Mechanical Systems

[JayC 10]

How long, generally, do you like to back test a trading system, to become confident? It seems the markets are always evolving and changing. Will a true mechanical trading system, work in all markets? Or do they constantly need tweeking?

 

Just wondering some of others' thoughts on this.

 

JC

[Guest Tresor]

Backtest it for at least a 3 year period and then optimize when the equity curve starts to deteriorate.

 

Regards

[Head2k 10]

How long, generally, do you like to back test a trading system, to become confident? It seems the markets are always evolving and changing. Will a true mechanical trading system, work in all markets? Or do they constantly need tweeking?

 

Just wondering some of others' thoughts on this.

 

JC

Hi, when I was first attracted to trading I started with mechanical systems. Programming, backtesting, optimizing etc. Then I started trading live one of the systems I developed. It took me about one week to find out that this was not the way. I couldn't be confident with my system, no matter what results it showed in backtest, because I simply didn't understand the market. I think confidence comes from understanding, knowing why you are doing what you are doing. And then from testing. Some people are probably OK with testing only to be able to gain confidence, but that wasn't my case.

To your second question:

I don't think one system will work in all markets or market phases. Surely you will adapt your systems for different vehicles. Also there are ways of detection the phase of the market (sideways / trending) to some extent. There are also ways of detection of volatility. Hence you can make your system adaptive to some extent (so it is at least not losing much during an unfavorable phase). And to smooth your equity curve you can run more uncorrelated systems at the same time.

[januson 10]

Hi Head2K

 

Is it an automated system you're talking about or you sitting there executing the trades according to some signals printed at a chart? Day or intraday?

 

If your system needs optimizing you should discard it right away! You should however develop a system to a standard stock, i.e. high volume most of the time. Forget about sideways or trending, when you think like that you're allready starting to see problems instead of solutions. If a stock is moving sideways, it just mean no money or at least less money, your system should react to this!!

 

I don't think one system will work in all markets or market phases.

It is a loser coming to talk, allready preparing himself to get ripped of!! (Sorry, but you deserved that )

 

Whats a system then? A system should certainly work under all market conditions, you've to build the the conditions into your system i.e. don't trade between the first hour, don't trade the last hour or if volume is lower than 40% of normal HALT system, 3 loses in a row in a choppy market HALT or change observation period/ view / whatever.

I don't think the markets are evolving, but I think the people are.... they enter, they trade, they get some experience, they get more experience, they go away of one reason or another.

[Head2k 10]

Reading your response, januson, I realize I didn't delve into the system development enough. And I am also glad I shifted to discretionary approach.

[januson 10]

Many people don't realize what a system actually is before they sit down and try to program it. If you can't code it then you'll find yourself strugling with confusion. And that's why so many chooses the dicretionary approach, because a discretionary approach is flexible all the time

[sevensa 10]

Many people don't realize what a system actually is before they sit down and try to program it. If you can't code it then you'll find yourself strugling with confusion. And that's why so many chooses the dicretionary approach, because a discretionary approach is flexible all the time

 

I think you are confusing trading on the seat of your pants with real discretionary trading. That is why many traders fail with a discretionary approach because they think like you do that it is flexible all the time. This is simply not true. You still need a very rigid trading plan and rules even with a discretionary approach to have the best chance of success.

[BlowFish 10]

Hi Head2K

 

Is it an automated system you're talking about or you sitting there executing the trades according to some signals printed at a chart? Day or intraday?

 

If your system needs optimizing you should discard it right away! You should however develop a system to a standard stock, i.e. high volume most of the time. Forget about sideways or trending, when you think like that you're allready starting to see problems instead of solutions. If a stock is moving sideways, it just mean no money or at least less money, your system should react to this!!

 

 

It is a loser coming to talk, allready preparing himself to get ripped of!! (Sorry, but you deserved that )

 

Whats a system then? A system should certainly work under all market conditions, you've to build the the conditions into your system i.e. don't trade between the first hour, don't trade the last hour or if volume is lower than 40% of normal HALT system, 3 loses in a row in a choppy market HALT or change observation period/ view / whatever.

I don't think the markets are evolving, but I think the people are.... they enter, they trade, they get some experience, they get more experience, they go away of one reason or another.

 

The thread is about mechanical systems whether automated or not is kind of academic it could be automatic, broker executed, or self executed.

 

I wonder if you are actually talking about fully mechanical systems? as most of what you are saying is contrary to my experience. It's also contrary to what most systems designers, who also publish, write. Do you trade a mechanical system(s)? How long have they been working?

 

I am interested cause it sounds like you have discovered the holy grail if you have a system that can use the same entries for trend and consolidation that works in all market conditions under all levels of volatility and never needs tweaking or putting on the shelf for a while

[Head2k 10]

As I understood it he is talking about something I would call a compiled system, which is basically a system that includes different signals for different market conditions. Could be understood as a bunch of standalone systems running together. I guess it depends on what one means by "system" - if it is one setup bound with certain criteria, or a whole trading strategy.

[BlowFish 10]

As I understood it he is talking about something I would call a compiled system, which is basically a system that includes different signals for different market conditions. Could be understood as a bunch of standalone systems running together. I guess it depends on what one means by "system" - if it is one setup bound with certain criteria, or a whole trading strategy.

 

Yes that occurred to me H2K. Interested to hear more

[darthtrader3.0beta 10]

A real eye opening experience is to program a system and optimize on things you know are completely bogus and random.

A system that only buys at 10:16 on a tuesday..if you use even the slightest bit of money management to have bigger winners than losers its not that hard to find a "profitable system" like this with an "edge".

The problem is the system has no edge, its just data snooping bias and purely random.

Even win rate has a ton of randomness in it.

Here is a system: long SPY at 10:00am everyday for the last 2 months, stop at 1/2 ATR and target at a full ATR..

89% win rate but it loses .13% before commisions..

To me the point is, with historical backtesting you simply can not know if you are exploiting a fundamental behaviour of the market or if you have simply found some good looking random patterns in your data...Even then with walk forward testing, there is nothing to say your are simply not getting lucky, especially since if you walk forward for 2 months and get great results you will probly have alot of false confidence in the system.

I do agree that I don't get why anyone uses optimized parameters in a system..

Currently, I simply see no way around using a monte carlo simulation. At least then you have a control to test against as far as purely random behaviour is concerned.

Not to mention that to me the retail algo literature focuses far too much on the entry with trade management and the exit an after thought.

[atto 20]

Way of the Turtle addresses many of these issues (especially backtesting biases), if anyone's interested in creating a mechanical system.

[bakrob99 11]

I do agree that I don't get why anyone uses optimized parameters in a system..

 

Everything you said can equally apply to a discretionary system. So what? If you are live trading your system and making money, just be aware that at some time in the future it may stop working. How will you know? Look at your equity curve and treat it like a price curve. Stop trading it when it is trending down and wait for it to turn back up gain, or set a new high or higher high.

 

One of the big advantages to computer generated trading is the automatic record keep and statistics maintained (in my case by TradeStation). And I know that having just one (type of) system forme is not helpful.

 

In my case, I have a breakout system, a pullback system and a consolidation system. They all work - some times better than others but on average work more consistently than just 1 by itself.

 

Be sure to reduce size. Trade small. Build an edge and consistency on a daily basis.

[darthtrader3.0beta 10]

Everything you said can equally apply to a discretionary system. So what?

 

Well I'm not sure what you mean. I mean if you look at discretionary trading from an algo standpoint, the discretionary trader is basically forward testing and optimizing the most powerfull neural network software ever devised..

That is vastly different than if your optimizer tells you to use a 21.22 and 51.054 moving average because its slightly more profitable than a 20 and 50 moving average.

From trying to get into auto trading, I feel that at the retail level we are still stuck in the 90s while there have been huge advances in auto trading in the last 10-15 years. I totally agree with your multi strategy statement, our software should open up to a page that shows the portfolio of strategies and how they interact against real market data and purely random data. Instead we get alot of point and click strategy wizards and useless indicator optimizations.

This sounds like an interesting new book on the subject but I had been waiting for someone to post a review which someone did yesterday. Probly worth checking out as far as stuff "filtering down" to our level.

http://www.amazon.com/Quantitative-Trading-Build-Algorithmic-Business/dp/0470284889/ref=sr_1_1?ie=UTF8&s=books&qid=1220109693&sr=1-1

[januson 10]

Well I'm not sure what you mean. I mean if you look at discretionary trading from an algo standpoint, the discretionary trader is basically forward testing and optimizing the most powerfull neural network software ever devised..

Yes I can agree what that, but only to a certain degree. The human brain is also very complicated and even small insignificant things can affect the outcome in a negative way! Consistency is needed

 

That is vastly different than if your optimizer tells you to use a 21.22 and 51.054 moving average because its slightly more profitable than a 20 and 50 moving average.

No one should ever built a system that need that kind of optimizing, I think the main reason for this confusement is that the majority of new players at the market are overwhelming these times.

 

From trying to get into auto trading, I feel that at the retail level we are still stuck in the 90s while there have been huge advances in auto trading in the last 10-15 years. I totally agree with your multi strategy statement, our software should open up to a page that shows the portfolio of strategies and how they interact against real market data and purely random data. Instead we get alot of point and click strategy wizards and useless indicator optimizations.

The Market isn't random so why should one ever concentrate on building something at a unrealistic background, besides it would be fun :o

But I agree with you

[Dacamic 10]

To me the point is, with historical backtesting you simply can not know if you are exploiting a fundamental behaviour of the market or if you have simply found some good looking random patterns in your data...Even then with walk forward testing, there is nothing to say your are simply not getting lucky, especially since if you walk forward for 2 months and get great results you will probly have alot of false confidence in the system.

I do agree that I don't get why anyone uses optimized parameters in a system..

Currently, I simply see no way around using a monte carlo simulation. At least then you have a control to test against as far as purely random behaviour is concerned.

Not to mention that to me the retail algo literature focuses far too much on the entry with trade management and the exit an after thought.

When done properly, back tests can be very useful for developing trading systems. As you said, though, consideration must be given to doing more than simply running strategies against a batch of data. You mentioned three such additional factors maybe worth pondering: Monte Carlo simulations; parameter optimization; and, money/risk management.

 

In regard to the original questions, my individual strategies are initially evaluated against at least 250 bars of data (I am an EOD trader).

[bakrob99 11]

One of the biggest advantages I have received from spending the year of development effort devising a winning stragey or 2, is the understanding of what it takes to be a winner in the market. You can get this by discovering what money management rules work , and which ones do not.

 

Then - should you decide to take some discretionary trades - you are in a much better position to win over time by following the rules which have proven merit from your backtesting.

 

Case in point: I used to think it was profitable to scale out of trades with as little as 10 ticks on the YM (4 on the ES) and try to hold onto a 2nd position. I now realize that trying to scalp the market is ultimately a losing proposition.

 

Cut down you trade frequency - get into a trend and trail your stop a long way behind the market. You need to take a profit at least 2 x your stop size to be a long term winner.

[sevensa 10]

Cut down you trade frequency - get into a trend and trail your stop a long way behind the market. You need to take a profit at least 2 x your stop size to be a long term winner.

 

I don't agree with this statement. Just looking at risk:reward, is looking at only half of the equation. The higher your accuracy, the less your risk:reward needs to be to be profitable and visa versa. You need to look at expectancy and not just risk:reward.

[bakrob99 11]

My point is: for my entry techniques I learned by backtesting that taking profit too soon was an account killer. If you are successful scalping the market - congratulations.

[bakrob99 11]

What I mean is that while optimizing money management parameters for my automated (mechanical) systems, I learned that, especially if you are scaling out of a position, it is the last part of the position which provides the major portion of the account eequity growth.

 

Staying in the move is essential to build up the value of the Average Winning Trade.

 

This was not what I expected when I first seriously tried to develop automated strategies. And it was only after I gave up trying to take 10 points of profit for the 1st contract in the YM (as an example) and held for longer gains (34 ticks on the 1st half) - and then left my stop at breakeven while the 2nd position runs for as long as possible that the strategies really began to have a tradeable edge.

[Dacamic 10]

... You need to look at expectancy and not just risk:reward.

I agree and tend to focus more on expectancy.

[bacchus 10]

Hi every one

 

Can you imagine that somebody has developed a mechanical system that works

and he is prepare to sell it for a few $$$$?Come on guys,we know these kind of systems are for people who don't have a bloody idea about trading and they are prepare to part of their money.The best system is learn to SEE the chart

with no systems and no indicators.Then you will make some $$.Why am I say all these?Because I never heard anybody to be successful using systems of that nature.

[Guest mifgjkar]

Is it an automated system you're talking about or you sitting there executing the trades according to some signals printed at a chart? Day or intraday?

[Dacamic 10]

The best system is learn to SEE the chart

with no systems and no indicators.Then you will make some $$.Why am I say all these?Because I never heard anybody to be successful using systems of that nature.

It's been awhile since I read Marty Schwartz's book (Pit Bull), and so may not correctly remember certain details. If my somewhat faded memories are reasonably accurate, Marty earned 25% compounded monthly returns using -- among other tools -- moving averages, stochastics and Terry Laundry's Magic T indicator for his trades.

 

Discussions of this nature sometimes get hung up on definitional disputes about "system" and "indicator". As a guess, it's the term "mechanical" that is causing some people to have mental heartburn. To offer one perspective in that regard, my trading did not become consistently profitable until I began developing and trading mechanical systems, possibly because it afforded me a more structured environment.

[edabreu 10]

Currently I am using a mechanical system. What is does is generate a buy or sell signal, and I chose when or if to enter into a trade. Since I have not taken that last step (left it to place the trades for me) then I think it falls under the heading of mechanical. Would a rule based indicator trade signal that the trader would enter into at their discretion be the same thing?

 

If it were totally automated, then it would fire the signal and place the trade for me.

 

The auto scalper system (that's what we call it) is currently about 85%. Its a scalper, so the targets are close - 1, 1.25, 1.75 points on the S&P. I don't trade 3 contracts with it, so for me its a true scalper - 1 point all in all out. The stop is 2.5 points, but in the back testing stats, 85% average negated the occasional full stop out. What's unique is that if it starts to move against you, you can get out before a full stop out (which I often do) and so keep it the stops to about 2 points.

 

It's not my system, but so far it has performed extremely well in real market trading.

 

There are a few real market tested systems out there that have performed well consistently, and I have read about them in Futures Truth magazine. The BWT Zones is the main one that comes to mind. It is still in use today.

 

I am discovering (in my research) that the best automated systems are NOT for sale. They are leased, and run either through strategy runner or some other automatic trade entry software at the broker's host site, and the cost is x amount per contract traded.

Edited January 18, 2009 by edabreu
Add more thoughts