How is quantitative system trading profitable?

[goponcho]

Hi, only still in my early stages of learning about systems trading. I've been learning about the use of amibroker slowly.

One thing that I am still unclear on is where we profit in systems trading. In order for us to profit, we need to be exploiting an inefficiency in the market right? What process goes into discovering these systems which exploit these inefficiencies? Trial and error after having an idea?

I get that we can use Amibroker to explore, backtest and optimize a system, but what process is required to develop a system with a positive expectation, not just a system.

After I have spent more time studying and learning how to create a system, will i definitely be able to find a system that will gain money?

Dont know if i explained myself properly, but yeah. Thanks guys!!


Also, as it is probably relevant, i am aiming to eventually trade a 100-200k account in a medium term time frame.

 

[skc]

Hi, only still in my early stages of learning about systems trading. I've been learning about the use of amibroker slowly.

One thing that I am still unclear on is where we profit in systems trading. In order for us to profit, we need to be exploiting an inefficiency in the market right? What process goes into discovering these systems which exploit these inefficiencies? Trial and error after having an idea?

I get that we can use Amibroker to explore, backtest and optimize a system, but what process is required to develop a system with a positive expectation, not just a system.

After I have spent more time studying and learning how to create a system, will i definitely be able to find a system that will gain money?

Dont know if i explained myself properly, but yeah. Thanks guys!!


Also, as it is probably relevant, i am aiming to eventually trade a 100-200k account in a medium term time frame.

You made a good point. Many people get fancy with their system using all sorts of parameter and indicators, without really understanding why such a combination of numbers worked in the past, or why should/would it continue to work in the future.

And if and when the system stops working, they don't really know how to fix it, as they never knew why it worked in the first instance.

IMO you need to watch the market first, identify an inefficiency first before back testing. Otherwise you are creating a blackbox for yourself...

 

[goponcho]

You made a good point. Many people get fancy with their system using all sorts of parameter and indicators, without really understanding why such a combination of numbers worked in the past, or why should/would it continue to work in the future.

And if and when the system stops working, they don't really know how to fix it, as they never knew why it worked in the first instance.

IMO you need to watch the market first, identify an inefficiency first before back testing. Otherwise you are creating a blackbox for yourself...

Thanks for your response. What kind of inefficiencies are there that we can use? Can you give some examples, even if they're not true just so i can get an feel for what I'm looking for?

Are they just simple, like if MA1 crosses MA2 we buy, and MA2 crosses MA3 we sell and historically we profit? This example doesnt make sense logically, but is just a correlation which would have made us profit in the past. Are the 'inefficiencies' we are looking for logical ones?

I dont think i really know what the inefficiencies are that I'm tlaking about, just know that they exist.

I am planning to keep an eye on the markets after my initial reading, though i have been in the past.

 

[tech/a]

You need to find opportunities where demand outstrips supply which will cause price to rise.
You then need to identify where demand is outstripped by supply which causes price to stall or fall.

How you do that is wide and varied
You'll need to spend the 1000s of hrs investigating the art of technical analysis and how to apply it--- just like everyone else.

 

[AlterEgo]

I don’t think it’s really an inefficiency. You’re just recognising the way stocks tend to move due to human behaviour. If you’re aiming to trade over a medium term timeframe, then I think what you should probably be focusing on is trends – ie. identifying what the current trend is and how to determine when the trend has likely changed. Trends can often persist for a considerable amount of time. The idea is to ride the trend as far as it goes, then exit when it reverses. Maximise your profit when you’re right, and minimise your loss when you’re wrong.

 

[goponcho]

You need to find opportunities where demand outstrips supply which will cause price to rise.
You then need to identify where demand is outstripped by supply which causes price to stall or fall.

How you do that is wide and varied
You'll need to spend the 1000s of hrs investigating the art of technical analysis and how to apply it--- just like everyone else.

Hi tech,

I dont mind spending the hours learning about technical analysis, just wanted to know that there are profitable ways of applying it find inbalances in supply and demand. There is a lot of noise for a beginner trader to filter through.

So the positive expectation comes from greater knowledge of techincal analysis than the market, at buying and selling points?

I don’t think it’s really an inefficiency. You’re just recognising the way stocks tend to move due to human behaviour. If you’re aiming to trade over a medium term timeframe, then I think what you should probably be focusing on is trends – ie. identifying what the current trend is and how to determine when the trend has likely changed. Trends can often persist for a considerable amount of time. The idea is to ride the trend as far as it goes, then exit when it reverses. Maximise your profit when you’re right, and minimise your loss when you’re wrong.

Thanks AlterEgo. I will probably be trading with the trend, as it seems simpler for a beginner. Yeah ill probably look for low risk situations and let it run with maybe a trailing stop.

 

[skc]

Thanks for your response. What kind of inefficiencies are there that we can use? Can you give some examples, even if they're not true just so i can get an feel for what I'm looking for?

Here are a couple of (ficticious) examples.

- CBA and WBC are both influenced by the domestic economy the same way. So their share price movements should and do more or less mirror each other. Now for no apparent reason the share price has diverged. You place a bet on the price reverting to the mean. This is pairs trading using statistical / quantitative tool backed by actual logic.

- The SPI has never risen more than 10 days in a row. That kind of over-enthusiasm is shown to be smacked back down 65% of the time after 7 days, 72% of the time after 8 days etc etc. You have an edge placing a short as supported by the statistics.

With chart patterns, trends, support and resistance etc, you can probably come up with similar thoughts and logic, and test quantitatively whether such logic prevailed.

 

[goponcho]

Here are a couple of (ficticious) examples.

- CBA and WBC are both influenced by the domestic economy the same way. So their share price movements should and do more or less mirror each other. Now for no apparent reason the share price has diverged. You place a bet on the price reverting to the mean. This is pairs trading using statistical / quantitative tool backed by actual logic.

- The SPI has never risen more than 10 days in a row. That kind of over-enthusiasm is shown to be smacked back down 65% of the time after 7 days, 72% of the time after 8 days etc etc. You have an edge placing a short as supported by the statistics.

With chart patterns, trends, support and resistance etc, you can probably come up with similar thoughts and logic, and test quantitatively whether such logic prevailed.

Coooool! Gives me a taste of what i will be looking for.
With the first example, it seems quite specific to a particular pair of stocks. Is it suitable to create an entire trading system with(ficticiously)? From my limited reading, trading systems usually involves applying an idea across a whole range of stocks to find multiple trades, can we do this with such a specific idea?

The second example seems much more testable, and clearer on how the system will form.

I like how you tied how the logic related to creating a trading system, rather than just scouring a whole bunch of data for trades that would have worked in the past, and happen to work in out-of-sample data, without a reason. This is kind of how it seems in my recent reading, which would not leave me confident trading.

 

[wayneL]

I don't know much about quantitative trading, but I always thought quantitative trading was something distinct from technical trading.

 

[skc]

Coooool! Gives me a taste of what i will be looking for.
With the first example, it seems quite specific to a particular pair of stocks. Is it suitable to create an entire trading system with(ficticiously)? From my limited reading, trading systems usually involves applying an idea across a whole range of stocks to find multiple trades, can we do this with such a specific idea?

The second example seems much more testable, and clearer on how the system will form.

I like how you tied how the logic related to creating a trading system, rather than just scouring a whole bunch of data for trades that would have worked in the past, and happen to work in out-of-sample data, without a reason. This is kind of how it seems in my recent reading, which would not leave me confident trading.

There's a whole thread on pairs trading here. https://www.aussiestockforums.com/forums/showthread.php?t=14508

On statistical testing etc here's a good source. http://quantifiableedges.blogspot.com/

 

[tech/a]

I don't know much about quantitative trading, but I always thought quantitative trading was something distinct from technical trading.

Yes you do!
Black Sholes ring a bell?

Technical analysis is simply the vehicle to derive the numbers used in quant analysis.
You can of course get numbers for quant analysis from other areas un related to tech analysis.
So I wouldn't call it distinct--- but then again the use of it (tech analysis) in quant would make it distinct.
It's basically what your doing when you use it in systems design.

 

[wayneL]

Yes you do!
Black Sholes ring a bell?

I guess so LOL. I Just never connected the dots.

Technical analysis is simply the vehicle to derive the numbers used in quant analysis.
You can of course get numbers for quant analysis from other areas un related to tech analysis.
So I wouldn't call it distinct--- but then again the use of it (tech analysis) in quant would make it distinct.
It's basically what your doing when you use it in systems design.

Again yes I guess so.

 

[village idiot]

you play poker dont you, OP?

 

[goponcho]

you play poker dont you, OP?

Yes sir. Do you feel I am thinking about things in an incorrect way because of this?

 

[village idiot]

not neccessarily. you are thinking about it in the correct way in that you need to find situations that are +EV and act upon that in a structured manner, which is a long way ahead of many who start out.

The line you are bringing from poker is that the job is to figure out which are the losing players and how best to exploit them.

Here it doesnt matter who is losing or whether the market is inefficient or not, just concentrate on identifying situations where the balance of probablilities is something other than random, by enough for a bet to be +EV. Which is the track you are on anyway

this is not a bad start;

Trial and error after having an idea?

and you are definitely on the right track with this;

I like how you tied how the logic related to creating a trading system, rather than just scouring a whole bunch of data for trades that would have worked in the past, and happen to work in out-of-sample data, without a reason. This is kind of how it seems in my recent reading, which would not leave me confident trading.

 

[howardbandy]

Greetings all--

Developing trading systems that are profitable is very hard.

Barriers to entry into trading are low, rewards for being right are high, the markets are very nearly efficient.

The basis of any trading system is to identify some inefficiency in the issue being traded -- identify some pattern that precedes a profitable trading opportunity. Making profitable trades removes a portion of the inefficiency making it more difficult to be profitable in the future.

Your competition is well educated, well trained, well paid, well equipped.

I agree with Geoff Colvin in "Talent is Overrated" and Malcolm Gladwell in "Outliers" that it takes on the order of 10,000 hours of high quality practice to become expert at anything. If anything, that is an underestimate for trading.

Thanks for listening,
Howard

 

[mazzatelli]

Technical analysis is simply the vehicle to derive the numbers used in quant analysis.

Hi tech,
Could you elaborate further? - I've hardly seen any technical analysis, unless you count plotting the underlying price on a 2d plane, used in quant analysis.

It's more akin to what skc describes - corr, auto corr,co-integration, residuals on corr etc , basically relative value metrics.

 

[tech/a]

Hi tech,
Could you elaborate further? - I've hardly seen any technical analysis, unless you count plotting the underlying price on a 2d plane, used in quant analysis.

It's more akin to what skc describes - corr, auto corr,co-integration, residuals on corr etc , basically relative value metrics.

I'm attempting to simplify.
Probably wont do a great job as the topic of Quant analysis could be discussed for ever.

The numbers derived from say stochastic can be the input function or part a group of functions with variables that can be tested against a data set to determine any advantage in their repeated application over time to a similar data set.

I dont wish to narrow the answer to my suggestion but offer it up as a part of the whole.

Now I maybe wrong so I will consult my in house Rocket scientist.

 

[mazzatelli]

I'm attempting to simplify.
Probably wont do a great job as the topic of Quant analysis could be discussed for ever.

The numbers derived from say stochastic can be the input function or part a group of functions with variables that can be tested against a data set to determine any advantage in their repeated application over time to a similar data set.

I dont wish to narrow the answer to my suggestion but offer it up as a part of the whole.

Now I maybe wrong so I will consult my in house Rocket scientist.

Hi tech/a,
I'd say 85% of quant space assumes the underlying is random via stochastic processes e.g. brownian motion - this assumption only useful for instruments with optionality.

This leaves the remaining, of which I have seen times series analysis and forecasts used to model for example momentum e.g. correcting for volatility differences to filter out random noise that might look like momentum, rather than using a stochastic oscillator calculation.

Or e.g. converting prices to log returns and running F-test in an attempt to define trendiness rather than use a moving average.

There is some crossover e.g. using breadth statistics but I wouldn't say ta provides the majority of input for quant analysis.

I noticed you cited using simulation e.g. bootstrap or Monte Carlo to test if a systems returns are statistically significant, with a stochastic oscillator as the/a input. I agree it is a quantitative approach, but debate the perceived common use of traditional ta indicators.

Relevance to OP? goponcho can look into stat-arb [pairs trading] as suggested by skc or loose arbs [ - buy X at n and looking to sell Y at n+1], since this is classed under traditional quantitative trading.

All semantics, and by no means a dish on technical analysis.

Is your rocket scientist now working in finance?

 

[tech/a]

Thanks Mazza
Beyond basics is above my fighting weight.

Rocket scientist is developing a cancer identification instrument through protien identification using light refraction,
So I guess you couldnt get further from finance!