Manual or Automated Trading?

[mslk 10]

as i have been reading thru the numerous threads about people techniques (phantom, maelstrom, optiontimer, etc), a question always pops in my mind - "is any of this automated?"

 

if not, is it possible? or does successful trading require more than just programming models into a system? do you need to develop some gut-instinct as well?

 

- mslk

[Tams 20]

as i have been reading thru the numerous threads about people techniques (phantom, maelstrom, optiontimer, etc), a question always pops in my mind - "is any of this automated?"

 

if not, is it possible? or does successful trading require more than just programming models into a system? do you need to develop some gut-instinct as well?

 

- mslk

 

if you can articulate your logic in plain English,

you can program it.

[calsprdr 10]

if you can articulate your logic in plain English,

you can program it.

 

Sorry Tams, but as a professional programmer with 35 years experience, I can unequivocally say that your statement just is not true.

Of course many systems can be automated. But the problem lies with systems where the signals are not generated by "when X is 2 cents below Y." There are traders I have worked with who can tell me exactly what they are doing and I can program an approximation, but the backtesting blows up because I can *never* find a Z% away from W that adequately captures their "artistic" entries.

I like programming systems but I have never got one to work that adequately approximates how *I* trade. Now that's frustrating! :crap:

[zdo 20]

The mindset of being an automated systems developer is quite a bit different from the mindset of being a discretionary / rule based systems trader. The best system developers I know would consider manual trading to be a complete waste of time. They find a niche edge and become masters of testing and accurately reading performance reports and real time tweaking instead of masters live ‘screen time’ trading.

 

After a while at it, I finally realized I had to spend months as one, then months as the other to get desired results – but maybe that’s just me. After almost 25 years of trading full time, I still look at some methods and fantasize about automating it but realize how extremely arduous it would actually be to even try – all the while without any guarantees of acceptable results. Been there done that – more than once. Also have been blessed to develop some automation that is still standing the test of tme...

 

Also, finding a programmer to get into a shared vision with you / really understand what you are trying to accomplish, and make the wholehearted commitment required has odds about like playing the ‘big game’ lotteries. The huge majority of programmers, including most 'do it yourself'ers, can’t even begin to think and code into the uncertainty and fuzziness of varying multiple weights / influence of parameters that is required.

 

It takes a great deal of work to fully automate most methods – largely because it’s very arduous to correctly program all the parameter fuzzinesses that individual's grayware handles quite a bit more routinely. Basically, the practiced eye can discern and unconsciously use information, but not be able to quantify it in code well enough where it works the same way automated as it does manually for close to the same percentage of signals. Generally most beginners find that code can generally handle setups ok, but anything but the simplest of context conditions can quickly become a very 'disencouraging' activity…

 

In the ‘distribution’ of methods, the great middle of systems ie analysis methods most traders use are the ones most difficult to automate. Very generally speaking, the systems that can be ‘easy’ to automate are in the tails. Both are ‘simple’ types of systems. They are either in or near the lower win rate tail (think long term trend, etc.) or they are in or near the high win rate tail (think HFT, etc.). The real challenge programmatically for both of these types of ‘easy’ automation is proper risk management. Without automated sizing too, the whole development effort can be a waste, etc etc. … am not jaded and I still love automation. Just trying to be real here

:missy:

hth

Edited July 28, 2011 by zdo

[Tams 20]

Sorry Tams, but as a professional programmer with 35 years experience, I can unequivocally say that your statement just is not true.

Of course many systems can be automated. But the problem lies with systems where the signals are not generated by "when X is 2 cents below Y." There are traders I have worked with who can tell me exactly what they are doing and I can program an approximation, but the backtesting blows up because I can *never* find a Z% away from W that adequately captures their "artistic" entries.

I like programming systems but I have never got one to work that adequately approximates how *I* trade. Now that's frustrating! :crap:

 

then you are talking non-logic... it cannot be coded.

[Tams 20]

the first step is getting your thoughts in clarity

 

human mind works in fuzzy logic

 

computers/languages we have access to works in sequential logic

 

before you can code anything,

you have to reduce your fuzzy feelings into cold hard facts.

[phantom 16]

Well said, zdo.

 

My experience working with my brother, who is a high level programmer, attests to your insights. We've tried, but we just couldn't do it.

 

I'm not saying that it is not possible.

 

I am saying that our results were not reliable.

 

Huge difference.

 

 

Phantom

[phantom 16]

if you can articulate your logic in plain English,

you can program it.

 

I think the issue here is: Can one articulate one's logic in plain English?

 

Talk is cheap, Tams.

 

How about you posting one of your trades where you coded the entry and exit and had the computer bring you a REAL profit?

 

Otherwise, we can assume you're only playing the Devil's Advocate...

 

 

Phantom

 

P.S. and if you reply with something lame, like, "you can assume whatever you want..." We'll have to "consider the source" from here on out.

[zdo 20]

Tams is absolutely correct about “if … logic in plain English, then…” – Theoretically.

 

But practically ?? A quick, incomplete example - It is ‘logical’ for 1) a certain condition to fall within a certain range to be ‘true’, 2) for that ‘trueness’ range to vary in size and location (still under logical but suddenly overly complicated conditions) … practically, the 'coding' practicality falls off rapidly and precipitously...

[zdo 20]

Taking the diff’s btwn manual and automated several layers further - In much manual trading, even cognitive ‘work’ with “all the parameter fuzzinesses that our grayware handles…” isn’t logical.

 

‘Worse’ yet for attempts at coding, beyond that we also have the issue of subjective probability accessed not from the cognitive but from adaptive utilization of the ‘soft’ data of emotions. You can actually train yourself to consciously benefit from awareness of common cause and special cause and ambiguity aversion,etc. Knightian uncertainty - Wikipedia, the free encyclopedia etc etc

Let’s see you code that sht, bit! :rofl: ( …jerk gotta scoot so finished this up tersely instead of nicely )

[phantom 16]

Taking the diff’s btwn manual and automated several layers further - In much manual trading, even cognitive ‘work’ with “all the parameter fuzzinesses that our grayware handles…” isn’t logical.

 

‘Worse’ yet for attempts at coding, beyond that we also have the issue of subjective probability accessed not from the cognitive but from adaptive utilization of the ‘soft’ data of emotions. You can actually train yourself to consciously benefit from awareness of common cause and special cause and ambiguity aversion,etc. Knightian uncertainty - Wikipedia, the free encyclopedia etc etc

Let’s see you code that sht, bit! :rofl: ( …jerk gotta scoot so finished this up tersely instead of nicely )

 

WOW. Now THAT was a mouthful!

 

Think I'll stick to manually trading breakouts of ranges with evidence of price rejection...

 

I gotta go wipe my forehead off...

[mslk 10]

then you are talking non-logic... it cannot be coded.

 

tams, is your trading 100% automatic? i am also a programmer by trade and thus automating is always my natural instinct. on one hand, i see your point - if your rules are logical and clear, in theory it should be programmable. in practice, i see some issues with sequential languages and maybe a rules-based language might be better. but back to the topic, how much of your trading is automated?

 

I think the issue here is: Can one articulate one's logic in plain English?

 

you hit the nail on the head phantom!

 

Talk is cheap, Tams.

How about you posting one of your trades where you coded the entry and exit and had the computer bring you a REAL profit?

 

lets not get into a 'prove' yourself flamewar .. tams i'll take your word on whatever your response is.

 

any other traders have any input? timracette, thenegotiator, mms, steve46, thalestrader, urmablume?

 

thanks

-mslk

[Tams 20]

tams, is your trading 100% automatic? i am also a programmer by trade and thus automating is always my natural instinct. on one hand, i see your point - if your rules are logical and clear, in theory it should be programmable. in practice, i see some issues with sequential languages and maybe a rules-based language might be better. but back to the topic, how much of your trading is automated?

 

...thanks

-mslk

 

my trading is not automated.

 

I have audio and visual alert on key events,

but the signals are not automated.

[phantom 16]

 

lets not get into a 'prove' yourself flamewar

 

My apologies.

 

 

Phantom

[xioxxio 10]

I am a trader with an IB, I have two developers working with me developing "strategies" to assist my trading. At the end of the day human instinct/gut felling is where the real money is.

 

The markets are driven by emotions (fear/greed), no program can cover that. It is easy to fit a strategy to historical data. However, humans do irrational things at expected and unexpected times which could never be accounted for.

[mslk 10]

My apologies.

Phantom

 

no worries, i've just seen other threads where this stuff kills the thread (e.g. most threads where steve46 is involved!)

 

anyhow, my motivation for this question is really around picking the right broker\platform. obviously, if success can be achieved via programming then i need to be more selective and my options are more limited. i've been playing with ninja but also looking at tradestation as its broker integration is nice. for any of the programmers out there, what are you using?

 

thanks

- mslk

[mslk 10]

However, humans do irrational things at expected and unexpected times which could never be accounted for.

 

are you referring to other traders or yourself? if its other traders, is the idea for you to spot these and take advantage of?

 

if its yourself, then isn't the point of automation to try and eliminate these (assuming its programmed correctly with the right fail-safes). thanks for the input.

 

anyone know what the big institutions\hedge-funds do?

 

thanks

- mslk

[Tams 20]

there are 3 types of autotrading

 

1. trend following

 

2. scalping, also known as buy low sell high. most of HFT are in this category.

 

3. spread/differential trading

[Tams 20]

there are 3 types of manual trading

 

1. trend following

 

2. scalping, also known as buy low sell high. most of HFT are in this category.

 

3. spread/differential trading

[calsprdr 10]

i've been playing with ninja but also looking at tradestation as its broker integration is nice. for any of the programmers out there, what are you using?

 

thanks

- mslk

 

I use IB and have done a little programming against its java API...not perfect, but could be worse. Let me warn you, don't use Tradestation Futures edition. It plain sucks. The regular Tradestation 8 for equities is OK. By the way, I use ninja for charting, it gets its data feed from IB with no problem except IB doesn't send tick data

[mslk 10]

there are 3 types of manual trading

 

1. trend following

 

2. scalping, also known as buy low sell high. most of HFT are in this category.

 

3. spread/differential trading

 

tams - i don't know if your being sarcastic or not. instead of speaking in code, what do you mean by these 2 posts?

 

thanks

- mslk

[Tams 20]

tams - i don't know if your being sarcastic or not. instead of speaking in code, what do you mean by these 2 posts?

 

thanks

- mslk

 

autotrading is just a reflection of manual trading.

[xioxxio 10]

are you referring to other traders or yourself? if its other traders, is the idea for you to spot these and take advantage of?

 

if its yourself, then isn't the point of automation to try and eliminate these (assuming its programmed correctly with the right fail-safes). thanks for the input.

 

anyone know what the big institutions\hedge-funds do?

 

thanks

- mslk

 

The hedge funds that trade through us use multiple strategies. Unfortunately we may only be seeing one side of the trade. Hedge funds tend to use multiple prime brokers to execute through.

 

Even your Quant funds use research reports, i.e. Fundamental analysis

 

What I was referring to if you could predict human behaviour making money in the markets would be easy.

[Tradewinds 12]

n!

permutations

factorial

 

Factorial - Wikipedia, the free encyclopedia

 

Let's assume you have 2 Moving averages on one chart, and 2 Moving averages on another time frame chart. Then you have identified 3 basic behaviors of each. Then you look at support and resistance, and have identified 3 basic behaviors of each. Then you have one lower study, and are looking for 4 basic behaviors from the lower study. So you have 12 conditions that you are trying to monitor.

 

• 3 conditions from Chart One with two MA's
  
• 3 possible conditions from Chart 2 with 2 MA's
  
• Support and resistance with 3 possible conditions you are filtering for
  
• 1 Lower study with four specific conditions you are looking for. E.g. Over 40, Under -40, Just went under zero, Just went over Zero. There could be many more.
  

 

There are 479,001,600 possible combinations of those 12 conditions. So you need to find a way to either exclude a lot of those possible conditions, or combinations of conditions, or you need to be able to evaluate over 479 million different possibilities.

 

So if you want to rely on pure computing power, and a complete analysis of every possible combination, it would seem that your system would need to somehow filter through hundreds of millions of possible conditions. And you would need to define what each of the millions of possible combinations means.

 

Or you need some way of figuring out why 99.999999% of those 479 million possible combinations are irrelevant, and define which combinations are the only relevant ones.

 

When you realize how many possible combinations of the conditions there can be, it's understandable how easy it is to miss what is really going on in the market. No one can possibly monitor hundreds of millions of possible combinations in real time.

 

Excellent trading knowledge might be able to easily discount the vast majority of those combinations as meaningless to a trading decision, but still, even if you are left with hundreds of possible combinations of your trading inputs, that's still not easy to quickly evaluate in real time.

 

We don't consciously understand how much data the brain is processing in real time. It's a lot.

[daVinciLite 10]

in practice, i see some issues with sequential languages and maybe a rules-based language might be better

 

-mslk

 

I think a combination of an artificial intelligence language like Prolog and a sequential language like C++/C# would provide a framework for a workable 'expert system'. This would be a long-term project of at least a few years.

 

From my experience a good programmer makes a terrible discretionary trader and a good discretionary trader makes a sub par programmer. In order to create an 'expert system' you also have to be an expert to begin with. :-) I wonder how many good traders know enough programming to put their subconscious rules into a set of programmable rules?

 

dVL