Mechanical System Traders - How Long Did You Take To Get Your First System?

[TWI]

Did you look at the return volatility vs the number of commodities for this portfolio?

The system was the same for each commodity the only changes being adjustments for unique characteristics such as price ranges and roll-overs. It was all EOD. I looked at various combinations of portfolio as the products were split into sectors of Grains/Softs/Energy/Metal but worked basis the fact that we needed to include all these sectors to obtain something that would be saleable to funding. Spent a lot of time working on sector weightings and individual product weightings within the sector and ended up with something similar to the components of GSCI.
One of the frightening things was that shortly after we decided that we would move on due to lack of real interest (only interested investor was from US), it took an 18% draw, that would have been our first 3 months, so would have been very nasty and maybe we wouldn't have made it anyway. A shame as this year it jumped on the energies.

it does start to get difficult to trade with so many different systems

This has been a real problem because I spend most time concentrating on the STIRS. I have concluded that automated execution is the only way to minimise slippage and the extra bro cost is more than made up for by this.

 

[Denny]

Mechanical Investing

Tufty,

"Denny said trading / Investing is an art rather than a science and if mechanical systems worked then the designer would be a billionaire within months. Following this logical argument through then wouldn't the most artistic person be a billionaire within months too"


Sorry I cannot accept that logic at all.

Science is built on fundamental principles which cannot be changed (unless disproved of course) the law of physics for instance. Following a certain path using these laws will always produce the same result as with mathematics.
Art is different as there is no definitive right or wrong as it is mostly subjective to opinion.
Probably the reason why nobody ever fails arty subjects at university providing they just turn up.

Denny

 

[jmreeve]

The system was the same for each commodity the only changes being adjustments for unique characteristics such as price ranges and roll-overs. It was all EOD. I looked at various combinations of portfolio as the products were split into sectors of Grains/Softs/Energy/Metal but worked basis the fact that we needed to include all these sectors to obtain something that would be saleable to funding. Spent a lot of time working on sector weightings and individual product weightings within the sector and ended up with something similar to the components of GSCI.

Quite apart from system design, portfilio construction and weighting is an art in it's own right.
Would be interested in you experiences as to how you approached this. My feeling at this point is that the drawdowns of individual systems should be equalised across the portfolio. However, this raises the question of how do you measure the draw-down of a system in a robust way?

This has been a real problem because I spend most time concentrating on the STIRS. I have concluded that automated execution is the only way to minimise slippage and the extra bro cost is more than made up for by this.

I have put a considerable amount of effort into automation for similar reasons.

 

[Tuffty]

Denny, You obviously have strongly held beliefs which means that for you mechancial trading won't work. For me its just a methodology to exploit the markets. From what I see the methodology is just as discretionary as any other trading (i.e. you make up the rules, choose the markets, choose which systems to trade etc).

I used to do 'pure' discretionary trading based on intuition, fundamentals, and technicals. However developing trading ideas where the belief is based on statistics and probabilities fits my personality better. Unfortunately its taken me a long time to realize this and it was the Turtle's story combined with noting certain market characteristics that inspired me to look again at systems trading.

 

[TWI]

I have put a considerable amount of effort into automation for similar reasons.

What have you found the best solution for this?

Quite apart from system design, portfolio construction and weighting is an art in it's own right.
Would be interested in you experiences as to how you approached this. My feeling at this point is that the draw-downs of individual systems should be equalised across the portfolio. However, this raises the question of how do you measure the draw-down of a system in a robust way?

Needed to know we could move at least $60mln through the system which meant some of the less liquid commodities could not have too much volume.FCOJ and Oats being prime examples.
Worked on assumption that if lowest volume commodity had 1 lot what would that mean for others based on GSCI and CRB and scaled it up accordingly. Tried that for start. Then split up each sector and ran equity curves for compositions of each sector based on liquidity considerations. Ended up between the two and tended more towards GSCI composition with the provision that certain markets would never exceed more than a certain lottage. There are more quantitative ways to approach this but this way worked out OK.

 

[jmreeve]

Worked on assumption that if lowest volume commodity had 1 lot what would that mean for others based on GSCI and CRB and scaled it up accordingly. Tried that for start. Then split up each sector and ran equity curves for compositions of each sector based on liquidity considerations. Ended up between the two and tended more towards GSCI composition with the provision that certain markets would never exceed more than a certain lottage. There are more quantitative ways to approach this but this way worked out OK.

Interesting approach based on liquidity. What made you choose to structure the portoflio along the lines of these two indices?

Re:automation, I have ended up coding my own. The actual automation and order routing side of things is relatively easy. However, the problem of getting very reliable, very clean real-time and historical data for a wide range of markets into multiple trading applications is truly fearsome. There are comercially available systems but these are aimed at hedge funds and are priced accordingly.

 

[TWI]

Interesting approach based on liquidity. What made you choose to structure the portoflio along the lines of these two indices?

Because they existed I think is the best answer. When we got to the allocation part that is what we knew about, so it seemed a sensible starting point. All the portfolio stuff I had done previously was due to allocated capital in limited portfolios or single markets so I did not have experience of solving this problem. I had had a bad experience with the oats market a number of years before when I sent it limit down getting out of a long "at market" so knew, all to well, those sorts of issues.

Have you done much work on this?
I could not find a lot that made sense last time I looked just found a few CTA's that published the spread of exposure across commodity sectors.

 

[jmreeve]

Have you done much work on this?
I could not find a lot that made sense last time I looked just found a few CTA's that published the spread of exposure across commodity sectors.

I have not done a vast amount of work. My starting point was that if short term equity variations were normally distributed and had no correlation then equalising the equity variations between systems would give the smoothest equity curve. However, in practice there is some correlation and the variations are not normally distributed.

The attached paper gives an approach I quite like as it does not make many assumptions and is useful in practical portfolio design.

 

[Denny]

Mechanical trading

Tufty,

I am not sure we are that far apart on this argument actually as pure trading on a "wing and a prayer" is not my idea of trading either.
I have set rules also for entering and exiting trades but without definitive values which. I tend to trade with the longterm trend and wait for a pullback to the short term mean. However the entry point especially is based on the graph rather than exact values. I particularly like triangles and breakouts from the triangles as it seems to have worked for me as has MACD divergence. I get it wrong often and get out when I am. However I know that I could not adhere to a strictly mechanical system because I could not handle the stress of major drawdowns and not knowing whether it was a drawdown or that the system was simply not working anymore.

Each to his own I say and as long as it makes money that is fine, whatever the method.

Denny

 

[picker]

Hi, anyone mech system traders wanna share stories of their startups?
I'm a fledging newcomer trying to absorb all the information overflow i have gathered attempting to trade mechanically these 3 mths. Thankfully my tuition fee was small (US$500).

How long did it take for you to get ur first system up?
Countless years of number crunching or just a couple of months? Any sob stories?

Trend following may be for you, if you have the discipline and money management rules!! Try one or two moving averages and see which fits your personality!!
hope this helps "picker".It will take you about two years to learn then you will need to capitalise your account with serious money to be a trend follower for a living.

 

[meanreversion]

This is an interesting thread, even if I did get to it 5 years late. In my experience, the vast majority of active traders (as opposed to buy and hold stock investors) like to get in and out of positions fairly frequently and prefer occasionally being flat. Trend following suffers in that regard as you might be in a trade for six months, during which time equity swings are unpleasant. Try trading on a shorter time frame (1 - 4 hours) and if the market is fractal, then the "trend" might last only a few days using a short time frame.