Machine designed strategies. Do they work?

gmst · Apr 18, 2012

Amazing industry, Let me ask you an unrelated question. For how many years have you been trading, and how many *different* strategies have you devised based on observed market dynamics? And is most of your work on US Equities?

Most of the strategies I have developed have had their origins on persistent market behaviors and had a logical underpinning to them. Its clear that you believe in this approach. But after a certain amount of time and work, the ideas start to get exhausted and one starts going round and round as far as developing new strategies is concerned. Just want to benchmark against you to see after how many years, how many different strategies have you developed ? Thanks.

braincell · Apr 18, 2012

Quote from amazingIndustry:

but dont you see that this is in summary the whole fallacy of curve fitted models...

That's your experience, and you probably have more than I do, I admit. You know when you were a kid and you knew best, and parents try to tell you not to jump the bike down 50 stairs, but you just gotta learn on your own and mess up anyway. Then again, you know about the other kids who got lucky for not listening. Same analogy, there's that attitude that i'll make it regardless of what others say right now. It's also enforced by the stories of the minority that did make it. It's ultimately a risk i'm aware of and willing to take. Hey, what's life without taking some risks. Ambition is a bitch. The only true lessons are learned on your own mistakes. So if that blows up my account, oh well, i'll just sell trading software.

I also agree about the models derived from trading experience. I can also say that some of it can be implemented in machine learning by limiting concepts and targeting certain modes of operation. There are numerous methods for machine learning that can be combined to create new exoctic ones, but i digress. Still, the curvefit will be there. Getting that kind of trading experience takes years, and it's time i dont have right now. In my early days i basically made "machine learning" and "my learning" equivalent - not really the case right now. The relationships in data i'm finding seem good, and in that regard i'm of the opinion that most statisticians would tell me to "go for it". Your opinion is different, but in the end all i can do is wait and see, right.

amazingIndustry · Apr 18, 2012

agree on most parts, and I laud you for your attitude. You seem to be willing to put the work in. Hope it pays off.

Quote from braincell:

That's your experience, and you probably have more than I do, I admit. You know when you were a kid and you knew best, and parents try to tell you not to jump the bike down 50 stairs, but you just gotta learn on your own and mess up anyway. Then again, you know about the other kids who got lucky for not listening. Same analogy, there's that attitude that i'll make it regardless of what others say right now. It's also enforced by the stories of the minority that did make it. It's ultimately a risk i'm aware of and willing to take. Hey, what's life without taking some risks. Ambition is a bitch. The only true lessons are learned on your own mistakes. So if that blows up my account, oh well, i'll just sell trading software.

I also agree about the models derived from trading experience. I can also say that some of it can be implemented in machine learning by limiting concepts and targeting certain modes of operation. There are numerous methods for machine learning that can be combined to create new exoctic ones, but i digress. Still, the curvefit will be there. Getting that kind of trading experience takes years, and it's time i dont have right now. In my early days i basically made "machine learning" and "my learning" equivalent - not really the case right now. The relationships in data i'm finding seem good, and in that regard i'm of the opinion that most statisticians would tell me to "go for it". Your opinion is different, but in the end all i can do is wait and see, right.

alexvnew · Apr 19, 2012

Quote from amazingIndustry:

Heck, I say this with an advanced degree in Stochastic Calc and Financial Engineering. But I deeply believe in what I just said.

You use very vague and basic terminology, not the kind that someone with an advanced degree in these fields would use. Maybe you studied them but 99.99% of those who study these fields do not understand them because Universities nowadays are mass production factories rather than educational environments. Your thinking is too complicated for trading. You are lost into rivers of vague notions but trading and especially trading systems is much easier than what you may think. You probably worked as a quant isolated in a cubicle and the only sounds of real life you were hearing during your 14 hours of daily work were the farts of the guys working in the next cubicles and you never had real interaction with experienced traders other than you getting to know what they wanted you to know. You probably developed systems in a group that afterwards the traders told you they didnât work in an effort to discourage you from selling them. I suggest you start from scratch trading simple moving average systems with mini contracts and forget about what you were forced to learn to this date. It seems to me that your thinking is chaotic and your head is full of garbage you do not need. You must clear your mind first of your misconceptions and understand that trading is much easier than you think. Good luck to you - try not to be a quant but a trader.

amazingIndustry · Apr 19, 2012

Interesting that you had NOTHING whatsoever of value to add to this discussion so far. And then you come to personally attack others? Why don't we let others judge?

P.S.: Oh by the way, there is a reason hft firms, hedge funds, and sell side firms hire only the brightest and most promising (unfortunately they often get it wrong though) candidates that they can get their hands on (before the googles and amazons and facebooks snap them up). Its because trading is not easy. And not all trading is about flipping couple stocks around. I started out at an exotics desk and I am not sure anyone would take you seriously if you had no clue about measure theory, stochastic calculus, PDEs, and a deep understanding of probabilities, among others. But it was and still is a highly profitable business. You only prove you have not the slightest clue what you are even talking about. Are you one of those who somehow did not make the cut and now you need to rebuke others who you think are inferior to you but got the chance to step up to the plate? Then I pity you because there are many ways to happiness in this world, its not all about money, and careers.

I have two things to give to you on the way:

âOrdinary fools are all right; you can talk to them, and try to help them out. But pompous fools-guys who are fools and are covering it all over and impressing people as to how wonderful they are with all this hocus pocus-THAT, I CANNOT STAND! An ordinary fool isn't a faker; an honest fool is all right. But a dishonest fool is terrible!â
Richard Feynman

and

If you can't explain it simply, you don't understand it well enough.
Albert Einstein

Quote from alexvnew:

You use very vague and basic terminology, not the kind that someone with an advanced degree in these fields would use. Maybe you studied them but 99.99% of those who study these fields do not understand them because Universities nowadays are mass production factories rather than educational environments. Your thinking is too complicated for trading. You are lost into rivers of vague notions but trading and especially trading systems is much easier than what you may think. You probably worked as a quant isolated in a cubicle and the only sounds of real life you were hearing during your 14 hours of daily work were the farts of the guys working in the next cubicles and you never had real interaction with experienced traders other than you getting to know what they wanted you to know. You probably developed systems in a group that afterwards the traders told you they didnât work in an effort to discourage you from selling them. I suggest you start from scratch trading simple moving average systems with mini contracts and forget about what you were forced to learn to this date. It seems to me that your thinking is chaotic and your head is full of garbage you do not need. You must clear your mind first of your misconceptions and understand that trading is much easier than you think. Good luck to you - try not to be a quant but a trader.

braincell · Apr 19, 2012

This forum and the financial industry as a whole is like sitting at a poker table. You can never have a true idea about who the other person is, and what the reasons for his actions are - in this case, why he is saying what he's saying, and why i'm saying what i'm saying. Everyone seems to have an agenda. It wouldn't surprise me if most people here said the opposite of what they really believed. That's my experience so far (based on a few years of interaction) anyway and it's not easy to adjust. It's not an academic field - trading and finance. It's a business. I'm sure that by now if all the work and great knowledge acquired by the quants of the world was published, there would be a few scientific breakthroughs in all sorts of sciences - from random theory to mass psychology, etc. But it isn't. So whatever is truly worthwhile, i doubt anyone will talk about. Why take anything you read here so seriously then? It's all poker, and a bit of a laugh...

I mean, why don't we all post our CVs on ET and have the mods check our home address and phones, etc. Maybe then the discussions would change a bit. :/

gmst · Apr 19, 2012

Quote from amazingIndustry:

I started out at an exotics desk and I am not sure anyone would take you seriously if you had no clue about measure theory, stochastic calculus, PDEs, and a deep understanding of probabilities, among others. But it was and still is a highly profitable business. You only prove you have not the slightest clue what you are even talking about. Are you one of those who somehow did not make the cut and now you need to rebuke others who you think are inferior to you but got the chance to step up to the plate?

Mate, you have a good degree and you worked on an exotics desk - alright understood. What OP is saying is that you don't need much complicated maths to make money trading delta1 in a systematic/quant way. Invoking your understanding of stoccal and PDEs doesn't mean that OP is wrong. Infact OP is correct.

Anyways, you have been harping on here about you being a quant trader and ibank experience and masters in finance and StocCal and PDEs and Measure theory and what not. Care to mention your sharpe and PF for last 2 yrs? I bet a few traders on this board will beat you on those measures. Its good to keep your ego in check.

Anyways, do you realize you have been running away from engaging in a point by point technical discussion with braincell. You are just commenting on general points. I know your intention is good but you lose credibility when you run away from serious discussion. I was like you when I was in grad school. Know it all guru types. I am no more that way - markets humbled me

amazingIndustry · Apr 19, 2012

* braincell is not the OP (and neither the OP nor braincell claimed what you point out)
* "x does not mean OP is wrong. So OP must be right". I am afraid I do not follow your line of reasoning.
* I am sure some people will have higher Sharpe ratios than my strategies have. I never claimed otherwise. I am not sure how this is related to my ego. I respect those that have high PERSISTENT risk adjusted return measures to show for their work.
* I traded many years after grad school. And your comment "I was like you when I came out of grad school" makes the assumption you are far ahead of me. I think you go out of a limb here with your assumptions, especially knowing nothing about my track records or more detailed background.
* Please point out in detail(!) where I evaded in-detail discussions. I was never asked to comment on machine learning software application specifics and never claimed I could or would. I made general comments about the flaws in machine learning and neural networks when specifically applied to financial algorithms. I am not going further into this because I think I made my point clear, I laid out several examples to support my point. If this is not enough then I am sorry you may need to engage with others who are posting as much as your 1600+ pieces in less than a year here on this site, I am a little too busy with my own work.

Dude, simply ignore my advise and spend your time on learning MLAlgos, be my guest. Do whatever you please to do. I volunteered advice, if all you can to add value is to make false accusations then save my and your time and move on....

Quote from gmst:

Mate, you have a good degree and you worked on an exotics desk - alright understood. What OP is saying is that you don't need much complicated maths to make money trading delta1 in a systematic/quant way. Invoking your understanding of stoccal and PDEs doesn't mean that OP is wrong. Infact OP is correct.

Anyways, you have been harping on here about you being a quant trader and ibank experience and masters in finance and StocCal and PDEs and Measure theory and what not. Care to mention your sharpe and PF for last 2 yrs? I bet a few traders on this board will beat you on those measures. Its good to keep your ego in check.

Anyways, do you realize you have been running away from engaging in a point by point technical discussion with braincell. You are just commenting on general points. I know your intention is good but you lose credibility when you run away from serious discussion. I was like you when I was in grad school. Know it all guru types. I am no more that way - markets humbled me

gmst · Apr 20, 2012

Quote from amazingIndustry:

* Please point out in detail(!) where I evaded in-detail discussions.
if all you can to add value is to make false accusations then save my and your time and move on....

First, I am not making false accusations, I will copy paste in this message and prove it, will also point out where you evaded in-detail discussion.

I have already said in an earlier post of mine that on MLearning and naive data-fitting, we are on the same page. My issue with you is that you are evading specific questions and trying to invoke your pedigree to win the argument instead of point by point discussion. The examples you have given are very general examples like butter prices in Bangladesh against SP performance, and don't deal with same/similar asset class.

By the comment I was like you when I was in grad school, I am not implying that I am far ahead of you now. No. I am saying when I was in grad school I had these kind of tendencies to invoke my prior background among my fellow grad students (which was stronger than others in the class) and talk about generalities and avoid specific discussions.

Anyways, I will point you out where you have refused to answer or evaded the specific questions.

1. I asked you about your PF and Sharpe. I didn't ask about your PL or your identification information. You should have answered that. If your sharpe is 0.8, we know how much weight to put into your comments vs if your sharpe is 2.2. When on one hand, you are invoking your PDEs and financial industry knowledge to establish your guru status, you should not run away from mentioning your trading performance parameters. You are obviously not a snakesoil salesman but you are displaying similar tendencies, by disclosing favorable information to establish your online identity and running away from disclosing information which might put a dent in your online identity.

2. I mentioned that simple mathematics and very simple trading strategies can make money and to make money doing delta1, you don't need any StocCal or things like that. If you agree, you should say so, if you disagree I would be happy to prove you wrong.

3. Again, you evaded following because answering these might have put a dent in your supposed guru status:

Quote from gmst:

Amazing industry, Let me ask you an unrelated question. For how many years have you been trading, and how many *different* strategies have you devised based on observed market dynamics? And is most of your work on US Equities?

4. Following is a very specific example in which braincell tried to engage you in a specific discussion, you should go back and read the response you wrote to him. Just generalities to a specific question!!! Common dude, if you don't want to engage in a little technical discussion, then don't invoke your pedigree to suggest what is right or not. Otherwise, answer the specific question.

Quote from braincell:

Ok now we get to walk forward testing. If we get two OOS samples, 25% of data each, we can find 100s of systems and see if positive OOS1 data correlates to positive OOS2 data and the scatter graph looks nice, what, you're saying that means nothing - again? My opinion is if there is correlation in OOS1-OOS2, we can safely assume that when we get only OOS1 that is positive, OOS2 (which now becomes live trading) should have a statistical chance of holding up which is equivalent to the previous OOS1-OOS2 tests. Anything wrong with that logic?

amazingIndustry · Apr 20, 2012

a) I gave a lot more examples but I guess you were to preoccupied with other things

b) generally people coming right out of grad school are a lot more specific and detailed because they lack working experience in most cases and fail to often see the big picture while career practitioners are often generalizing rather than digging into details from the outset of a discussion. Hence, you should actually observe the opposite. Those who can describe things in simple terms and can extrapolate imho are more grown and mature than those who try to win arguments biting their teeth into a single detail and lose sigh of the actual point they try to make. If you do not comprehend what point I tried to make then I failed explaining it or your intellect could not grasp it, make a choice

c) I should have answered the Sharpe ratio question? I had to? you forced me? What a troll you are. I have to do nothing! So what if I tell you some of my hft strategies sport an annualized 8-10 Sharpe based on daily returns and return vol with 95% confidence? Next you need to see proof? I am not engaging with you in a pissing contest, so rest it. I never made any claim whatsoever in regards to performance.

d) I never said all approaches to trading involve higher level math. Quote me please.

e) I traded for over 14 years 6 of which are post grad school in several tier one ibanks and hedge funds, after which I started running my own hft venture. I posted those numbers before if you did not pay attention then it speaks to your ignorance.

f) I answered his question in much detail, my point AGAIN was that correlations alone do not matter much, neither does statistical significance in isolation. One example is that the OOS could be way too short a period to capture different market dynamics. Even you have 10 different statistically significant OOS results that speak for an implementation of your strategy you may not at all have captured different market dynamics. There are aspects involved in arriving at a truly promising strategy that can only be learned from actual risk management and risk taking experience. No book will teach you that, no time series analysis will get you that.

Sorry but you are a total troll because you have no issues with snake oil salesmen who are cocky and arrogant but you take issue with those who may be arrogant and cocky but definitely have put in the work and have plenty enough experience and the educational background. For whatever reason you appear very insecure by attacking on a personal level. If you believe I pulled out all the things I wrote out of my ass or copied some wiki pages then I rest my case. This is my last msg to you I have no inclination to battle about irrelevant things.

Quote from gmst:

First, I am not making false accusations, I will copy paste in this message and prove it, will also point out where you evaded in-detail discussion.

I have already said in an earlier post of mine that on MLearning and naive data-fitting, we are on the same page. My issue with you is that you are evading specific questions and trying to invoke your pedigree to win the argument instead of point by point discussion. The examples you have given are very general examples like butter prices in Bangladesh against SP performance, and don't deal with same/similar asset class.

By the comment I was like you when I was in grad school, I am not implying that I am far ahead of you now. No. I am saying when I was in grad school I had these kind of tendencies to invoke my prior background among my fellow grad students (which was stronger than others in the class) and talk about generalities and avoid specific discussions.

Anyways, I will point you out where you have refused to answer or evaded the specific questions.

1. I asked you about your PF and Sharpe. I didn't ask about your PL or your identification information. You should have answered that. If your sharpe is 0.8, we know how much weight to put into your comments vs if your sharpe is 2.2. When on one hand, you are invoking your PDEs and financial industry knowledge to establish your guru status, you should not run away from mentioning your trading performance parameters. You are obviously not a snakesoil salesman but you are displaying similar tendencies, by disclosing favorable information to establish your online identity and running away from disclosing information which might put a dent in your online identity.

2. I mentioned that simple mathematics and very simple trading strategies can make money and to make money doing delta1, you don't need any StocCal or things like that. If you agree, you should say so, if you disagree I would be happy to prove you wrong.

3. Again, you evaded following because answering these might have put a dent in your supposed guru status:

4. Following is a very specific example in which braincell tried to engage you in a specific discussion, you should go back and read the response you wrote to him. Just generalities to a specific question!!! Common dude, if you don't want to engage in a little technical discussion, then don't invoke your pedigree to suggest what is right or not. Otherwise, answer the specific question.