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

I was just referring to the argument that a certain group of inputs can be automated to make money consistently , that argument is flawed and counterproductive ....

If you mean fixed inputs then I might agree.
If you have adaptive inputs based on price, then I would say it is possible.
 
Common sense.. walking.. visual identification as if imitating the human iris - not one of these has anything to do with profiting from range-bound squiggly lines. They are GCSE level debating arguments of humans vs machines. Not one of them is relevant. In no way are we asserting machine superiority over humans in general, but if you think you can out-trade a decent institutional automated system using channels, S&R, legs, key areas, box breakouts, higher timeframe swing hi/lows, then you're, respectfully, near enough clueless of what goes on in your own supposed industry. As Liquid says, no HFT or long term algorithmic code has ever hit the internet in full (and it never will because many of the core libraries are heavily obfuscated - the UBS and Goldman thefts were only partial - only individual python and java classes will be known inside out by one individual).

Why is there no human equivalent of Kasparov at the top of a modern institution defeating Deep Blue over and over...?

No, your argument is specious. I am explicitly asserting that human capacity for pattern recognition and response, when trained correctly is far more effective at taking money from a market than the best algo. I am not just talking about FX here or HFT either.

As for your Kasparov vs Deep Blue argument, I thought you might be beyond a reductio ad absurdum type argument - worthy of a 6th form debating society but nonetheless, adolescent. I wonder how Deep Blue would have done against Bobby Fischer with no prep from IBM?

Algo trading is the answer to profitability and scaling by large participants. It is in itself not the pinnacle of performance; It merely raises the average.

I find what you say very interesting btw - it was in fact one of your posts that made me think and start this thread.
 
Last edited:
No, your argument is specious. I am explicitly asserting that human capacity for pattern recognition and response, when trained correctly is far more effective at taking money from a market than the best algo. I am not just talking about FX here or HFT either.

As for your Kasparov vs Deep Blue argument, I thought you might be beyond a reductio ad absurdum type argument - worthy of a 6th form debating society but nonetheless, adolescent. I wonder how Deep Blue would have done against Bobby Fischer with no prep from IBM?

Algo trading is the answer to profitability and scaling by large participants. It is in itself not the pinnacle of performance; It merely raises the average.

I find what you say very interesting btw - it was in fact one of your posts that made me think and start this thread.

Well thanks for that, but I also take mild umbrage to some of what you say.

You're asserting that this 'trained' pattern trading is superior without evidence or experience of algorithmic trading, thus it is just a guess on your part...? So why say it without proper proem saying that? And what do you mean by effective? Win rate? Anyone who risks their mortgage on a forward tested 'feel' win rate with anything less than 250 grand (most people here out) is borderline insane and belongs ranting with 15 min et al. It's just a massive punt based on reckless comments that markets are 'simplistic repeating patterns blah blah' - retail traders are so relentlessly stupid sometimes that it's genuinely hard to actually feel for them the more I hear from them.

How do you know algorithmic trading is not the pinnacle of performance exactly? I never stated it as such, but you stated the opposite, thus I assume you know this and how??

I already admitted my Deep Blue argument was silly and meant to undermine far more ridiculous and entirely uneducated (people who have never coded, people who have never worked in an IB, people who have never worked in a fund, people who don't even have a positive P/L even) comparisons being drawn on the other side, but it was kind of you to bring it up again despite my very obvious labelling of it as so to Peter.

Bah.
 
Random i wasn't referring to banks trading as a whole they make profits because they have many divisions and they are diversified , maybe they autotrade thousands of systems simultaneously , not to mention they are market makers as well so that is not the point they make money at least most of them although i wouldn't be surprised if all of them blow up at some point ! I was just referring to the argument that a certain group of inputs can be automated to make money consistently (directional trading ) , that argument is flawed and counterproductive ....

Didn't mean to get on my high horse tar, sorry for that, but I've been doing what you say isn't possible for years and I drawdown on 1 in 7 trading days, which is an incredible luxury. I've hit nothing like the win rate from algo trading with anything else and my point to point drawdown is the smallest I have ever ever experienced allowing me to scale up far more rapidly over the years than I could have otherwise. I don't know if our definition of directional is different - I take as many as 20 trades per instrument a day.

I have six ''optimisable' (not a word) inputs in total for about 90 million permutations - I optimise them in an iterative fashion 3 times (iterations done 3 times cover most probabilities, just like grossing up a single integer 3 times using a circular ref covers most of the end result) because my computers can't handle any more than that in a weekend as the data I use is bar magnified (i.e. reconstructed exactly as it occurred tick by tick to properly test stops - OHLC is quite flawed).

This is why I couldn't agree to what you were saying as I use it, the bank I left use it and many others do too.
 
Anyone who risks their mortgage on a forward tested 'feel' win rate with anything less than 250 grand (most people here out) is borderline insane and belongs ranting with 15 min et al. It's just a massive punt based on reckless comments that markets are 'simplistic repeating patterns blah blah' - retail traders are so relentlessly stupid sometimes that it's genuinely hard to actually feel for them the more I hear from them.

I would say anyone who risks a penny is insane (based on forward tested feel), granted we all do to some point, as we must test the waters. 250k seems a bit steep tho:-0

If we are to lose, may as well do it with as little as possible. Everyone needs the chance to learn, but must curb that enthusiasm first, hence the need to be burnt at the beginning, then the energy from the enthusiasm can be used to better effect.

After this point, if they continue in their reckless manner, then I totally agree with your last sentence(y)
 
Didn't mean to get on my high horse tar, sorry for that, but I've been doing what you say isn't possible for years and I drawdown on 1 in 7 trading days, which is an incredible luxury. I've hit nothing like the win rate from algo trading with anything else and my point to point drawdown is the smallest I have ever ever experienced allowing me to scale up far more rapidly over the years than I could have otherwise. I don't know if our definition of directional is different - I take as many as 20 trades per instrument a day.

I have six ''optimisable' (not a word) inputs in total for about 90 million permutations - I optimise them in an iterative fashion 3 times (iterations done 3 times cover most probabilities, just like grossing up a single integer 3 times using a circular ref covers most of the end result) because my computers can't handle any more than that in a weekend as the data I use is bar magnified (i.e. reconstructed exactly as it occurred tick by tick to properly test stops - OHLC is quite flawed).

This is why I couldn't agree to what you were saying as I use it, the bank I left use it and many others do too.

Well good luck , but for me it is a counterproductive argument to say there is a directional mechanical system that if you followed to the letter you will make money forever let alone automating it , if such a codeable system exist it will definitely be a holy grail and is worth billions but then again these developers and fx traders you've mentioned are not even self employed and still depend on their daily jobs , any system if it exist it will become obsolete in no time in these competitive markets that's why we've never seen or heard of such a system . I exclude market making , HFT front running , arb ... etc which is a totally different animal and it is not related to predicting the next move " directional " .
 
Well thanks for that, but I also take mild umbrage to some of what you say.

You're asserting that this 'trained' pattern trading is superior without evidence or experience of algorithmic trading, thus it is just a guess on your part...? So why say it without proper proem saying that? And what do you mean by effective? Win rate? Anyone who risks their mortgage on a forward tested 'feel' win rate with anything less than 250 grand (most people here out) is borderline insane and belongs ranting with 15 min et al. It's just a massive punt based on reckless comments that markets are 'simplistic repeating patterns blah blah' - retail traders are so relentlessly stupid sometimes that it's genuinely hard to actually feel for them the more I hear from them.

How do you know algorithmic trading is not the pinnacle of performance exactly? I never stated it as such, but you stated the opposite, thus I assume you know this and how??

I already admitted my Deep Blue argument was silly and meant to undermine far more ridiculous and entirely uneducated (people who have never coded, people who have never worked in an IB, people who have never worked in a fund, people who don't even have a positive P/L even) comparisons being drawn on the other side, but it was kind of you to bring it up again despite my very obvious labelling of it as so to Peter.

Bah.

I can't substantiate my claim Random - I would be lying if I said I could. I do know more than you give me credit for about the industry and about professional software development but I just don't really fancy doing my life story if you don't mind.

WRT measurement I was thinking of Sharpe's as a good measure over a long period. I would have thought that the long term performance of a good human trader would possess a Sharpe that continues to rise over the length of their career and doesn't actually top out - I'm not talking about guys that pack in after 5 years btw. I mean people that have been through a number of cycles. Conversely I would expect the Sharpe of an algo to remain relatively stable over the same period and not demonstrate the same rise as it's primary attribute is to be effective from the word go but not necessarily to exhibit any material improvement. Doing something like a 2-sided T-Test on their Sharpes over say 20 years based upon monthly returns or something like that would indicate either a +ve or -ve deviation of performance away from the algo.

As for your general opinion on retail traders and their oversimplification of the subject and attitude to risk, I do tend to agree with you.
 
Last edited:
BTW, the basis for my assertion on performance is that even a sophisticated algo does not have the capacity for adaptation and improvement that a person has and is comparatively limited in it's function. This is based upon the brains massively parallel processing capability which is (currently) far superior to that of any quadcore/hexacore CPU. Speed is not important for many trading styles, parallelism is.

The human brain is broadly comparable to about 100 dual cores operating in parallel but it's clock speed is about 400mhz. It has two processing areas, conscious (pre-frontal cortex) which can process data at about 40 bit/s and the subconscious (rest of brain) which handles about 10mbit/s of data without us knowing, mainly from our senses.

Unconscious competence as performed by any expert in their field, especially wrt cognitive behaviour (like trading or driving) is the process of moving processes from the pre-frontal cortex which is slow and laboured in it's processing into the area of the brain responsible for running routine behaviour and reaction. This is what I am referring to as trained - using internalised and learnt processes and constantly adapting and improving them.

Finally, the only figure I can give you for algo trading performance which I can substantiate but unfortunately it is a single data point is about 60% pa off a fund size of about $1bio which is as an oversimplification, trend trading. I do better than this, and most importantly the head of trading at this fund can outperform this as a human. It's just the case that to get people into the fund, investors want algo based trading. In short their fund would not have grown had they not developed algos. These investment decisions weren't made based upon returns - they were made based upon fashion. If fashion was involved, human folly isn't too far behind and all I can see is an industry that believe's it's own propaganda - the madness of crowds in action once again.

Ergo, if algo trading is prevalent because it is a good way to scale a trading operation and investors want it, it is easy to believe that this is the right and best way to trade. I think you carry this confirmation bias with you Random and although your arguments and clearly your knowledge is way beyond the retail trading world, I just think you have bought into the industry dogma. Anyone for a Tulip?
 
Last edited:
Random

Do you keep your systems up all the time or you do turn it off at times "market news or volatility changes... etc" ?
 
Finally, the only figure I can give you for algo trading performance which I can substantiate but unfortunately it is a single data point is about 60% pa off a fund size of about $1bio which is as an oversimplification, trend trading. I do better than this, and most importantly the head of trading at this fund can outperform this as a human. It's just the case that to get people into the fund, investors want algo based trading. In short their fund would not have grown had they not developed algos. These investment decisions weren't made based upon returns - they were made based upon fashion. If fashion was involved, human folly isn't too far behind and all I can see is an industry that believe's it's own propaganda - the madness of crowds in action once again.

Good post Rob, I pretty much agree on the human brain side of things,
all algo's have human intelligence as an input at some stage.

Its the assertion of algo's as a fashionable client draw that I have trouble completely agreeing with.
For me its more about order management - iceberg order being the most well
known example.
It may be a fashionable draw, I also think an algo is a key element in reducing
staff costs, earning the spread on the smaller iceberg constituents,
and increasing the efficiency of execution.
All that adds a decent amount to bottom line of any fund.

Although I certainly wouldn't doubt that there is an element of perceived
capability used as a marketing tool, which may or may not be warranted.
 
BTW, the basis for my assertion on performance is that even a sophisticated algo does not have the capacity for adaptation and improvement that a person has and is comparatively limited in it's function. This is based upon the brains massively parallel processing capability which is (currently) far superior to that of any quadcore/hexacore CPU. Speed is not important for many trading styles, parallelism is.

The human brain is broadly comparable to about 100 dual cores operating in parallel but it's clock speed is about 400mhz. It has two processing areas, conscious (pre-frontal cortex) which can process data at about 40 bit/s and the subconscious (rest of brain) which handles about 10mbit/s of data without us knowing, mainly from our senses.

Unconscious competence as performed by any expert in their field, especially wrt cognitive behaviour (like trading or driving) is the process of moving processes from the pre-frontal cortex which is slow and laboured in it's processing into the area of the brain responsible for running routine behaviour and reaction. This is what I am referring to as trained - using internalised and learnt processes and constantly adapting and improving them.

Finally, the only figure I can give you for algo trading performance which I can substantiate but unfortunately it is a single data point is about 60% pa off a fund size of about $1bio which is as an oversimplification, trend trading. I do better than this, and most importantly the head of trading at this fund can outperform this as a human. It's just the case that to get people into the fund, investors want algo based trading. In short their fund would not have grown had they not developed algos. These investment decisions weren't made based upon returns - they were made based upon fashion. If fashion was involved, human folly isn't too far behind and all I can see is an industry that believe's it's own propaganda - the madness of crowds in action once again.

Ergo, if algo trading is prevalent because it is a good way to scale a trading operation and investors want it, it is easy to believe that this is the right and best way to trade. I think you carry this confirmation bias with you Random and although your arguments and clearly your knowledge is way beyond the retail trading world, I just think you have bought into the industry dogma. Anyone for a Tulip?

Good post and you may be right about the fashion, and if so, then that's good news for us, however, I would like to challenge your statement about the power of the human brain. It's one thing saying the human brain is capable of this sort of level of calculation, and is equivalent to such and such processing power, and it is quite another to apply that level of computation to a task. A calculator with next to no processing can perform better calculations than almost all humans. My computer can remember large documents faultlessly, I can't remember my previous post before this one. Actually I can't even recall this one. So while we may have that level of processing potential, we're mostly unable to use it, as it's occupied in several other tasks.


In trading I believe a high level of processing can be seen over time by the human brain, whether it is being applied at the subconscious level, I don't know, I'm not an expert in how brains work, but certainly there are various 'a-ha' moments along the way, and they come from somewhere. These algos can't do that. They can't write a novel, they can't do maths proofs, creatively they're weak. No doubt about that. So then in a way this thread comes down to whether creativity is an advantage in live trading. Of course it is in trading overall. But the creativity in trading is outside of actual live trading (for me!). I come to the trade with my trading strategy, and I play that strategy all day. Any deviation from changing the rules mid trade etc has ended in poor results for me in the past. Which is not to say that you or anyone else can't do it. In fact I have followed the S&P thread and you seem to do that, but it's hard to know how much of what you do could be automated. Perhaps it's the next step for me in my learning, but i'm not there yet, and I like my fixed rules, it gives me a base to judge things from. The creativity can take place outside of actual trading (for me).
 
Last edited:
Its the assertion of algo's as a fashionable client draw that I have trouble completely agreeing with.

Well I can only speak from my direct experiences with my friend. He's actually one of the 3 partners for the fund. I know him (the trading partner) very, very well and the other partner (responsible for software development) pretty well. I've watched them build the fund from scratch.

They have a couple of client do's every year. I go to them, generally keep his wife company whilst he talks the talk with clients and other interested parties. If there is one constant over the 7 years when making chit-chat with his clients, it's about the brilliance they perceive of investing in a fund that is almost completely automated. The fact he has outperformed the algos and has a record from the last 20yrs that is not unusual but is certainly better performing than his algos over a longer period of time is of no relevance to his clients. His perceived value is in the area of trading strategy development, which is predominantly what he does when he isn't entertaining and expanding the business for the fund.

Good post and you may be right about the fashion, and if so, then that's good news for us, however, I would like to challenge your statement about the power of the human brain. It's one thing saying the human brain is capable of this sort of level of calculation, and is equivalent to such and such processing power, and it is quite another to apply that level of computation to a task. A calculator with next to no processing can perform better calculations than almost all humans. My computer can remember large documents faultlessly, I can't remember my previous post before this one. Actually I can't even recall this one. So while we may have that level of processing potential, we're mostly unable to use it, as it's occupied in several other tasks.

Wrong part of the brain Shakone - you are trying to use your pre-frontal cortex and it's 40bit/s processing capability. The part I am interested in and believe is most relevant once you get well into conscious competence is your subconcious processing; The part that takes over when you drive; The part that controls your walking; The part that is responsible for the chess move that jumps out at you because you have played 1000's of games; The part that plays that piano piece without music whilst you are looking away from the keyboard; The part that says go long now because this sell off has finished; The part that controls 95% of pretty much everything that you do as a human being and exerts so much control over your pre-frontal cortex and associated conciousness that it fools you into thinking you rationalised what you are doing.
 
I come to the trade with my trading strategy, and I play that strategy all day. Any deviation from changing the rules mid trade etc has ended in poor results for me in the past. Which is not to say that you or anyone else can't do it. In fact I have followed the S&P thread and you seem to do that, but it's hard to know how much of what you do could be automated. Perhaps it's the next step for me in my learning, but i'm not there yet, and I like my fixed rules, it gives me a base to judge things from. The creativity can take place outside of actual trading (for me).

Me too. I always plan the night before, adjust based upon what happens in the European session and then trade around my plan in session. What I do in execution of that plan may vary though. Say for example I go into the session with a bias long and take a pullback off a test of say, the overnight low. If it looks weak and can't make it to the other side of the range, I come out and my bias then switches to short as I know from experience that a re-test of the low set by the initial balance usually fails. I then look to see if there is a viable target wrt risk in doing this and if so, go short. If not, I stay out. Sometimes it works out, sometimes it doesn't. They are just rules/guidelines - it's not magic and I probably could get about 80% of it modelled if I put my mind to it. I only really trade variations of 3 strategies.
 
Well I can only speak from my direct experiences with my friend. He's actually one of the 3 partners for the fund. I know him (the trading partner) very, very well and the other partner (responsible for software development) pretty well. I've watched them build the fund from scratch.

They have a couple of client do's every year. I go to them, generally keep his wife company whilst he talks the talk with clients and other interested parties. If there is one constant over the 7 years when making chit-chat with his clients, it's about the brilliance they perceive of investing in a fund that is almost completely automated. The fact he has outperformed the algos and has a record from the last 20yrs that is not unusual but is certainly better performing than his algos over a longer period of time is of no relevance to his clients. His perceived value is in the area of trading strategy development, which is predominantly what he does when he isn't entertaining and expanding the business for the fund.

Although I still don't completely agree that is the only reason, I certainly think
there is an element of that present.
If that is your personal experience, I can't argue with that.
All I can offer as a counter argument is rentec's medallion fund:
Best Hedge Funds: Jim Simons Medallion Fund?s Returns and Alpha - Insider Monkey
That fund is only open to current and past employees and their families.
At least with that fund, there is no reason to impress anyone, as they are all
directly involved anyway.

I've always said this, and I stand by it now, I don't think automation is efficient
as good discretionary trading.
I also don't believe there is a best way, just the one that suits the individual.
As mentioned before, yes there is an undeniable client draw factor as well
as cost reduction etc.
Personally, I use it because I don't want to actively monitor myself,
as my typical trade duration is usually about as long as you can go without
holding overnight, unless its cut due to loss trigger.

BTW, good thread you started :)
 
trade what you see, or what you think, or what you hear, or what you smell, or what you ....... just as long as you trade your edge, you know its character anything else is gambling. I love to gamble but it doesn't pay the bills.
 
Well here's an example of direct competition:
http://www.trade2win.com/boards/futures-options/160450-algos-stirs.html
Although its fair to say most people aren't affected to that extent, or even at all.
I can see your point of view as you run scalp entries, so they probably help you.

I think there is still good structure to the market right now.

Well, OK - the ES has been a bit awful for the past 2 weeks but it's definitely showing signs of coming to life again with todays action and I think it will start giving position day traders decent entries again.

The bund is still fine in terms of structure and giving position day traders plenty to work off.

In terms of the really thick markets like the US Treasuries, I have heard that game has been tougher the past few weeks but I don't think that it's 'over' for them long-term by any means.

For those ultra thick instruments like STOXX, I have heard that this has pretty much died from a prop perspective but isn't it just like C was before the split and how BAC is now? BACs average volume is 122,203,560 shares per day and the ATR is 30c. When I think of algos killing something, that's what I think of.

I don't think of algos in terms of dicking around with outright directional positions but I could be wrong.

Volume in the ES right now is what 1.4M - 1.8M on an average day? SPDR is 50M with a 1.04 range. Seems OK to me.
 
Top