Dr. John Clayburg seminar on Creating and Using Automated Strategies

[MultiCharts]

MultiCharts is committed to helping traders learn and grow by inviting industry experts to present webinars. Different presentations cover trading methods, trading ideas, and our MultiCharts trading software. Everyone from new traders to experts can learn something new, so pick ones that interest you and click Register to reserve a spot!

http://www.multicharts.com/multicharts/webinars/

http://multicharts.webex.com/

MultiCharts is pleased to present a seminar with Dr. John Clayburg on Tuesday, May 22 at 4:30 p.m. EST.

Dr. John Clayburg has been programming in the EasyLanguage environment for over 25 years. John’s automated strategies are regularly rated in the top 10 systems by Futures Truth, and published “Four Steps to Trading Success” in 2001. He has been a featured presenter at trade shows and seminars throughout the US and Europe, and is best known for the development of Parallel Function User Technology. His current website is www.onlinesystemtrading.com .

Topic: Creating and Using Automated Strategies

Description:
1. The basic "4 Step Method" used to set the basic rules for any trend following system.

2. The basic steps necessary to create a logical system - all the do's and don'ts.

3. Why use an automated system?

Significant reduction in the stress of trading.
The ability to accurately trade multiple markets and timeframes simultaneously.
The ability to backtest a strategy before you actually trade it with real money.
Confidence that each and every trade is generated in the exact same method.
Programmed exits enforce automated money management routines that are difficult to manage discretionarily.

 

[bashatrader]

Multicharts is an excellent program. It is very user-friendly and EL compatible. I highly recommended it to anyone out there who is looking for a good alternative with portfolio testing capability.

That being said, with all respect, I think the old ways of developing trading systems are now outdated. It is becoming extremely hard to find an edge that is free of optimization/curve-fitting and data snooping bias. IMHO, machine designed trading systems have been the main drive behind the profits of many system traders during the last 5 years and that will be true in the future. Platforms will only serve the purpose of automating the orders.

Maybe next time you can have a presentation along the lines of machine designed strategies, like those you get from Adaptrade, Price Action Lab, Trading System Lab, etc.

 

[jcl]

I wouldn't get too excited about "machine designed strategies". The strategies are human designed. The computer just optimizes them with a machine learning algo that can find patterns. Those patterns fulfil conditions that are either hardwired in the algorithm, or set up by the developer of the strategy. This is not very different to old fashioned TA strategies.

The disadvantage of "machine designed strategies" is that the pattern finding process makes them very sensible to curve overfitting and data snooping. This is not saying that they are useless, but you must absolutely know what you're doing for working with them. Otherwise you're just wasting money with purchasing such software products.

 

[intradaybill]

Quote from jcl:

This is not saying that they are useless, but you must absolutely know what you're doing for working with them. Otherwise you're just wasting money with purchasing such software products.

You are also wasting your time if you purchase a simple charting software and you start drawing lines all over the place and plotting indicators.

You are definitively wasting your time if you start writing hundreds of lines of code for testing systems since the probability of such systems working in real trading is 0.000001 or less.

You are also wasting your time reading books, going to seminars or working with mentors, since the probability is also very low that you will make money.

So what is your point? Do not try unless you already know it all? I’m having hard time understanding what your point is. Every endeavor has a learning curve.

 

[jcl]

Quote from intradaybill:

So what is your point? Do not try unless you already know it all? I’m having hard time understanding what your point is. Every endeavor has a learning curve.

Read the discussion. It was suggested that human strategy development is outdated and "machine designed strategies" are less susceptible to curve fitting and data snooping. Both is wrong. I'm using machine generated algorithms myself and am far from condemning them, but they have their limitations and are no replacement or successor to manual strategy design.

 

[NetTecture]

I take the counter argument that by definition machine designed strategies are MORE curve fitted - there is no human brain doing a last validation

 

[goodgoing]

Quote from NetTecture:

I take the counter argument that by definition machine designed strategies are MORE curve fitted - there is no human brain doing a last validation

Valid point but I think it depends on how it is done. I tried several programs that work that way. I have spent countless hours working with them. For example, Adaptrade starts with random populations of systems that use a mix of indicators, patterns, etc. Then it uses genetic programming which is really a buzzword for a simple evolution scheme where members of populations that best perform in maximizing a fitness function are randomly chosen to come up with new populations and so on. This is curve-fitting but also involves data-snooping and selection bias, amongst many other shortcomings. Eventually this process will fit a system on both the in-sample and put-of-sample if you try many times. On the other extreme three are programs like Price Action Lab that have a fixed population of similar alternatives and select from that the best fit members. This is subject to selection bias only, a less severe but also potentially adverse condition. The difference is that in the latter program if the population selected from the in-sample does not pass the out-of-sample test the first time around the process terminates and there is no additional fitting going on by constructing new populations. What is important to realize is that all these approaches emulate what a human usually does manually but do it significantly faster and more thoroughly. The gain comes from saving time.