Hours spent developing a trading system

[pavilion103]

I currently work a full time job and am new to trading. I am curious as to what people consider an adequate amount of hours per week to spend working on a trading strategy/system.

At the moment I am spending an average of 14 hours per week doing this, including week nights and weekends. Is this an adequate amount of time to commit per week if I am looking to commence trading within the next couple of months (I have already spent 2 months working on this)?

Any thoughts or advice would be much appreciate. Thanks

 

[Tanaka]

Quite an individual thing I guess. I spent about 20 hours a week for 6 months until I started to get the results I wanted. I was clueless when I started. I had no understanding of the fundamentals of a good system, position sizing, risk/reward ratios, expectancy, slippage, draw downs etc… Actually, it probably took me 6 months to find which volatility level suited my character.

If you are looking at mechanical/technical systems having a good backtesting program will greatly speed up the process.

 

[pavilion103]

Quite an individual thing I guess. I spent about 20 hours a week for 6 months until I started to get the results I wanted. I was clueless when I started. I had no understanding of the fundamentals of a good system, position sizing, risk/reward ratios, expectancy, slippage, draw downs etc… Actually, it probably took me 6 months to find which volatility level suited my character.

If you are looking at mechanical/technical systems having a good backtesting program will greatly speed up the process.

What are some of the better backtesting programs? I am new to trading and have no idea which ones to even look at.

 

[pixel]

Quite an individual thing I guess. I spent about 20 hours a week for 6 months until I started to get the results I wanted. I was clueless when I started. I had no understanding of the fundamentals of a good system, position sizing, risk/reward ratios, expectancy, slippage, draw downs etc… Actually, it probably took me 6 months to find which volatility level suited my character.

If you are looking at mechanical/technical systems having a good backtesting program will greatly speed up the process.

Six months rather full-time should give you a reasonable grounding - provided you have some sort of background that lets you relate to numbers, mathematical models, and a smattering of ability to read a financial report.
I had pretty much all of the above, so I thought I'd start trading - or rather: buying some shares and selling when they showed a profit.
After about three years, I found what I had been missing:
All my tutors and seminar leaders had assumed I knew all about diversification and capital management. So they hadn't even bothered mentioning those topics.
I had also assumed that those financial advisors, who had a license and/or proclaimed they were experts in a particular area, were in fact as clueless as the next pedestrian in a shopping mall.
OK, so I'm a slow learner - but it wasn't until I set up my own rules and traded stocks after backtesting my personal setup that I made consistent profits to live on.

Setting up a personal trading system and backtesting it took me about 3,000 man-hours. If anybody believes they have found a reliable mechanical backtesting program, please PM me. I'd be thrilled to know how they do it. I found that all systems will either err to the cautious side, or make unrealistic assumptions as to how many shares you bought at what price, and how many and when you sold them.
The data that you need: traded volume and price at every instant after your buy or sell signal popped up, are simply not available. It's great to scan the Market for a buy signal 2 hours after Market Close, then have the system "assume" you bought on that day, maybe even at the Intraday Low.
OK, so the system can give you a reasonable "hint", but it's then your job to dig out the daily (or intraday) trading data to find out, at what price you could have realistically bought the thing. Lots of manual work that can run into hours for the results of just one watchlist - let alone Index, such as XJO or All Ords.

Data of 10-15 years ago will also rarely include ALL shares available at the time, but only those that have survived to today.

 

[tech/a]

At the moment I am spending an average of 14 hours per week doing this, including week nights and weekends.

What is it you are doing with your 14 hrs a week if you are new to trading and dont have testing software?

Amibroker coupled with Tradesim Enterprise edition will do most.
I use Metastock only because I started with it.

OK, so the system can give you a reasonable "hint", but it's then your job to dig out the daily (or intraday) trading data to find out, at what price you could have realistically bought the thing.

Common Noob error.
Simply delay buy or sell by one period.
You Dont need intra day for EOD systems testing.

If your a complete novice as in have to learn some language then will take a few years.
Worth it though.
You'll learn much about what doesn't work in your journey to finding WHAT works and WHY.

Data of 10-15 years ago will also rarely include ALL shares available at the time, but only those that have survived to today.

Survivorship great topic but not for this thread.
If you KNOW WHY your system works (Thanks Nick Radge) survivorship becomes less of an issue.

 

[TulipFX]

The rule of thumb is that it take 10,000 hours to become an 'expert' at something. I have found this to be a pretty accurate number.

 

[howardbandy]

I agree with TulipFX.

Barriers to entry to trading are very low, rewards for being right are very high, competition is brutal. Many of the people on the other side of your trades have been at this all their lives, including multiple university degrees, on-the-job apprenticeships, and trading tools common folks have no access to.

Study everything you can, then pick a style that fits with both your personality and whatever else is going on in your life. You will need to be a programmer to write, modify, and test your own systems. You will need to be a statistician to know whether your systems are valid before you start trading them, and whether they are still working as you trade them.

10,000 is four years full time -- the time it takes to do a university program.

Thanks for listening,
Howard

 

[AMSH]

There's a really simple answer to this question - you work until you KNOW that your approach is statistically profitable. Putting arbitrary time limits on when to start trading isn't going to do you any good.

How long will it take? Like Tech said, knowing what isn't important is just as important as what is - unfortunately a lot of your time will be spent on figuring out the former (as there's a lot more that doesn't work than what does). It all depends on how you approach it.

The fastest way to figure out what's important and what isn't is to take a quantitative approach to the markets - treat your entire data set as just that - they're not charts, they're not companies, they're simply data containing (mostly) noise and (some) relationships. When you can discern a relationship, you can trade profitably. Whatever type of trading you want to do (discretionary, mechanical, fundamental, a mix of all), this is the ONLY way to go in my opinion.

As an example, you can sit looking at breakout charts for 2 years, trying to discern how price behaves given certain setups and conditions. You might become proficient at this type of trading and make money. BUT, if you spend 3 weeks learning to code, you can test (literally) tens of millions of different setups, combinations of conditions, time periods etc etc etc. Even if you want to be a discretionary trader - this is the fastest, most reliable way of building your knowledge of how price might behave - in any situation you care to look at. Mate, buy Amibroker, learn to code, read about the quantitative approach (Howard's books are very good) and enjoy it. You'll enjoy trading far more if you're making money even if it takes you longer to start.

Sorry if that was a bit self-indulgent.

Cheers,

Al.

Quick P.S. to Pixel. It's quite simple to adequately model your entry prices when backtesting EOD systems - I generally add two ticks of slippage from the open (if buying on the open) for any system I want to trade - you can also test whether your entries are realistic after the fact by comparing your actual entry price to that indicated by your backtesting - if they differ, you simply alter your entry component accordingly. If you're interested PM me and I'll send you some code.

 

[Grep]

It all depends on how much you already know about what works in systematic trading.
Do you know what the components of a trading system are ? Do you know what your final system will look like, e.g make the biggest annual return possible or provide the lowest drawdown possible. Maybe you want both.
Can you write code to test these components over some historical data.

For example, you have an entry idea, an exit idea and a position size idea.
You might discover through testing that when I tighten the exit price I make more money but have a lower win percentage.Do you like this outcome.Do you understand why that happened to your system.

Eventually you will figure out what each component does and how they interact with each other.You might even discover that small changes in one component always has a bigger effect on the outcomes than similar changes in the other components,regardless of what the system is.

You might find that really simple ideas work better than cosine adjusted fractal oscillators that have been optimised with artificial intelligence algorithms.

I use Amibroker and excel.

 

[brty]

I find these comments from some esteemed posters to be interesting...

Can you write code to test these components

Mate, buy Amibroker, learn to code,

You will need to be a programmer to write, modify, and test your own systems

If your a complete novice as in have to learn some language then will take a few years.

If the above were true, then successful trading approaches have only worked since the invention and use of computers. Before that time everyone had to be a loser. I'm afraid I don't buy the argument of needing to be able to code. Perhaps I'm wrong.

Can anyone code "meandering" or perhaps "good support" or "strong resistance" ??
I find my understanding/use of these types of events hugely beneficial to the bottom line of trading.

brty

 

[pavilion103]

Thanks for the help guys.

I had a look on the AmiBroker website. I had looked at it previously but not purchased it. I'm assuming I should get the professional version for $279 rather than the standard for $199?


In response to what I have been doing so far: I have ready about 10 trading books, some of the better ones being "Trade Your Way to Financial Freedom", "Way of the Turtle" and "The Universal Principles of Successful Trading".
I've created an Excel trading "template" that calculates the R for each trade, % gains and other statistics. In fact I'd love to send it to anyone who is willing to look over it for me? Just to see if I am completely going in the wrong direction. It would mean a lot to me. Can I post it on here?

Thanks guys, much appreciated.

 

[MRC & Co]

From what I am aware, most use TradeStation if you have a bit of money to spend and once you get even more, a lot use Mechanica now.

But the ones suggested are probably better if you don't have much money and are just starting out.

HowardBandy, what are these trading tools common folk have no access too?

But I agree, to properly test most things, you will probably need a programmer/quant.

Someone also said it above: "You might find that really simple ideas work better than cosine adjusted fractal oscillators that have been optimised with artificial intelligence algorithms." - smart lad.

You will need to learn what works in trading before even bother to start coding (for that, you can just pay someone once you know what will make you the $$$$ and they will help you analyse/optimise certain conditions).

 

[Gringotts Bank]

I find these comments from some esteemed posters to be interesting...

If the above were true, then successful trading approaches have only worked since the invention and use of computers. Before that time everyone had to be a loser. I'm afraid I don't buy the argument of needing to be able to code. Perhaps I'm wrong.

Exactamundo. Aside from TA and FA, there's a whole universe of ways to go about finding stocks that go up in price. I know a guy whose research consists of reading the Fin Review each day. If he likes the story (story>FA, though they are linked of course) he buys. He does extremely well and admits he has trouble reading mining reports, so could not be considered an FA guy. Also possible to make money if you have a cluey broker with connections and insider info. Another way to do well is to find a poster on a forum like this or HC and buy what they buy. Don't laugh, my limited research in this area indicates that you can do fairly well following those who consistently place well in the HC competition. If you follow a certain poster, try to find one who is low key and posts out of interest rather than for notoriety. It's easy to spot the difference. Then there's investment newsletters. I know one guy who comfortably lives off a dividend type portfolio designed with the help of Huntley. There;s many other ways as well. Don't limit yourself...

 

[motorway]

I find these comments from some esteemed posters to be interesting...

If the above were true, then successful trading approaches have only worked since the invention and use of computers. Before that time everyone had to be a loser. I'm afraid I don't buy the argument of needing to be able to code. Perhaps I'm wrong.

Can anyone code "meandering" or perhaps "good support" or "strong resistance" ??
I find my understanding/use of these types of events hugely beneficial to the bottom line of trading.

brty

I agree with brty

the only things worth testing are real things
This is a very much smaller universe than unreal things

looking at things people do test
seems to involve a lot of things that do not mean much and never will
eg they are testing in the universe of the unreal..

should my indicator be X look back
should my signal be X number of bars from etc

Motorway

 

[Wysiwyg]

If anybody believes they have found a reliable mechanical backtesting program, please PM me. I'd be thrilled to know how they do it. I found that all systems will either err to the cautious side, or make unrealistic assumptions

My thoughts too. After countless hours back testing and walk forward testing I have come to the same conclusion. The mind, with experience/knowledge, can process all the incoming data and adjust accordingly while a mechanical approach is not flexible.

How is this for simplicity?

An entry price with a higher probability based upon and/or

a) F/A
b) T/A

An exit price based upon and/or

a) Time
b) %/$ gain or loss
c) T/A
d) F/A

coupled with risk controls relative to tolerance or bank account.

 

[AMSH]

Brty, Motorway et al you're right - you don't need to be able to code to trade profitably; people trade using all sorts of methods. But we're talking about a beginner who's trying to find the best way to approach learning how to be a profitable trader - what are the best tools to learn how the market really works? Is it to slog it out in front of a screen looking at 1,000 charts, or is it to get your software to do that in 10 seconds, get a summarised version of the results and then move on to the next idea?

Discretionary trading is in some ways far superior to a quantitative approach - there's some stuff that simply can't be coded. But to grow your knowledge base and gain information to become (perhaps) a great discretionary trader, using software and being able to code is a necessity. Again, only my opinion.

Isn't it logical that if you trade a method without knowing how it's going to behave, you're essentially gambling. I'm pretty sure that if anybody could pick up a newsletter, trade the recommendations and beat the broad index, 1) there wouldn't be any question about whether the market is efficient or not (it's not but that's beside the point) and 2) there would be far more newsletters.

Motorway, I think indicators are useless and I'm not trying to promote them - OHLCV is all the information that you have, and it's all that's needed - indicator simply distort the basis of what you're trying to measure.

Cheers,

Al.

 

[chrislp]

But we're talking about a beginner who's trying to find the best way to approach learning how to be a profitable trader - what are the best tools to learn how the market really works? Is it to slog it out in front of a screen looking at 1,000 charts, or is it to get your software to do that in 10 seconds, get a summarised version of the results and then move on to the next idea?

If you're a beginner I personally think it's the former.

I don't see how you can have ideas without seeing how the market works first.

 

[Gringotts Bank]

A couple of things any beginner could do to make a quick start:

1. Plot a Parabolic SAR(.017,.017) on the All Ords. Stay out of trading (or reduce your exposure) when it's below the line. Sure, it's curve fitted, but as far as broad indicators go, it's ok. Most of the time, when it's below the line, the market is bearish. Sometimes it can be below the line and just 'taking a breather'. You tell the difference by using #2.

2. Monitor the number of micro-cap stocks that are in the +10% range for the day (on decent volume). This is an easy way to get a feel of short term sentiment. The All Ords will not give you this information.

3. Of the stocks in #2 above, how many of these are well known trading stocks?*
eg. If amongst your +10% stocks you have SDL, CCC, SSN, CDU, etc. making new short term highs, then the market is sufficiently bullish in the short term. If on the other hand your list of +10% stocks contains only biotechs and unheard of lightly-traded mining stocks having a one-day burst, the short term is much less likely to be bullish.

4. When #1-3 are in your favour, choose a method of stock picking, understand your time frame is "short term", and roll the dice.

5. if you win, roll again using a bigger stake. If you lose, reduce your stake size. If you make a lot of losses, stop and review your stock picking method.


*Trading stock: Any stock that features consistently in forums like HC, or any stock that has decent 'daily value $ traded' and has made 100% or more over the past few months.

 

[Frankie]

I would like to make a distinction between:

a) Developing a trading system
b) Applying a trading system.

Developing a trading system

Personally, I am always trying to develop or improve my trading system. I look for ideas when scanning the paper. I look for ideas when analysing how others trade. Development is an ongoing process and comes out of my deep interest into the subject of trading.

Applying a trading system

I apply my system by going through my trading checklist. I apply my system before market opens, during market open and at market close. By applying my trading checklist I spend around 30mins or less on trading each day.

I believe in following a trading plan, and taking as many shortcuts as possible!

 

[tech/a]

I would like to make a distinction between:

a) Developing a trading system
b) Applying a trading system.

Developing a trading system

Personally, I am always trying to develop or improve my trading system. I look for ideas when scanning the paper. I look for ideas when analysing how others trade. Development is an ongoing process and comes out of my deep interest into the subject of trading.

Applying a trading system

I apply my system by going through my trading checklist. I apply my system before market opens, during market open and at market close. By applying my trading checklist I spend around 30mins or less on trading each day.

I believe in following a trading plan, and taking as many shortcuts as possible!

You have a trading Plan or hypothesis.
You DONT have a Trading System.