Is 100 percent mechanical trading possible?

[forker]

Would that be emotional attention or maintenance and tweaking?

 

[the hare]

Would that be emotional attention or maintenance and tweaking?

Its mainly ongoing maintenance and data analysis. I'm not really at the stage where I trade a single strategy that I'm 100% happy with, which means I'm trading multiple instances of systems spread across multiple brokers, and that increases the workload.

Establishing reasons why a strategy is performing better or worse than expected in particular market conditions takes quite a lot of analysis and really becomes a process of ongoing research.

I dont really have much of an emotional attatchment to results, if I could get away with it I'd probably only check once a quarter, so in that aspect at least mechanical trading has its advantages.

 

[forker]

Do you know what you're working towards or are you changing things around as you think best in each instance and hope for better results?

I have to admit I don't think I could ever automate the way I trade even with savvy programmers at hand

 

[the hare]

Do you know what you're working towards or are you changing things around as you think best in each instance and hope for better results?

I have to admit I don't think I could ever automate the way I trade even with savvy programmers at hand

Yeah, the basic model which I've been trading for almost 8 years now has always been the same. A lot of the grunt work is simply establishing that everything is running OK and within expected parameters.

I came to the conclusion a while ago that if you are trading based on TA or statistical analysis there's a fair amount of random chance which effects the final result, and therefore any sort of walk forward optimisation of model parameters is practically useless. I've even seen proprietry research that suggest that an effective edge could be developed by fading an optimised parameter set ! Thats one of the reasons for the diversification, which of course brings its own set of problems trying to understand how best to dynamically allocate capital (an area I'm really struggling with to be honest).

The approach I use for automated trading is almost a complete polorisation of how I previously traded, but I suspect over time there may be elements I'd like to incorporate.

 

[forker]

Cool, thanks for sharing..

 

[numbertea]

The only statistical analysis I apply to my trading system I'm developing is comparing the winners to losers and the acceleration of price while I am in trades. My system is based completely on how I physically look at and interpret price movements on the actual charts, I programmer my system to replace my eyes and do only what I would do when I have a clear head. Of course the computer does it much faster so much more trade opportunity is afforded. I have given up a long time ago on oscillators and other fancy indicators.

 

[Adamus]

I've even seen proprietry research that suggest that an effective edge could be developed by fading an optimised parameter set !

It sounds like you've got a lot of experience in mechanical trading and that bit I'm quoting above is fascinating. Of course everyone always asks if you (I mean me) tried fading a system you wrote that has bad results, but I never took the question seriously.

Thinking about it now, do you remember whether the research was looking at just a random system that the researchers had developed themselves, or was it targetted at any particular widely marketed trading system advertised to make millions on the back of curve-fitting to the last year?

 

[the hare]

Thinking about it now, do you remember whether the research was looking at just a random system that the researchers had developed themselves, or was it targetted at any particular widely marketed trading system advertised to make millions on the back of curve-fitting to the last year?

It wasnt a random system (the opposite of random is of course random, which leads to similar results)

I do know that there is information in the public domain from others who have taken a similr approach. The guy who wrote the wave59 charting app has certainly published a couple of articles based on this kind of approach.

 

[Ferru]

I am yearning for a 100 % mechanical trading system that is profitable?. Is this possible?

Yes - a guy called John W Henry bought the Boston Red Sox and Liverpool FC on the back of purely mechanical trading!

I'd recommend a book called Trend Following, by Michael Covel. It actually gives examples of statistically backtested mechanical systems.

Jeff

 

[Adamus]

No system is 100% mechanical because a human is always making decisions - with trend following systems, you have to decide when to scale back or even turn off the system in times of severe drawdown.

 

[the hare]

No system is 100% mechanical because a human is always making decisions - with trend following systems, you have to decide when to scale back or even turn off the system in times of severe drawdown.

I disagree. A key element of sytem design (ad indeed ongoing analysis) is establishing the point at which a system should cease trading. That sort of decision definately shoudnt be discretionary.

 

[Ferru]

No system is 100% mechanical because a human is always making decisions - with trend following systems, you have to decide when to scale back or even turn off the system in times of severe drawdown.

Hi Adamus

With a properly designed and backtested system (ie one with a proven edge and a safe staking plan), any severe drawdowns can be confidently attributed to bad luck. Therefore, if you get severe drawdown with such a system, then surely the best approach is to carry on plugging away, and let mean reversion take its course.

Also, surely High Frequency Trading is proof that profitable mechanical trading is possible, given that computers can make literally millions of trades a day and come out with an overall profit. OK, I know those servers can access the market in ways your average trader at home can't, but it shows that human judgement isn't always a prerequisite for successful trading.

Jeff

 

[Adamus]

I disagree. A key element of sytem design (ad indeed ongoing analysis) is establishing the point at which a system should cease trading. That sort of decision definately shoudnt be discretionary.

OK in an ideal world I guess you can program your software to implement those decisions ahead of time but they are still decisions that have to be made, whether before or during. Of course if your system shuts itself down automatically, you also have to make the decision when to turn it back on.

@Jeff: I have no experience or inside knowledge of the Goldman Sachs / Citibank / etc HFT systems. I would assume though that there is a high level of human supervision.

But don't get me wrong, I totally agree that theoretically purely mechanical trading is possible, with all human interaction carefully programmed in - but I'll tell you this for sure, it's not done with NinjaTrader or Interactive Brokers or by small retail traders constantly developing their systems to leverage limited trading capital.

 

[DDI]

I am going through programming now hoping to automate my trading system one day..

I believe that if (big IF) I manage to code my strategy successfully and the computer trades it accurately without any bugs, the system would outperform me.

If after a while its in the red constantly, I would have done worse manually.

It would first eliminate human discipline problems. When I sit all day by the PC and I see no valid set ups, I just start trading on invalid signals, particularly when I am frustrated seeing valid signals missed overnight (I trade mainly forex). The computer should address both these problems.

I have faith in my system and if I wanted to judge my performance on valid signals over the past few years my results have been good.

However, having said that I do have some concerns that linger in the back of my mind:

1) I see people who are experts on programs (and T/A) charging hourly rates to teach. I automatically think surely this guy has enough knowledge to set up his own blackbox and make lots of money hassle free instead of chasing $100 here and there (or selling indicators).

2) Throughout my trading life, whenever I have spoken to those that trade professionally for a firm, I notice they hardly use (or in some cases havent even heard of) technical analysis methods/chart set ups.

3) Finally a computer sitting in the corner of the room and making money while one can live a leisurely life, sounds too good to be true. And when something sounds too good to be true, it usually is!

 

[Ferru]

It does, but the fact is that some of the most successful traders in the world do precisely that. Sometimes the simplest solutions are the best solutions...

I'd recommend you read Michael Covel's book Trend Following - it focuses on purely mechanical trading, and shows how people have made fortunes totalling hundreds of millionds of dollars with that approach. He's also written a book about the Turtles, a group of novice traders who were taught a 100% mechanical system.

The difficulty with non-discretionary trend following is mainly psychological - 2/3rds of your trades will fail, and you need the stomach to stay focused and disciplined in the face of big drawdowns.

Jeff

3) Finally a computer sitting in the corner of the room and making money while one can live a leisurely life, sounds too good to be true. And when something sounds too good to be true, it usually is!

 

[DDI]

Thanks Jeff,

will definitely read it.

but by mechanical system do we mean systems that use price action set ups (or/and based on T/A) or complicated mathematical models.

If both then it seems the latter is more popular as there are billions invested in them whereas, as I said before, I have hardly seen a professional trader use simple methods such as MA crossover or pinbars etc....

 

[Ferru]

Trend following actually uses simple mathematical models (and doesn't use TA at all) - it uses stuff like MA crossovers or 'buy when the market reaches the highest point of the last x days, and sell when it reaches the lowest point of the last y days'.

The reason you won't have seen a trader use a technique like using that is probably because such menial trades are best done by computers! Why pay a trader £100K a year plus bonuses to do something that requires no judgement, when a computer can do that kind of trading for you 24/7, scanning thousands of markets and guaranteeing flawless execution every time?

Jeff

Thanks Jeff,

will definitely read it.

but by mechanical system do we mean systems that use price action set ups (or/and based on T/A) or complicated mathematical models.

If both then it seems the latter is more popular as there are billions invested in them whereas, as I said before, I have hardly seen a professional trader use simple methods such as MA crossover or pinbars etc....

 

[numbertea]

Hey DDI I had the same concerns. I am still not in a position to comfortably trade my system yet but I'm closer every day. When I finally got some of my entry signals coded it was just amazing to see work because there is no more second guessing. The system is in place and it doesn't faulted of get scared. Now I'm working on exits... Cut losses and let profits ride.

 

[BeginnerJoe]

I made a 100% automatic trading system. It's very easily possible. However, it was not profitable because the idiot who gave me the algo hadn't a clue. If anyone has a working algo, I'd be happy to go in a partnership and share the profits.

 

[i0n]

I have found this thread fascinating and informative, thank you.

I am a programmer with very little (almost none) experience in the financial markets. However I am good at what I do (programming), and I have developed several high concurrency systems using a test driven "agile" methodology to solve problems for clients in other business disciplines. While predictive or reactive algorithms do represent a leap from what I am used to working on, this is currently my primary focus.

At time of writing I have identified three main issues that I need to rectify in order to attain the desired outcome (making money autonomously with calculated or managed risk)

1. I need to understand the financial markets better. How they fit together. Examples of successful manual strategies. Clear and unbiased analysis on the tools (software mainly but perhaps hardware too) that is available. Clear and unbiased analysis of data feeds, access to trade and the costs involved.

2. Infrastructural support. Basically other skilled and driven people with the same objective. Another programmer with similar aims would help drive things along I think. Preferably enthusiastic, articulate and agreeable to pair programming and constructive criticism (both able to give criticism of my code and receive criticism of theirs!) Also, as outlined above, being able to liase with a successful trader/expert in the financial markets would be hugely beneficial and would dramatically shorten the time it will take me to get up to speed in this area. Their insight and "feel" for market opportunities would be invaluable, and I think that this human expert element is a prerequisite to attaining the most successful results.

3. Programming Language/Skill Set issues.
This issue is of the least concern to me. Understanding underlying programming paradigms and syntax generally means that I can pick up new languages effectively. Nevertheless, this may be another barrier that has to be overcome if other software requirements dictate it. For example I have already happened upon Metatrader, and by the looks of things it has it's own proprietary language (MQL?)

As is probably most obvious by now I am committed to being successful in the financial markets, and if after careful analysis I think that the only way to succeed is to trade manually (as part of a larger organisation or as an individual) then that is what I will do. However I currently do not believe this to be the case. I believe that all systems can be explained mathematically. If a system appears to be chaotic or random then it appears that way because of complexity. It is still a system. For proof look at the universe. It appears random and chaotic but it's behaviour can be expressed (at present imperfectly I'll admit) as an equation. It's just my opinion and I'm open to being proven wrong!

If any of what I have said resonates with you and you think that you have some of the skills I have outlined above, please get in contact. I currently live just outside of London (England), and I'd be happy to chat to you over email, IM or even in person if you would prefer.

Thanks again to the people on this thread that have already shared their insights.

PS Sorry about the length of this message. My verbosity is something that I am aware of. What can I say, I love words!