Trade what you see, not what you think....

The 'correct' path is the one that personally suits you.

Lack of awareness does not mean impossible.
Generally though, I will agree that the higher end complex stuff is the domain
of the PhD's and server farms.

Liquid - very nice post. You addressed a lot of the incorrect assumptions about automated trading. I find it a struggle to understand why people think automated systems are so limited in scope and application, yet are happy to play FIFA with their underpowered Xbox and have a good challenge there and think nothing of the technology behind it...

I can offer the following observations about forex for anyone who cares to read them (which is generally the instrument of choice for T2W people I think and from where I have real world experience):

- that humans mainly trade exotics these days - these were becoming more and more profitable when I left Mitsubishi UFJ and very much favoured human style accumulation trading and dare I say it, a bit of manipulation (moving, busting and reversing, teasing, nothing illegal). Though I understand exotics have had a lot of work put into them by the development teams since then, but I don't know what the split is nowadays.

- majors and minors are very heavily traded by computerised systems. Many smaller 'institutions' (like Knight Capital did...) put almost their entire credit line into automated trading. The bigger players like JPM obviously don't do that, but it makes up an awful lot of their day to day trading and as I have said before, most human traders have been let go from FX desks, I think even JPM barely has 1,200 now. The manual teams could only match automated returns using exotics, and that was a return based on ratios, not actual returns, due to the liquidity being crap in exotics. News trading and accumulation biases may still be managed by humans, but executed by the same systems.

- that optimisation to a relevant sample is all important in automated trading. Since the market 'shape' ultimately phases based on the number of market moving speculators currently engaged in it rather than due to anything mystical, you can adapt your system as the market changes on the fly, usually at a ratio of the system up time vs sample time at a factor rate determined optimal by the bank per instrument. A lot of people backtest systems and find something that works well for 2012 and are terribly disappointed to see it failing in 2011 and don't trade it for fear 2013 will be a new 2011. Well, a bit more thought and work might let them open their minds to the reasons for market changes and how they might actually employ the system. Though the market can change suddenly, it NORMALLY doesn't. And this is hugely profitable for well optimised systems. (The BOJ was constantly intervening 2 years ago at exactly midnight and 4am GMT with the Yen. They were giving money away to automated traders by doing this for more than a few weeks and then they stopped being so predictable. Simple adaptation required, minimal drawdown and maximum run up.)

- People also hugely underestimate what 'can be programmed' because most have been exposed, at most, to 100 line EAs or lack any sort of programming experience outright. You can program very many advanced conditions - look at the games and programs available to us in the modern world as even a basic example. When you write a program from scratch, you try and put your thoughts and opinions about how your brain solves a certain issue or situation into high level code and honestly, there is little that can't be coded in trading, except of course the ability to have random ideas pop into your head, which has limited application with trading!

My own optimised systems are as lean as I can make them... which is nothing compared to the institutions because my computers simply cannot run a bar magnified optimisation across a trillion permutations every Saturday. I do manually trade as well as use auto trading, but (crude example coming) I know that the former is like doing accountancy with an abacus and a journal, while the firms have Excel. I can get it right using my abacus, but I'm pretty slow and ill suited to competing with the bigger players and their spreadsheets... I just have to find a niche where my crap old method still works and accept what I am and where my competition are in relation to me. But I really shouldn't claim that my abacus is better when I've never ever used or even seen Excel at work.... :whistling
 
In answer to the initial question, I don't think a human at his peak can beat the machines in a fixed system with fixed rules. The evidence for that in trading is the rise of the machines and their dominance. The reason why, is not because a human is any less than a machine at his human peak, but simply because a human at his peak is less than a human at his peak plus machine speed capabilities. After all, these HFT systems and algos are not (in the main) computers learning for themselves. I realise there are some who invest in neural nets, to try to learn, but this is not what the dominant algos generally are (imo). There are experienced and successful traders (most likely much more than you or I) who have had input into these algos, and they have been tested to death. And if it's not an experienced trader inputting, it's most likely a team of researchers who are at their peak in maths/stats/physics etc. IF it were just man against machine, then that would be far easier for us.

I wouldn't despair just yet though, for a few reasons. First you don't need to be better than all of them to make money. They are fighting with eachother, that's where the big money is, not us, we're often collateral damage, and there will always be opportunity regardless, in the case of forex, business and exchange needs to be done, government policy will occur, and it doesn't have to care about every single pip.

And secondly, the system is not fixed, everything changes. This is a universal law. As Liquid Validity mentioned, what would the effect of a transaction tax be? What if there were a miniscule tax on placing limit orders? These could probably decimate a large part of the HFT industry. I don't expect that to happen any time soon as the industry spends quite a lot of time convincing everyone that they are beneficial and didn't cause any flash crashes etc.

Finally the markets are still emotionally driven and based on feedback loops. I realise this is hard to quantify, so you are quite welcome to say I'm writing BS. I can't give any convincing evidence. But even looking at Liquid Validity's post using the nanex data to show liquidity drying up before a news annoucnement, the question is why? Why would all these algos withdraw liquidity before news? Fear?
 
I find it a struggle to understand why people think automated systems are so limited in scope and application, yet are happy to play FIFA with their underpowered Xbox and have a good challenge there and think nothing of the technology behind it...

People also hugely underestimate what 'can be programmed' because most have been exposed, at most, to 100 line EAs or lack any sort of programming experience outright. You can program very many advanced conditions - look at the games and programs available to us in the modern world as even a basic example. When you write a program from scratch, you try and put your thoughts and opinions about how your brain solves a certain issue or situation into high level code and honestly, there is little that can't be coded in trading, except of course the ability to have random ideas pop into your head, which has limited application with trading!

I agree about the source of most commonly held views,
and to a certain extent, I agree with them considering the typical source.
We are all lumped in with the clueless EA sharks and zulu high strike rate blow up
merchants.

In reality, the overwhelming majority of public domain programming is
either deliberate or unintentional garbage.
Anything with any value in the retail / insti domain is kept under wraps for obvious reasons.
Based on the easily available information and evidence, the typical view
of automation is to be expected, justifiably so...

In a funny kind of way, I actually think its a good thing that these misconceptions exist.
If it deters people form getting involved with programming who have done little
or no research, therefore under estimating the time and effort involved,
that is maybe a good thing.
I'm certainly wary of encouraging anyone to do this for those reasons alone.

Starting off as a discretionary trader and building up market experience,
and more importantly understanding the basic trading principles such as
risk, drawdown, sizing, probability and so on, is a safer bet for sure.
Without that foundation, tackling automation would be a nightmare.
Although on the other hand, a lot of that can be learned from automation as well.
Although its more likely that someone without that foundation will tend
to curve fit using nonsensical parameters.

that optimisation to a relevant sample is all important in automated trading. Since the market 'shape' ultimately phases based on the number of market moving speculators currently engaged in it rather than due to anything mystical, you can adapt your system as the market changes on the fly, usually at a ratio of the system up time vs sample time at a factor rate determined optimal by the bank per instrument. A lot of people backtest systems and find something that works well for 2012 and are terribly disappointed to see it failing in 2011 and don't trade it for fear 2013 will be a new 2011. Well, a bit more thought and work might let them open their minds to the reasons for market changes and how they might actually employ the system. Though the market can change suddenly, it NORMALLY doesn't. And this is hugely profitable for well optimised systems. (The BOJ was constantly intervening 2 years ago at exactly midnight and 4am GMT with the Yen. They were giving money away to automated traders by doing this for more than a few weeks and then they stopped being so predictable. Simple adaptation required, minimal drawdown and maximum run up.)
Very good points, completely agree.
Agree on your optimisation points, that's an area that can cause all sorts
of problems as its usually misunderstood and abused.
Usually its a case of arbitrary parameters that give the best results - aka curve fit.
Any optimisation, as you intimate, should make sense and target specific types
of behaviour.

I think that's one of the root causes of the general misconceptions that arise.
Plenty of software is capable of automated optimisation, so people
blindly trust and use it without a thought given about the logic behind the
resulting code.

My own optimised systems are as lean as I can make them... which is nothing compared to the institutions because my computers simply cannot run a bar magnified optimisation across a trillion permutations every Saturday. I do manually trade as well as use auto trading, but (crude example coming) I know that the former is like doing accountancy with an abacus and a journal, while the firms have Excel. I can get it right using my abacus, but I'm pretty slow and ill suited to competing with the bigger players and their spreadsheets... I just have to find a niche where my crap old method still works and accept what I am and where my competition are in relation to me. But I really shouldn't claim that my abacus is better when I've never ever used or even seen Excel at work.... :whistling
Same here, agree due to hardware and resource limitations.
Annoyingly enough,my main limitation is still programming, but its a constant
learning process, which I accept.
In the meantime I work around the issue as best as I can.
 
Machines/Computers currently outperform the human brain only when it comes to processing arithmetic, but they don’t even come close to the processing power humans take for granted. Michio Kaku, an American theoretical physicist said that if you combined all the computing power on earth today you would barely have the intelligence of a lobotomised cockroach. I 100% agree with him.

Just look at how much is being invested in voice and face recognition software and they still don’t work as well as a 5 year old or even a 3 year old child’s brain does. A child immediately recognises its own mother’s face or voice among a crowd of people even with all the other distractions.

Robert Zubrin, an American aerospace engineer, is keen to have a manned mission to Mars whereas NASA favours Rover exploration. He posed this simple challenge to NASA: Let’s have an Easter egg hunt where NASA can have 1000 Rovers Vs one 10 year old child and we’ll see who ends up with the most eggs!
 
By the way, successful trading isn't about outperforming anyone/anything. It's about recognising what is being performed and capitalising from it. Buying and selling whether performed by humans or machines all looks the same on the tape.
 
I wouldn't despair just yet though, for a few reasons. First you don't need to be better than all of them to make money. They are fighting with eachother, that's where the big money is, not us, we're often collateral damage, and there will always be opportunity regardless, in the case of forex, business and exchange needs to be done, government policy will occur, and it doesn't have to care about every single pip.
Good summary, as long as price move, opportunity exists, no matter how you trade.

And secondly, the system is not fixed, everything changes. This is a universal law. As Liquid Validity mentioned, what would the effect of a transaction tax be? What if there were a miniscule tax on placing limit orders? These could probably decimate a large part of the HFT industry. I don't expect that to happen any time soon as the industry spends quite a lot of time convincing everyone that they are beneficial and didn't cause any flash crashes etc.
If the SEC wanted to curb HFT action, a tax on limit orders would be less damaging.
A lot of HFT activity revolves around quote stuffing:
Nanex ~ 14-Dec-2012 ~ Quote Stuffing Bombshell
Nanex ~ 01-Nov-2012 ~ The Denial of Service Algo
Nanex ~ 15-Nov-2012 ~ Dangerous Order Types

I certainly wouldn't rule out SEC action at least, as they have been sabre rattling
for a while:
Nanex ~ 16-Sep-2012 ~ SEC Speaks "Nanex"
Nanex ~ 14-Sep-2012 ~ Watershed Event! SEC Fines NYSE

Finally the markets are still emotionally driven and based on feedback loops. I realise this is hard to quantify, so you are quite welcome to say I'm writing BS. I can't give any convincing evidence. But even looking at Liquid Validity's post using the nanex data to show liquidity drying up before a news annoucnement, the question is why? Why would all these algos withdraw liquidity before news? Fear?
I can understand your point in that algos are still programmed by humans.
Although they execute with no emotion and target other algos who operate the same way.
Yet despite that, there is still an element of emotion in the thought processes
involved in creating the code - human ideas.

As for the example of liquidity drying up, what that shows is they have advanced
to the level similar to a flock of birds feeding on the ground.
In normal conditions its business as usual.
If one signals a warning, they all react, taking flight.
So I suppose you could say its fear.

The main point is how easily a stampede can occur,
potentially leading to another flash crash.
 
I still can't figure out why people get so worked up about HFT...as far as I'm concerned it is pure media propaganda BS!
 
Machines/Computers currently outperform the human brain only when it comes to processing arithmetic, but they don’t even come close to the processing power humans take for granted. Michio Kaku, an American theoretical physicist said that if you combined all the computing power on earth today you would barely have the intelligence of a lobotomised cockroach. I 100% agree with him.

Just look at how much is being invested in voice and face recognition software and they still don’t work as well as a 5 year old or even a 3 year old child’s brain does. A child immediately recognises its own mother’s face or voice among a crowd of people even with all the other distractions.

Robert Zubrin, an American aerospace engineer, is keen to have a manned mission to Mars whereas NASA favours Rover exploration. He posed this simple challenge to NASA: Let’s have an Easter egg hunt where NASA can have 1000 Rovers Vs one 10 year old child and we’ll see who ends up with the most eggs!

Yes there is no argument based on adaptive and subjective intelligence.
Its the weaknesses that adaptive intelligence creates that
have caused markets to be run by computers.

No one is saying humans are redundant or its impossible for humans
to trade discretionary.

At the HFT end where the money is due to various currently legal loopholes,
computers win hands down simply due to speed of decision making and execution.
That does alter the characteristics of markets controlled by computers.
As I said earlier, as long as price moves, opportunity exists, no matter how you trade.
 
Its the weaknesses that adaptive intelligence creates that
have caused markets to be run by computers.

Really? Where is the proof? There is a group of conspiracy theorists who say the world is being run by The Rothschilds or the Illuminati...Personally I think the markets are being run by the aliens from Omicron Persei 8.
 
Really? Where is the proof? There is a group of conspiracy theorists who say the world is being run by The Rothschilds or the Illuminati...Personally I think the markets are being run by the aliens from Omicron Persei 8.

Proof of what?
Markets are run by computers, I don't see how you can argue against that fact.
Look at some of the links I posted if you want proof.

Whether its due to the reasons I gave, or simply cost cutting,
computers won.
Right or wrong, that is an irrefutable fact at this point in time.
 
Proof of what?
Markets are run by computers, I don't see how you can argue against that fact.
Look at some of the links I posted if you want proof.

Whether its due to the reasons I gave, or simply cost cutting,
computers won.
Right or wrong, that is an irrefutable fact at this point in time.

What? Irrefutable facts are you serious? The article knows all the participants in a market? Who is doing the bulk of the buying and the selling and from where? C'mon! Take a look at GATA's claim about manipulation of the gold and silver market. There is never 100% proof of anything, only inferences and incorrect, illogical or irrational ones at best.
 
What? Irrefutable facts are you serious? The article knows all the participants in a market? Who is doing the bulk of the buying and the selling and from where? C'mon! Take a look at GATA's claim about manipulation of the gold and silver market. There is never 100% proof of anything, only inferences and incorrect, illogical or irrational ones at best.

Evidence of computers being responsible for most of the volume:
http://www.trade2win.com/boards/dis...you-see-not-what-you-think-2.html#post2071196
Or maybe all the human traders decide to jump at the same time :rolleyes:

Evidence of human weakness as a collective whole, sample of prop traders:
http://www.trade2win.com/boards/first-steps/166016-95-lose-6.html#post2068914
Watch floored again as well.
Not hard empirical proof, but where is any proof to back up what you say?

What exactly what are you saying, just to be clear?
It looks to me like you are suggesting that computers are not
responsible for most trading volume.

You also appear to be suggesting that all human traders as a whole
are equally efficient with no weaknesses.
Off course there are exceptions.
Are you saying that?
 
The real question here is there a profitable system that can be automated ? speaking directional trading , there isn't period ! There is no such thing as a group of inputs that will make you consistently profitable , if there is then these inputs will become obsolete immediately cuz everyone will jump in , so this argument is counterproductive , it is a very competitive environment . Ofcourse that rules out "non-directional" algos : MM algos , HFT front runners , arb algos .. etc why ? because here it is not about predicting the next move but rather it is about technology and risk management .
 
Machine wins chess. Machine flies planes, drives trains. Machine executes 1000's of trades per second. Humans need to know their limitations and don't let their ego take over.

oh dear ....we are all confusing a machines capability to add up numbers with being able to apply the common sence we as humans accrue over the years.....;)

N
 
The real question here is there a profitable system that can be automated ? speaking directional trading , there isn't period ! There is no such thing as a group of inputs that will make you consistently profitable , if there is then these inputs will become obsolete immediately cuz everyone will jump in , so this argument is counterproductive , it is a very competitive environment . Ofcourse that rules out "non-directional" algos : MM algos , HFT front runners , arb algos .. etc why ? because here it is not about predicting the next move but rather it is about technology and risk management .

Based on what evidence T........ ?

N
 
Based on what evidence T........ ?

N

Self defeating process : " if there is then these inputs will become obsolete immediately cuz everyone will jump in , so this argument is counterproductive"
 
If I were trading in the market side by side with a machine (NOT trading against it, trading the market) there’s no doubt the machine would produce greater returns than me every time. However, if I were trading against a machine it would have an advantage at first. Eventually, given enough time and trades, I could decipher what it’s doing. I could gain the upper hand because I can think, react, and adapt where the machine can’t. That’s what is meant by gaining experience.

The basis of AI is getting the machine to think and reason so humans could not gain that advantage. Maybe 100 years from now?

Peter

kinda agree .............

I think working with machines is the best of both worlds.....they feed the decent signals and we make the calls......

Humans have the capability for exceptional performance as well as equally Jaw dropping lapses in focus and cognicence

N
 
Self defeating process : " if there is then these inputs will become obsolete immediately cuz everyone will jump in , so this argument is counterproductive"

so no-one makes money trading then ?

N
 
Top