 4:25 pm
Location: Toronto

Post by JAM »

Okay, I will bite. How much is it worth to you for me to peform these calculations on the software I purchased? I am more than willing to do as many calculations as you like, for the right price.

Alternatively, you could just buy the software yourself...but if you want to pay me, I am more than happy to oblige.
Tim Arnold
Site Admin
Site Admin
Posts: 9015
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold »

Hi pakua42,

Keep the questions coming -- best way to learn!

I've attached the results for a test run of the Triple Moving Average system on 28 futures for 2005, using a variety of parameters and assumptions. The value of this information is limited, as you have no way to verify for robustness without a testing platform.

Please note that Trading Blox is a strategy develoment and testing platform, we do not sell systems, and we do not have default parameters. We offer many public domain systems in our package so that people can learn how all the various test assumptions interact. It's important when looking at systems to really get under the covers if you are going to have the confidence to trade through any sort of drawdown.

Hope this helps!

Tim
Attachments
tma2005.jpg
tma2005.jpg (57.6 KiB) Viewed 21157 times
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

Tim, Tim, Tim. I keep trying to gently suggest that you include non-US futures in your testing portfolio, but it appears I still haven't succeeded.

I find that when I run the 3MA system for CY2005 using the same number of markets ( 28 ) as you ran, but employing an international portfolio of futures markets, the results are noticeably better. This portfolio includes quite a few overseas instruments, such as CSI Commodity Numbers 58, 131, and 565. They are traded in frightening, exotic locations like the steamy jungles of Winnipeg, Frankfort, and London. Oooohh, scary. :)
Attachments
3MA system settings
3MA system settings
params.png (30.45 KiB) Viewed 21124 times
2005 results
2005 results
eqc.png (24.68 KiB) Viewed 21121 times
Tim Arnold
Site Admin
Site Admin
Posts: 9015
Joined: Tue Apr 06, 2004 1:41 pm
Location: Boston, MA
Contact:

Post by Tim Arnold »

Great point, Sluggo, thanks for that.

I keep to the usual US suspects for the benefit of our customers that don't have access to world futures in their CSI subscription.

But perhaps your point is that we should do more to encourage our customer base to pay a few extra bucks for the data, and look beyond our gentle shores for opportunities.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Fair enough, Sluggo, Tim. But I wonder what Pakua 42 will feel he has gained? I wonder if the notion "too good to be true" might prompt him to ask himself (or indeed you) a few further questions. Or ask you to run a few further tests maybe.........
Post Reply