Traders' Roundtable

Hello everybody, I am pretty new to systems trading but I have realized there is many experience people around here. My question is regarding the system I am testing. I have been playing around with parameters, risk and money management settings, etc, what is a realistic/good MAR I can aim for?

My system is medium/long term trend following. I am looking for robustness and there is a fairly long set of parameters that give me MARs between .7 and 1 in a 20 year fairly diversified set of futures in backtesting. Is it this a realy good measure knowing the actual DD are going to be worse?

after fees, generating a MAR of 1 would have you breathing very rarified air.

I should clarify, in actual trading, over a decent number of years, in a scalable program, after 2 & 20, acheiving a MAR of 1 would place you among elite traders. In backtesting, I think it is fairly easy to come up with a MAR of 2 or 3, depending on how good you are at fooling yourself.

Do most people feel the same as RGD? This is an interesting topic because I am sure that answer is different for many traders.

edit:

heres an attachment of a long term trend following system of mine.

Hi Trackstar,

It is important to note that those are not my feelings, but observable facts which can be discovered by anyone willing to invest a few hours perusing CTA databases.

I do know traders who use non-scalable, short term strategies that achieve high MAR & sharpe ratios, but the strategies are constantly changing and seem to have a limited shelf life as other quants exploit the same pattern. I do know a couple discretionary traders who achieve exceptional numbers as well, but they too are not scalable, and rely heavily on the intuition of the trader. Such skills seem to be far more scarce than those who are able to create robust, systematic strategies to trade futures markets.

rgd wrote:after fees, generating a MAR of 1 would have you breathing very rarified air.

On a dollar-weighted basis, without adjusting for survivorship bias, I consistently come up with an average MAR ratio of 0.65 for reporting CTA's. It's closer to 0.75 if you equal weight instead.

Using the 50 longest running track records for reporting CTA's yields an average MAR ratio of 0.46. Using the 50 shortest track records (but with at least 12 months of data) yields a MAR ratio of 1.9.

HI rgd,

That is more or less what I was thinking. A ratio of 1-1.2 MAR for backtesting should be OK since the real (future) number will be somewhere around .7 which would make me very happy (if I can stick to the system). I say this because from my experience with CTAs track records, a .8 MAR is indeed "elite" (because of the long track record and 2-20 fees). I just wanted to know what was a realistic MAR for Backtesting considering all that, I think I am pretty good at not fooling my self becuase the system has very few parameters and they seem to be robust (the performance is pretty similar across a wide range). Actually following many Market Wizards advice I am trying not to play with the parameters and try to improve MAR through money and risk management. Thank you.

ecritt wrote:Using the 50 longest running track records for reporting CTA's yields an average MAR ratio of 0.46. Using the 50 shortest track records (but with at least 12 months of data) yields a MAR ratio of 1.9.

Would be interesting to know what the 50 longest and 50 shortest track records MAR ratios were over *JUST* the last 12 months.

Not that I'm a fan of MAR as a criterion, just that I can imagine several possible contributing factors to the difference between longest and shortest track record MAR, and limiting to the last 12 months would eliminate one possible factor.

nodoodahs wrote:

ecritt wrote:Using the 50 longest running track records for reporting CTA's yields an average MAR ratio of 0.46. Using the 50 shortest track records (but with at least 12 months of data) yields a MAR ratio of 1.9.

Would be interesting to know what the 50 longest and 50 shortest track records MAR ratios were over *JUST* the last 12 months.

Not that I'm a fan of MAR as a criterion, just that I can imagine several possible contributing factors to the difference between longest and shortest track record MAR, and limiting to the last 12 months would eliminate one possible factor.

Yes, if a CTA has been around for 30 years and initially traded with high leverage / low AUM and now trades low leverage / high AUM, they could have a lifetime MAR of 0.5, even though their MAR on a rolling basis could be greater than 1. This is why the Calmar ratio exists.

LeviF wrote:Yes, if a CTA has been around for 30 years and initially traded with high leverage / low AUM and now trades low leverage / high AUM, they could have a lifetime MAR of 0.5, even though their MAR on a rolling basis could be greater than 1. This is why the Calmar ratio exists.

Could also be a function of market conditions or LTTF strategy performance in the last 12 months, as compared to the last decade or more. [Edit for clarity, I want to stress the change in market conditions in which the long-running programs operated then Vs. now, rather than stress the change in leverage level under which those long-running programs operated then Vs. now.]