Machine designed strategies. Do they work?

[braincell]

Quote from jimbojim:

Really? Why don't you go back to your job of programming in assembly code before you get fired for spending too much time on forums and let questions about trading systems to more qualified individuals?

Do you think there are people in this forum in a sane state of mind who would spend 60K+ for curve-fitted systems? You know, the type you develop...

I had the same opinion for months after using various search algos, that it's all a pipe dream and that products like TSL are scams. I was really down. It took that much time and patience to learn how to use this type of software. There aren't exactly textbooks filled with statistics on various approaches to selecting input/output values, fitness functions, analysis of market dynamics and suitable patterns/indicators, a range of side-ways and WFO out-of-sample testing, etc. You have to try and learn all by yourself. So i don't blame you for thinking that, a lot of people think so too, even some that have had a lot of experience. In the end, my opinion still holds, that once you learn to use it the curve-fitting disappears and you are able to create good looking diversified portfolios that have a low probability of letting you down. You just sound like you don't have experience doing all of the above with search algos, especially given the low quality of commercial applications that are available out there, so it's pretty expected that you would think so. Maybe, a year from now, i might conclude that my current thoughts were just based on luck and that it really cannot work in the long run, but for now, my attitude doesn't change. I could also repeat my posts from before on how machine learning is just a different form of human learning if done properly, but i'm not sure there's much use of repeating myself. I dont like claiming i'm the smartest guy in the world with my "advanced search algos" and stuff when there are people with far more experience here and elsewhere. The reason i'm not listening to experienced traders when they say "its all a curvefit" is because a very small number of them has had enough experience in this area which is still very very new. The few that i've talked to in detail (with negative opinions) have demonstrated a lack of understanding for the subject and a general dislike for all things new and innovative.

Finally, our new member here can maybe tell us if he has live trading experience with TSL developed systems and what his portfolios look like in general terms. That would help the discussion, i'm sure.

On the other hand, i agree with you that it's silly to spend 60k on such software, especially given the above fact - how much time and effort it takes to learn to use it. Also, TSL makes it hard to test a lot of theories quickly for various reasons, but i won't get into that. An informed buyer should be weary.

Oh and, i can fire myself, but i'm not about to quite yet.

So i invite you to not generalise. Just like all my friends think everyone trading the markets is a billionaire, which is not true, it's also not true to think that every search algo or machine learning in general are bad.

 

[Jtrader500]

I am a TSL user and have designed a couple of high to medium frequency systems that are profitable in a variety of futures markets and stocks. I spent a lot of time trying different settings from bar size, entry tactic, etc. I did not have much luck in medium to longer term systems. I have tried AdapTrade Builder, Price Action Lab (when it was called something else, this is a re-branding of a product, which is suspicious). I have also worked with NeuroShell Trader. None of these came close to what TSL can do, still it is a lot of work and research to find a robust strategy.

The developer of TSL is often updating the software with new features to add value and to speed up processing.

The newest enhancement called EVORUN sounds like it a very valuable tool that does have the potential to find robust systems across a variety of bar sizes and trade types. Prior to this feature you had to try all the entry tactics manually. This will speed up things quite a bit, I haven't had a chance to try this new feature yet, so I cannot say with certainty how valuable it will be.

TSL requires a lot of hours of work to get to something viable to trade in today's markets. This is not for the part-time trader in my opinion.

I never thought I would post on a forum like this, but someone brought it to my attention because they knew I was a client of TSL. It's pretty amazing so many people opining about a product they know nothing about and touting a program that shows paper profits but never used it to trade.

 

[tim888]

Quote from tim888:

Now here is the OOS test:

Quote from ssrrkk:

What is the "CL" column? Thanks.

Consecutive losers (max). PF is profit factor, etc.

 

[tim888]

Quote from Jtrader500:

I have tried AdapTrade Builder, Price Action Lab (when it was called something else, this is a re-branding of a product, which is suspicious). I have also worked with NeuroShell Trader. None of these came close to what TSL can do, still it is a lot of work and research to find a robust strategy.

I use Price Action Lab with great success. There is nothing suspicious about it. This program does exactly what it is advertized for and they - unlike other companies, including TSL - do not hype it up and build false expectations. In addition they provide a demo with some restrictions. Whoever is not providing a demo I believe is afraid of something.

TSL seems like a very expensive version of Adaptrade and I don't see why should anyone pay this kind of money for it. I have never tried it because as I said there is no demo but if I wanted to go ahead and curve-fit models using genetic programming I would go with Adaptrade at a fraction of the price.

In the Adaptrade forum a real TSL user claims that "TSL doesn't work. Tried it personally. Was not able to develop a single working strategy."

http://groups.google.com/group/adaptrade-builder/browse_thread/thread/a161fc8496ee6843?hl=en

I would take your claims in good faith but it smells fishy.

 

[milewski05]

Hey I also own TSL id love to talk to you and maybe share ideas. I have had GREAT success with longer time frames but not as much with the shorter time frames and would love to talk. could you PM your email? or I can PM you mine.
-matt

Quote from Jtrader500:

I am a TSL user and have designed a couple of high to medium frequency systems that are profitable in a variety of futures markets and stocks. I spent a lot of time trying different settings from bar size, entry tactic, etc. I did not have much luck in medium to longer term systems. I have tried AdapTrade Builder, Price Action Lab (when it was called something else, this is a re-branding of a product, which is suspicious). I have also worked with NeuroShell Trader. None of these came close to what TSL can do, still it is a lot of work and research to find a robust strategy.

The developer of TSL is often updating the software with new features to add value and to speed up processing.

The newest enhancement called EVORUN sounds like it a very valuable tool that does have the potential to find robust systems across a variety of bar sizes and trade types. Prior to this feature you had to try all the entry tactics manually. This will speed up things quite a bit, I haven't had a chance to try this new feature yet, so I cannot say with certainty how valuable it will be.

TSL requires a lot of hours of work to get to something viable to trade in today's markets. This is not for the part-time trader in my opinion.

I never thought I would post on a forum like this, but someone brought it to my attention because they knew I was a client of TSL. It's pretty amazing so many people opining about a product they know nothing about and touting a program that shows paper profits but never used it to trade.

 

[milewski05]

if you have questions im here to answer them im not going to argue with you, i know it works cause of the amount of taxes i just paid on trading profits in 2011

Quote from alexandermerwe:

Why asking for people to trust you? This is not an interrogation...Everyone here has his own opinion like everyone here has (at least) one computer...

Risk aversion has nothing to do with this. If a program generates complete systems that work then nobody should sell it because with little money, risk averse trading and compounding you can become rich in 5 years without disclosing the edge.

If the program needs a lot of work done by the user to get a complete system then you introduce curve-fitting, selection and data-snooping bias, subjectivity, emotions, and a host of other problems. In that case I agree it makes sense to sell it but the chances of it working for most people are slim to none. It must be priced accordingly then. The price of the product you mentioned is not justified for the amount of work one must put and the possibilities. There are much cheaper programs that maybe do a better job.

 

[milewski05]

come on your guys, you guys are fighting like women can we please stay on topic and talk about the actual form post and no go back and forth arguing about things that are not related.

Quote from luckyputanski:

I didn't say that. Please keep rereading my post until you understand it. Hint - teaching skills means more people have them.

 

[milewski05]

yes i did and it was a i would do it again and again. if you have questions please ask them. I am not here to argue.

Quote from jimbojim:

Did you really spend 60K+ to purchase TSL or you are an April fools' hoaxer ahead of time?

If you did I really fee sorry about you paying that amount of money for something you can do in excel, i.e. curve-fitting.

 

[milewski05]

Hey braincell thanks for the questions finally someone on here who does not want to argue and give me opinions about something they have never used.


For a larger number of runs on different instruments and trade types, how many OOS charts were profitable in % terms. Was it 10, 20, 60%?

It really depends on the size of the bar. i would say if you are working with high time frame bars 60min and > that the OOS is very very high like 70% once you learn how to handle the data you can get that even HIGHER. i really only trade futures so this only applies to futures if you would like i can give you an example.. just give me a symbol. as far as orders i always use market orders when developing so i know what my slippage is depending on the symbol. if i have to use limit orders then i would only take a system with a very high average trade so i know i can take slippage if i need to.

Quote from braincell:

Thanks for the input, but uh-oh, i have to warn you that since Mike Barna no longer allows people to interview previous clients of TSL, there are many potentially interested people here who would like to hear about your experience. Get ready for a torrent of questions, or quickly dismiss that you used TSL.

For a larger number of runs on different instruments and trade types, how many OOS charts were profitable in % terms. Was it 10, 20, 60%?

It really depends on the size of the bar. i would say if you are working with high time frame bars 60min and > that the OOS is very very high like 70% once you learn how to handle the data you can get that even HIGHER. i really only trade futures so this only applies to futures if you would like i can give you an example.. just give me a symbol. as far as orders i always use market orders when developing so i know what my slippage is depending on the symbol. if i have to use limit orders then i would only take a system with a very high average trade so i know i can take slippage if i need to.

I too would be interested in your experience, such as:
- For a larger number of runs on different instruments and trade types, how many OOS charts were profitable in % terms. Was it 10, 20, 60%?

I ask because i think Mike only shows the good results in his flash videos, and also seems to calculate limit fills when the price touches it, not when it breaks (which is unrealistic).

PS - dont mind people like jimbojim, he isn't sure what he's talking about.

 

[braincell]

Quote from milewski05:

Hey braincell thanks for the questions finally someone on here who does not want to argue and give me opinions about something they have never used.


For a larger number of runs on different instruments and trade types, how many OOS charts were profitable in % terms. Was it 10, 20, 60%?

It really depends on the size of the bar. i would say if you are working with high time frame bars 60min and > that the OOS is very very high like 70% once you learn how to handle the data you can get that even HIGHER. i really only trade futures so this only applies to futures if you would like i can give you an example.. just give me a symbol. as far as orders i always use market orders when developing so i know what my slippage is depending on the symbol. if i have to use limit orders then i would only take a system with a very high average trade so i know i can take slippage if i need to.

Thanks for answering. You say it works on longer time frames, our new member Miketrader500 is claiming high and medium frequency, so you see that's interesting. It's good to have input from you guys, and keep your hats on, the "elite" on this forum can get worked up rather easily.

I sent you a PM with my email. If the forum blocked it because you are a new member, i can ask questions here also. Otherwise, email me and we can continue over there if you are uncomfortable with public forums.

Cheers