Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other strategy-fiding algos ?

[Lx Jedy]

Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other  strategy-fiding algos ?

[Hugh Hammer]

What other strategy finding algos are you referring too....trying to get an accurate picture

Hugh

On Tue, Apr 18, 2017 at 12:37 AM, Lx Jedy <lx....@gmail.com> wrote:

Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other  strategy-fiding algos ?

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.
To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-builder+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Brian Heyliger]

Here's two I know of: 

http://www.priceactionlab.com/

http://www.tradingsystemlab.com/

Limited experience with either...

On Tue, Apr 18, 2017 at 3:39 PM Hugh Hammer <hugh...@gmail.com> wrote:

What other strategy finding algos are you referring too....trying to get an accurate picture

Hugh

On Tue, Apr 18, 2017 at 12:37 AM, Lx Jedy <lx....@gmail.com> wrote:

Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other  strategy-fiding algos ?

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-buil...@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-buil...@googlegroups.com.

[Hugh Hammer]

The TSL is always been way out there.... price wise

First time I have seen pricelab ....looking thru manual for me, was confusing

Hugh

On Tue, Apr 18, 2017 at 3:45 PM, Brian Heyliger <heyl...@gmail.com> wrote:

Here's two I know of: 

http://www.priceactionlab.com/

http://www.tradingsystemlab.com/

Limited experience with either...

On Tue, Apr 18, 2017 at 3:39 PM Hugh Hammer <hugh...@gmail.com> wrote:

What other strategy finding algos are you referring too....trying to get an accurate picture

Hugh

On Tue, Apr 18, 2017 at 12:37 AM, Lx Jedy <lx....@gmail.com> wrote:

Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other  strategy-fiding algos ?

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-builder+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-builder+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-builder+unsubscribe@googlegroups.com.

[Jorgen Morgen]

Other Trading System "finding" applications worth mentioning are:

1) Neuroshell Daytrader (neuroshell.com)

2) Chaoshunter (chaoshunter.com) same people who make Neuroshell

3) StrategyQuant (www.strategyquant.com) Adaptrade Builder "Clone" Many concepts and program features were copied from Adaptrade Builder. Its interface is made with Java, which makes it run rather slow, although search algorithms are optimized and compiled in faster languages.

4) Discipulus 5 (rmltech.com) Kind of like genetic programming for finding formulas to different input values, can be used to curve fit data. The engine behind Discipulus 5 is what powers the "price wise impossible to obtain" TradingSystemLab. Discipulus 5 evaluation is very very buggy, so the licensed version must be buggy as well.

5) GeneXProTools (www.gepsoft.com) Is another formula finding application similar to Discipulus 5 and Chaoshunter. More inclined to statistics, but can be implemented for markets trading if you have the patience to get your hands messy.

I chose to purchase Chaoshunter because it was the easiest to use in comparison to Discipulus or GeneXProtools. Chaoshunter does not include a free trial like Adaptrade Builder, but they offer a 30 day satisfaction money back guarantee. Also, on their website they show a lot of examples of how to use Chaoshunter for trading different markets. It produces formulas which are similar to excel and they offer a free separate application you can use to convert Chaoshunter formulas to Tradestation/Multicharts/Metatrader.

My plan is to complement Chaoshunter with Adaptrade Builder, since Adaptrade Builder's engine is more focused on finding trading system rules, whereas Chaoshunter is more specialized in building formulas (indicators). Once you can pump out good indicators from Chaoshunter, you can plug them into Adaptrade Builder as indicators and it can make life easier for Builder.

After a long analysis of how TradingSystemLab, Chaoshunter and Adaptrade Builder worked (yes, I watched all the videos), I came to the conclusion that if you have Adaptrade Builder and Chaoshunter together, you can very much approximate what TradingSystemLab does. Remember, TradingSystemLab is an olympic platform at olympic prices only reserved for well funded "Deep Pocket Gods" willing to withstand Mike Barna's ego.

For a little bit of history, Adaptrade Builder was inspired by TradingSystemLab. You can read Mike's newsletters and articles and you will see that in one of his articles, he wrote that he was surprised at TradingSystemLab's unreachable price, and decided to start to code a very basic "Automated Trading System Generator" using at that time's Tradestation optimizer (There was no genetic optimizer in Tradestation at that time).


If you have watched TradingSystemLab's videos where Mike Barna explains how it works, you notice that he never uses commissions and slippages in his optimizations. This is a bummer. It is not a fair comparison if you use such a software and exclude these frictional costs. I bet I can use Builder and find equity curves as beautiful as what TradingSystemLab can get if I zero out all frictional costs, but I know this is unrealistic.

So to summarize, I have been watching Adaptrade Builder evolution since it started and trialed it back when it was version 1.7x or something. I even found out that in previous versions, it crashed when you inserted boolean indicators, but it didn't stop me from keeping a watchful eye from time to time on Builder's evolution. Mike Bryant is my hero.

I think that for the price, Adaptrade Builder is a gem in its own, you just have to play with it a lot and be willing to leave your workstation on 24/7 processing different optimizations and fiddling with it to find what fitness metric best suits your trading profile.

I can explain the difference in genetic programming methodology (Karva notation, Register Machine Language Neural Networks, etc), employed in the above mentioned applications, but I will leave that discussion for another post.

Jorgen

[Brian Heyliger]

Great post! Thanks Jorgen! 

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-buil...@googlegroups.com.

[Gregorian]

I am a user of Trading System Lab and StrategyQuant as well as Adaptrade Builder. Adaptrade is the easiest to use, with the best user interface. SQ is slow to generate strategies, very slow at being updated with new features, and does not always generate strategies whose results are replicatable in Tradestation/Multicharts and Ninjatrader.

TSL has one major advantage over the other two: It supports a broader variety of order types, including and especially the critical "on close" [e.g. buy this bar on close], which is the only order type that can be reliably used with many Advanced Bar Types. Adaptrade and SQ only support "next bar on open" orders, so they are of limited or no use with, for example, developing Renko bar strategies. TSL is also better at continuing to evolve increasingly improved strategies over multi-hour long runs. Adaptrade and SQ are better, however, at providing granular control over which indicators and order types are considered.

TSL does not support Ninjatrader or MT4, but then again, SQ and Adaptrade only support Ninjatrader 7, which is becoming less relevant as time goes on. MT4 will eventually be phased out in favor of MT5, so that, too, will become less relevant. EasyLanguage seems to endure.

You do get what you pay for. Mike Barna of TSL is very helpful and attentive, as he ought to be for the price he charges.

The holy grail would be the superior user interface of Adaptrade with the order type support and easier ability to continue evolving ever-better strategies of TSL, hopefully with Ninjatrader 8 support. I do think Mike Bryant is headed in that direction, and no one can question his commitment to his product.

[Brian Heyliger]

Hi Jorgen,

I'm also using Chaoshunter... interested in having a conversation offline...

Please contact me - heyligerb [at] gmail dot com...

Thanks,

Brian

On Wed, Apr 19, 2017 at 5:52 PM Jorgen Morgen <hsmg...@gmail.com> wrote:

--
You received this message because you are subscribed to the Google Groups "Adaptrade Builder" group.

To unsubscribe from this group and stop receiving emails from it, send an email to adaptrade-buil...@googlegroups.com.

[peter zwag]

There is a new but less mature product call GSB.
A comparison to builder is here.
https://www.youtube.com/watch?v=np87RA9H7c0

GSB is raw processing is about 10x faster than GSB (and Builder is fast)
However GSB architecture produces systems that have much higher chance of working out of sample compared to Builder (and probably the others too).



There is a free trial if you fill out the web form below.
http://www.trademaid.info/sysbuilder.htm

Builder can be made to work, and its nice to drive, with good support.
It tends to build systems with lots of redundant logic, and few go ok out of sample. It has very poor support for other data streams (unlike GSB)
Statquant has large following, but I couldn't get it to work, hasn't had a new version for a year and the documentation is woeful.

[Michael R. Bryant]

I would just caution against drawing comparisons between two products that seem to be quite different. To assert that the other program produces strategies that work better out-of-sample when the build comparison may not be similar seems a leap. Builder has a wide array of features that are designed to improve out-of-sample performance, none of which were used in the comparison in question.

Mike Bryant

[kj.w...@gmail.com]

What about a contest?

Mike and Peter will use their software to develop, test and validate their best strategy and share the code on this forum.
And after 12 months the results will be known.


Op dinsdag 18 april 2017 10:17:37 UTC+2 schreef Lx Jedy:

[Michael R. Bryant]

That’s what free trials are for. Trading contests rely too much on luck. I’ve been involved in futures trading for more than two decades and have seen my share of trading contests and participated in one years ago. The winners tend to be the ones who take big risks and benefit from luck. My customers and potential customers are very smart. I trust them to make an informed decision by evaluating the software based on the criteria that are most important to them.

 

Mike Bryant

[Jorgen Morgen]

Lets wait until Peter Swag comes out with a Free Trial of his Genetic Strategy Builder Software. Right now it is in alpha version and not so easy to download.
We have to take Genetic Strategy Builder with an open mind the same way we compare Trading System Lab with an open mind. By open mind I mean, to learn what other softwares can do and are good at or have needed features that Builder doens't have. Then, implement those features in further releases of Builder. Doing this will require that developers of each software to let pride aside and start adding and perfecting their softwares.

For example, GSB states that it doesn't repeat the same found strategies like Builder does. This would be a good feature to implement in future versions of Builder.
I think Mike should have already filled the web form of GSB in order to download the Alpha Trial it says it has in their website in order to study and dissect what good features GSB has and implement them in future versions of Builder.
This way, together, they can each develop better and better versions in the future for the benefit of their customers.

Swag's website has another interesting app in his website called Enhanced Walk Forward Optimizer which he states is superior to Mulchart's WFO. I think there is an idea here, where we could have WFO functionalities merged into Builder in order to add more robustness tests to the already varied repetoire Builder has, like allowing a Builder user to automatically perform different WFO tests without leaving Builder after a prospective strategy has been found.

Also, an interface API to builder would be interesting. In the past I programmed an interface automator for Builder upon a friend's request to do it for him (he is a licensed user) and he was able to read the backtest reports Builder generated and even click Builder's button "programatically" based on the automator's pre programmed decisions. It was like an external app reading Builder's reports, forms and windows and then pressing the buttons of Builder's interface based on the results of what it read previously. Using this method I was able to make sort of a Builder run restart if the results didnt match expectations after every X generations. All this was way before Mike implemented the Restart feature in Builder if objectives not met after X generations. I think my friend got this same concept of restarting the build from Trading Systems Lab. Mike Barna from TSL called this feature: "Multirun".
I programmed the automator with Auto-it.

Jorgen Morgen


----

[Brian Heyliger]

Very interesting discussion!

I second the API feature request. I'm currently in need of this feature for exporting trade-by-trade stats for automating the construction of multiple systems to into a portfolio with an equity curve that has a high Correlation Coefficient...

--

[Michael R. Bryant]

You’ve probably already considered this, but you can, of course, save the trade list of Builder to a MSA file, then use the portfolio feature in MSA to construct your portfolio.

 

Mike Bryant

 

From: adaptrad...@googlegroups.com [mailto:adaptrad...@googlegroups.com] On Behalf Of Brian Heyliger
Sent: Sunday, May 21, 2017 11:52 AM
Subject: Re: Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other strategy-fiding algos ?

 

Very interesting discussion!

[peter zwag]

"What about a contest?

Mike and Peter will use their software to develop, test and validate their best strategy and share the code on this forum.
And after 12 months the results will be known."

That’s what free trials are for. Trading contests rely too much on luck. I’ve been involved in futures trading for more than two decades and have seen my share of trading contests and participated in one years ago. The winners tend to be the ones who take big risks and benefit from luck. My customers and potential customers are very smart. I trust them to make an informed decision by evaluating the software based on the criteria that are most important to them.

{Mike}

I'm sympathetic to Mike's perspective of a free Trial. Most systems builders do this except TSL and Ward system Group.

There is an expression that beauty is in the eye of the beholder. Same thing with system builders.

5 years ago approx 100% of all systems I built in Wards chaos hunter failed in the OOS period.

Maybe its improved since, but I couldnt get it to work.

Stratquant seemed to have a really big following, and after spending over a week full time on it, I could not get it going.

Documentation very bad, critical bugs and promise of a new build 'soon'...  (nothing over a year)

Bottom line is I think the program works, but I cant get it work. I suspect its strong in FX market, which I am not. (yet) and I suspect Builder is not (yet) strong  in FX.

I had some success with Builder, and it well documented, good support, not buggy and well laid out.

I was also happy to purchase Builder, and I couldn't say this about any of the other competitors.

The point I make is that nothing is perfect, and some builders work better than others, and some people can use them, and others cant.

Hopefully they all get better as time goes on.

So I very much agree with "I trust them to make an informed decision by evaluating the software based on the criteria that are most important to them."

I don't see a need for a competition. It the users who decided what works best for them.

RC2ES was build by my earlier revision of GSB.

I can say however that RC2ES has been in futures truth top 10 for the last 6 months. That's impressive in that post Trump, daytrading ES has been really hard.

RC2ES was also the top S&P emini futures day trading system in futures truth  http://futurestruth.com/    by a long way. It also had over a year live brokerage statements.

I'm not so sympathetic to luck in trading. Robins world cup trading competition is often won by the same people for several times or won by people who have done the winners courses.

However their results can be so high because they use leverage above what is normally safe for long term trading. They can also have multiple accounts in the competition.

So they could have had 9 accounts blow up, and one make a lot of money.

You also need to factor in that markets go through good times and bad. During global financial crisis, my biggest shock was the tax bill. This last year on the S&P 500 was

the hardest since 2005.

One of the reasons I needed GSB was to develop systems on the markets that Im not yet good at trading.

Whats good about forums is people share ideas of what works, developers can take notes and improve the products. Everyone wins.

Peter

--
You received this message because you are subscribed to a topic in the Google Groups "Adaptrade Builder" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/adaptrade-builder/aMjTfGe7N-M/unsubscribe.
To unsubscribe from this group and all its topics, send an email to adaptrade-builder+unsubscribe@googlegroups.com.

For more options, visit https://groups.google.com/d/optout.

--

Thanks
Peter Zwag
pe...@zwag.id.au

[Brian Heyliger]

Yes... I did. It's been sometime however since I tried that. If I recall correctly, there wasn't a way to optimize the portfolio equity curve by Correlation however, so I had to use a Python script to help with that...

--

[Michael R. Bryant]

When you refer to “correlation coefficient” in this context, how are you using it for your portfolio? Do you mean you want the equity curve to have a high correlation coefficient with a straight line, like the correlation coefficient metric in Builder, or are you referring to the correlations between market systems?

 

Mike Bryant

 

From: adaptrad...@googlegroups.com [mailto:adaptrad...@googlegroups.com] On Behalf Of Brian Heyliger
Sent: Monday, May 22, 2017 8:26 AM
Subject: Re: Are there any data on how fast the Adaptrade Builders can find strategies and how it compared with other strategy-fiding algos ?

 

Yes... I did. It's been sometime however since I tried that. If I recall correctly, there wasn't a way to optimize the portfolio equity curve by Correlation however, so I had to use a Python script to help with that...

[mandelmus]

That's true, Mike has given us the ability to export strategies to MSA, but that is a tedious process involving lots of trial-and-error ... imagine trying to figure out which ten strategies out of millions of potential strategies to combine to come up with the 'best' portfolio.  

I would really like to see Builder do this for us during the build process.  Just like Builder creates and combines strategies to build better strategies, I would like to see Builder retain and combine the top N strategies into a portfolio that performs better than any individual strategy alone.  A diverse portfolio of top strategies has the potential to better handle a situation in which some of the strategies don't work so well out of sample.  It also has the potential to help us capture and spend more time in the market.

Similarly, even if we have Builder build across a group of symbols (FB, AMZN, NFLX, GOOGL), Builder applies each strategy across all symbols.  It does not figure out the best strategy for each symbol in our portfolio.  Thus, the performance of each strategy results from a single strategy applied across all symbols in the portfolio.  Obviously, each symbol will not work equally well with each strategy.  Thus, it would be nice if Builder would match the best strategy to each symbol and combine those symbols and their strategies into an 'optimum' portfolio.