Scientific Approach to Trading

[shyguy 10]

It seems very few people take a scientific approach to trading, and even fewer are willing to share their findings with the public. Since I first started researching trading I have heard people talking about the need to trade a system. That is, to have a defined set of criteria to enter a trade, and only enter a trade when said criteria are met. Technically speaking the same should be done for exits.

 

Once this has been done, the next step should be backtesting the strategy. Assuming the strategy has condiditon which cannot easily be programed into a computer (such as the price being at S/R levels), one should just start scrolling through old data until an entry condition is given, then examine the results. By keeping track of all the historical entries given, one can get a good idea of what to expect from said strategy in terms of win/loss rate, draw down, potential returns. If historically the strategy is profitable, and you start trading the strategy, this will let you know when perhaps it needs to be adjusted or temporarily removed.

 

Once you have developed a strategy, done histroical testing, obtained your historical performance data, and started demo trading the strategy, one should start over from scratch with a new strategy.

 

This process should be repeated until you are trading multiple strategies at the same time. I feel like if most successful traders have a trading strategy as with clearly defined entry/exit conditions (as in they never say "Hmmm should i enter this trade?") we could start compiling a list of these systems.

 

Many people have bveen generous enough to post their system on a variety of forums across the web, but it doesnt seem like anyone has taken the time to scientifically evalutate the effectiveness of the system through historical testing or even define specific entry conditions.

 

Why does nobody do this?

[steve46 12]

It seems very few people take a scientific approach to trading, and even fewer are willing to share their findings with the public. Since I first started researching trading I have heard people talking about the need to trade a system. That is, to have a defined set of criteria to enter a trade, and only enter a trade when said criteria are met. Technically speaking the same should be done for exits.

 

Once this has been done, the next step should be backtesting the strategy. Assuming the strategy has condiditon which cannot easily be programed into a computer (such as the price being at S/R levels), one should just start scrolling through old data until an entry condition is given, then examine the results. By keeping track of all the historical entries given, one can get a good idea of what to expect from said strategy in terms of win/loss rate, draw down, potential returns. If historically the strategy is profitable, and you start trading the strategy, this will let you know when perhaps it needs to be adjusted or temporarily removed.

 

Once you have developed a strategy, done histroical testing, obtained your historical performance data, and started demo trading the strategy, one should start over from scratch with a new strategy.

 

This process should be repeated until you are trading multiple strategies at the same time. I feel like if most successful traders have a trading strategy as with clearly defined entry/exit conditions (as in they never say "Hmmm should i enter this trade?") we could start compiling a list of these systems.

 

Many people have bveen generous enough to post their system on a variety of forums across the web, but it doesnt seem like anyone has taken the time to scientifically evalutate the effectiveness of the system through historical testing or even define specific entry conditions.

 

Why does nobody do this?

 

My process is very nearly the same, and many of my colleagues do much the same....So your comment "why does nobody do this?" is simply put....incorrect...

 

It would be more accurate to ask "why don't more retail traders do this"?....and of course the answer is...that this is one of the differences between professionals and amateurs...I think the bottom line reasons are along the line of laziness, incomptence, lack of education, etc.....your guess is as good as mine. As an amateur if I knew this was the difference between making money and losing it, I think it would motivate me to do the research and figure it out...

 

One problems with the concept of "scientific" system development and trading (in this venue at least) is that to do a high quality job, one needs significant educational background and skills. For example most folks who try to write a rule driven system do not understand that all systems are affected by cyclical stationarity (depending on time frame and market traded)...Because they don't see this problem, they may write up a set of rules that "backtest" nicely, but for reasons unknown, once they go live, starts losing money.

 

While I am thinking about it....it occurs to me that YOU might want to simply forego posting and go do this yourself...right?...if you know what to do (as you've just suggested), why not just go do it.....?

 

Good luck to you..

Edited March 31, 2011 by steve46

[SIUYA 11]

"Hmmm should i enter this trade?"

 

I usually think this and assess it against my ideas/rules/entry parameters. But then I apply context - this comes from years of experience.There is nothing scientific about it.

I would like to automate everything - but this seems to be a dream for me to date.....if so then I could back test and show you a set of possible scenarios.....but this is still not scientific.

 

But even after the entry the rest for me is then ""Hmmm should I exit or keep this trade"" again while a lot of rules can be used, in reality a lot of the time there is a certain amount of feel.

So much the same as the great post by Steve46, everything is largely process driven, but there is also lot of feel.

 

example; I trade stocks, today, a rare earths miner LYC - ASX raised capital. The first instance in many cases such as this is the share price goes down. However in this case the share price was likely to rally for a number of reasons..... 1) that area of the market is hot, 2) the capital raised is used to double production 3) it was well received and expected. So in this case I expected the stock to go up.....and it did.

For me the trade was to find a highly correlated related stock ARU-ASX..... this was what I bought on the open ....as the open was not too far from the close.....this then rose another 3-4% on the day.....I sold 1/3 of my position, and I will decide what to do again tomorrow, as I expect it will continue to rally - how hard who knows, but I am also quite happy to cut it at breakeven or a small loss.

I dont think it would be easy to make a scientific study of this, as you might be able to see my trades but not my rationale.

[steve46 12]

The gentleman doesn't really want to make a scientific study....as you may already have guessed....

 

I was tempted to provide some interesting "data" of my own, just for laughs, but really I cannot bring myself to do it..

 

Oh well, enough of this...2am here on the west coast and time to close up shop...

[wrbtrader 10]

It seems very few people take a scientific approach to trading, and even fewer are willing to share their findings with the public. Since I first started researching trading I have heard people talking about the need to trade a system. That is, to have a defined set of criteria to enter a trade, and only enter a trade when said criteria are met. Technically speaking the same should be done for exits.

 

Once this has been done, the next step should be backtesting the strategy. Assuming the strategy has condiditon which cannot easily be programed into a computer (such as the price being at S/R levels), one should just start scrolling through old data until an entry condition is given, then examine the results. By keeping track of all the historical entries given, one can get a good idea of what to expect from said strategy in terms of win/loss rate, draw down, potential returns. If historically the strategy is profitable, and you start trading the strategy, this will let you know when perhaps it needs to be adjusted or temporarily removed.

 

Once you have developed a strategy, done histroical testing, obtained your historical performance data, and started demo trading the strategy, one should start over from scratch with a new strategy.

 

This process should be repeated until you are trading multiple strategies at the same time. I feel like if most successful traders have a trading strategy as with clearly defined entry/exit conditions (as in they never say "Hmmm should i enter this trade?") we could start compiling a list of these systems.

 

Many people have bveen generous enough to post their system on a variety of forums across the web, but it doesnt seem like anyone has taken the time to scientifically evalutate the effectiveness of the system through historical testing or even define specific entry conditions.

 

Why does nobody do this?

 

Hi,

 

I'm a member of many different forums that are popular, not popular, small and large almost since their birth as in I've joined within a few months after their birth.

 

Most traders that have a more systematic approach (I prefer to use that word than scientific approach) including discretionary traders (those not using a coded system or automation) have done exactly what you've mentioned. Some have been very genereous and freely shared their methods, backtesting results and fewer are currently maintaining online trade journals.

 

Simply, I'm not sure why you say "nobody doing this" because I've seen many, many and many do such in their attempts to find a profitable trade method unless you meant why hasn't anybody posted the results of all these trading systems into one single thread instead of viewing them spread out at many different forums.

 

That's a daunting task for any trader. I just can't imagine someone finding the time & energy to compile such a list. Further, I'm going to make a good guess that I've seen at least 800 systematic trading methods along with variations posted online across many forums. I can easily see someone or a few spending a few years of full-time work making such a list and then maintaining it for accuracy in one thread like some "international one public library of systematic trade methods". I use the word "international" because there are also many trading forums in different languages besides English.

 

Lets put it this way, YOU can start by listing your favorite 10 systematic trade methods in full detail in this thread and please write the direct links to the forum where the method was originally revealed. This will give you an ideal about how much time & energy will be needed to maintain a list of just 10 systems in comparison to the hundreds or maybe thousands that's been posted the past 10 - 20 years online at many different online forums.

 

Regardless what was meant or how I interpreted what you said, most traders that aren't using an automated trading system will often discover that no matter what parameter they change (indicator or price action)...the biggest piece of the puzzle is not the method...instead it's them or the trading plan that involves many other variables working together (e.g. market experience, money management, capitalization, discipline, proper equipment, proper trading environment, costs of trading and so on).

 

My point is that the few profitable traders out there that's using a systematic trade method have discover that their profits involves many variables working together (called a trading plan) instead of just the trade method itself and arguably those other variables are just as important or more important than the trade method itself depending upon the profitable trader.

Edited March 31, 2011 by wrbtrader

[Tradewinds 12]

One problems with the concept of "scientific" system development and trading (in this venue at least) is that to do a high quality job, one needs significant educational background and skills.

 

My impression is, that most traders are not programers. In order to test rules with a computer program, a computer program needs to be created. I'm not saying that rules can't be tested without a computer program, but without the computer program, user discretion is very difficult to avoid. And if user discretion can't be avoided, then it is no longer a scientific method. If the trader doesn't follow the rules exactly, the data analysis can't really be matched to the rules.

 

I've studies charts for thousands of hours looking for patterns. Then I program rules to match a perceived pattern. Almost 100% of the time, the programed results show me that my perception was extremely flawed. The computer program has no discretion. It does exactly what I program it to do. When I am looking for price patterns, I see what I want to see; I avoid or ignore information that proves my perception wrong. I use wishful thinking.

 

I've been through the same situation probably 1000's of times. I think I see a cause and affect price pattern, and when I program a rule, it turns out to be almost nothing more than purely random. Unless a trader has experienced this, it's hard to believe, but true.

[Tradewinds 12]

I usually think this and assess it against my ideas/rules/entry parameters. But then I apply context - this comes from years of experience.There is nothing scientific about it.

I would like to automate everything - but this seems to be a dream for me to date.....if so then I could back test and show you a set of possible scenarios.....but this is still not scientific.

 

Trader experience and situational context would be extremely difficult to program. I'm not saying it couldn't be done, but it would probably take a team of people and millions of dollars. Like the program and computing power needed to play Jeopardy.

 

How would a computer program match how you interpret or strategize for a news event? I'm not saying it couldn't be done, but it would be a major undertaking.

[Tradewinds 12]

I have seen a retail trading platform that looks very easy to use for backtesting, and even has options for avoiding certain times of the year, or certain times of the day; when scheduled news comes out. I think it was Trade Navigator™. I don't own it, but I was impressed by the demo of the backtesting.

[shyguy 10]

My process is very nearly the same, and many of my colleagues do much the same....So your comment "why does nobody do this?" is simply put....incorrect...

 

It would be more accurate to ask "why don't more retail traders do this"?....and of course the answer is...that this is one of the differences between professionals and amateurs...I think the bottom line reasons are along the line of laziness, incomptence, lack of education, etc.....your guess is as good as mine. As an amateur if I knew this was the difference between making money and losing it, I think it would motivate me to do the research and figure it out...

 

One problems with the concept of "scientific" system development and trading (in this venue at least) is that to do a high quality job, one needs significant educational background and skills. For example most folks who try to write a rule driven system do not understand that all systems are affected by cyclical stationarity (depending on time frame and market traded)...Because they don't see this problem, they may write up a set of rules that "backtest" nicely, but for reasons unknown, once they go live, starts losing money.

 

While I am thinking about it....it occurs to me that YOU might want to simply forego posting and go do this yourself...right?...if you know what to do (as you've just suggested), why not just go do it.....?

 

Good luck to you..

 

I guess I should be asking, why does nobody do this and share their results online?

 

I have been doing this actually, it just takes a long time to do one system. It's about a weeks worth of solid work to complete back testing by hand. Now, to be clear, im not talking about having a computer test the system. What I've been doing for the system I've been testing, is bring up the 15 minute chart for the day in question, and scroll through candle by candle until I see what looks like an entry. This allows me to take into account factors that you would need some very sophisticated programming to consider. It basically lets me say "If I were trading this system on March 2, 2006, knowing what I know as of march 2, 2006, I would have taken the following trades as dictated by systemX, and they would have performed as such. Backtesting systems with a computer in this way is simply not possible (well, not profitable I should say).

 

I would not want to only trade this way. As SIUYA pointed out, mechanical trading puts limitations on you which make it extremely hard to account for and quantify all variables (such as news and S/R). I just feel like if you traded these mechanical systems in addition to how you trade normally, it would just increase your overall profit (if systems were profitable of course). You might, for example, make say 10% a month trading normally, and 1% from a system.

 

"Some have been very genereous and freely shared their methods, backtesting results and fewer are currently maintaining online trade journals."

 

@wrbtrader - if you could link me to a thread where someone has done this, that would be much appreciated. I have never found a system where someone has actually provided the results of trading it everyday. Lots of people log their trades, and post charts of the trades they took, but I'm looking for numbers.

 

It's obviously a lot of work to go through and test a few systems like that, but if we had a group of say 10 people, with two people testing each system, it would be practical. Once the historical testing is done, maintaining it would be no more difficult than keeping a trade log, or more simply, just entering your trades onto a spreadsheet. I'll try and post up an example of one I've been working on when I'm finished.

 

Thanks for all the interesting responses.

[emg 12]

 

Many people have bveen generous enough to post their system on a variety of forums across the web, but it doesnt seem like anyone has taken the time to scientifically evalutate the effectiveness of the system through historical testing or even define specific entry conditions.

 

Why does nobody do this?

 

Only a 3rd party educational vendors will sell their systems.

[Tradewinds 12]

but if we had a group of say 10 people, with two people testing each system, it would be practical.

 

I think a group effort would be great, but getting people together would be difficult I think. Someone would need to be the champion of the cause, and be able to coordinate and motivate the group. There would need to be a clear incentive, and a way to manage expectations and reward people for the end result.

 

You would need to find people interested the investment product you are looking at, the same time frame, the same style of trading and the same perspective.

 

And you need to be able to convince people that the effort would be worth it.

[steve46 12]

Only a 3rd party educational vendors will sell their systems.

 

In my office we have several bright young guys who do nothing but develop systems..they are paid to develop and then sell their system to our employer..so on the institutional side of this industry there are many people who sell their systems, in fact most of the intraday volume on the NYSE consists of automated execution of trading systems...Also to my knowledge there are several big firms who specialize in automated (basket execution of all kinds) and they are required to list their programmed trades publicly every day. These "systems" are developed and operated by employees and consultants of all kinds.

 

To be correct, "independent" free lance vendors catering to the retail crowd are the ones selling systems that simply don't produce consistent long term profitability...and there are many good reasons for that...

 

and I don't mean to single you out EMG because there are a lot of folks who have a screwed up idea of how this business works....and to the extent that some of them may want to eventually become professionals it may be a good idea to obtain a more realistic picture of what is actually going on...

Edited March 31, 2011 by steve46

[wrbtrader 10]

 

"Some have been very genereous and freely shared their methods, backtesting results and fewer are currently maintaining online trade journals."

 

@wrbtrader - if you could link me to a thread where someone has done this, that would be much appreciated. I have never found a system where someone has actually provided the results of trading it everyday. Lots of people log their trades, and post charts of the trades they took, but I'm looking for numbers.

 

It's obviously a lot of work to go through and test a few systems like that, but if we had a group of say 10 people, with two people testing each system, it would be practical. Once the historical testing is done, maintaining it would be no more difficult than keeping a trade log, or more simply, just entering your trades onto a spreadsheet. I'll try and post up an example of one I've been working on when I'm finished.

 

Thanks for all the interesting responses.

 

Hi,

 

I'm a little surprise that you haven't seen such. It seems to me to be common place at most forums. Lots of traders freely share their systematic rules or codes.

 

I'll give you one example of a thread where traders have freely shared the details of the "systematic rules" involving their trade methods @ Trading Systems @ Forex Factory and here's a specific thread where the rules (codes) I shared later in the thread @ Forums - Robotrading: CT + Trending Strategy on folios of futures

 

There's hundreds more and I'm sure you can join any particular discussion forum and do your own research now that you know they exist.

 

Yet, the problem you may have or probably already determine is that some of these systems rules or codes may require you to use the same testing software or accurately transcribe them into your own programming software. You'll obviously be required to "participate" in those threads to learn more information about the system along with asking questions about rule or code you don't understand. That's why I say this is a daunting task you're talking about.

 

The last time someone ask what you asked and I sent him links...he responded that he was looking for "easy systems that were profitable on his trading instrument"...I just thought it was odd that he started adding conditions (e.g. it has to be available for application in his programing software) on to his search as if someone was going to do all the work for him. :rofl:

 

Regardless, if their backtesting results are too short of a time period, you can do your own more extensive result after you learn to details of the methods if what they've shared is not enough info for what you want to do that will allow you to determine how good/bad the system.

 

However, I don't understand why you want to test something when these method have already been extensively tested and discussed et cetera as you'll see at some of those threads by others that are participating in the threads. Simply, you implying testing them on different trading instruments that the method designer doesn't trade ???

Edited April 1, 2011 by wrbtrader

[John Last 10]

Well imagine that sometimes it is a big responsibility to share strict mechanical rules. That is a responsibility beyond you can take.

 

Why?

 

Because you never know how your word can be interpreted and misunderstood.

 

That is why I just share tools and ideas.

[globestocks 10]

Free updates on the latest news and stock picks on the market! Get

free newsletter now!

[zupcon 10]

I guess I should be asking, why does nobody do this and share their results online?.

 

Because there's absolutely no benefit in doing so.