The Impact of Automated Trading

[oiltanker]

...I'm seeing two major types of 'bot. Momentum/break-out, and trend-reversal/market maker....

rnicoll

yes. algos like momentum especially breakouts on any time frame. indeed when you look prices is nearly all breakout of a range.

from the mathematical finances stuff i've seen forms of stochastics seems the idea and brownian motion and particle theory. trying to pick tops and bottoms must be a nightmare? Volume would seem key?

do you start from random walk? or trade the chartists?

 

[rnicoll]

i've been researching mathematical finance. there are courses in it at uni's. Basically mathematical finance regard most technical analysis as little better than 'a belief in santa' because they say price has no memory whereas technical analysis is nothing but price memory.

Thing is, why do they say price has no memory? Is there this opinion that traders aren't aware of what prices were yesterday? The number that look at the DJ at 8000-ish and think "Bloody hell, that was over 11,000 only a couple of months ago" is probably only a few, but you can bet most are aware it's dropped from the 9,000-ish it was earlier in the week.

The actual value of something has no memory, but this is poorly correlated to the price in my opinion. Does that make sense?

There are even courses called 'using black scholes to take on the chartists'. Presumably they think its easier to predict chartists and so make money from them than it is from the market?

Doing pure technical analysis by hand, without looking at fundamentals will indeed probably get your account eaten by the market. However, doing fundamental analysis and then looking at the technicals both to confirm, and provide entry/exit points, I think makes sense.

from lectures i ve read most reckon automated is now about 60% and say it will increase. the hot topics are neural networks and genetic programming. they say its hard to work out what is going on because anyone employed to work on it is signed to silence. (which sorts of blows the water out of the people selling their system- any real system would be a goldmine. who would tell anyone where the gold mine was unless there was no gold in it? )

Erm. Someone who sells their trade list based off the trade confirmation step of their system? I was trying this with Collective2; when I executed a trade, once it was confirmed it would be echoed to Collective2.

Also, if they're saying neural nets work, then that rather requires price memory. Sure, it's a lot more complicated than the technical analysis you'll read about in a book, but the idea of pulling market hints from the price is the same.

automation makes the market more efficient and will increasingly do so making it harder for non automation to compete. Sort of like tesco building a hyper store next to the corner shop.

however as long as markets are inefficient it is possible for individuals to make money and one would think it would never be 100% efficient. nothing is. And there will always be the golminers looking for the motherload....we've all been there.

Absolutely, and it's that efficiency I've been looking at primarily. Trying to find cases where currencies are overbought/oversold, and doing a trade the other way to make money from it. Right now, it doesn't work too well; the backtesting was too optimistic about actual prices to execute at, and then the market exploded. I haven't given up yet, though, either.

 

[rnicoll]

...I'm seeing two major types of 'bot. Momentum/break-out, and trend-reversal/market maker....

rnicoll

yes. algos like momentum especially breakouts on any time frame. indeed when you look prices is nearly all breakout of a range.

from the mathematical finances stuff i've seen forms of stochastics seems the idea and brownian motion and particle theory. trying to pick tops and bottoms must be a nightmare? Volume would seem key?

do you start from random walk? or trade the chartists?

My stuff is actually much more brutally simple, and trades based primarily on volatility. So, it tries to guage how much the price has moved relative to the value, and looks to trend reversals if the price moves faster than the value is probably moving. It doesn't try to get exact top/bottom (although I need to have it get closer), and instead just figures that if it's close it's good enough.

Another major advantage to an automated trader is that if it's right 51% of the time, that may be a tiny profit per trade, but it can trade relentlessy day in, day out, and without requiring a human. That's a good profit for relatively little ongoing effort.

 

[oiltanker]

the impression i got was as a rule they see price as random so anything that suggests otherwise means its probably something trading the chartists [who they see as more predictable?- you should read the hoots of laughter they have for things like elliot wave etc] rather than the price? there are some good course outlines for mathematical finance at oxford uni that give clues to what they might be on about. Obviously its a secret what they are really working on. There are some interesting 2008 lectures online at Gresham college on the history of mathematical finance that deals with what they are doing at the moment . part 3 2nd lecture if i remember.

One would think if they had really cracked it mathematicians would be billionaires? They do say this is a new subject maybe 50 years old so there is lots of room to improve as they look for 'laws'.

the indicator that, according to some, seems to have the greatest predictive value is the direction of net volume. Sort of like if the massive volume is equally balanced then all you need is a finger push to give it momentum in a direction assuming a frictionless environment. i suppose if one had a brokerage then one could see the data of buying and selling then develop models from that and use those to predict? But who has that advantage?

still its interesting stuff.It doesn't seem to matter if they are right or wrong. Its the fact they will increasingly dominate the market with those viewpoints which is the relevant thing?

 

[oiltanker]

...trade relentlessy day in, day out,...

yes machines can wear the human out. which is their advantage.

apparently the algos in the 1987 crash didn't include volatility as a consideration. they always assumed there would be liquidity [volatility is a function of liquidity] and so when there wasn't and prices gapped it freaked out the algos.

 

[rnicoll]

the impression i got was as a rule they see price as random so anything that suggests otherwise means its probably something trading the chartists [who they see as more predictable?- you should read the hoots of laughter they have for things like elliot wave etc] rather than the price?

I believe the chartists would argue that everyone's predictable, they're just aware of it. I think pure fundamental analysts are incredibly rare, and even if most traders don't sit there drawing lines on price charts, they are conscious of recent price movement when deciding their trades.

Can't comment on Elliot wave. It's always looked interesting, but I've never had time to actual sit down and run stats over it. It does sound a lot like anyone can make it say exactly what they want, though.

there are some good course outlines for mathematical finance at oxford uni that give clues to what they might be on about. Obviously its a secret what they are really working on. There are some interesting 2008 lectures online at Gresham college on the history of mathematical finance that deals with what they are doing at the moment . part 3 2nd lecture if i remember.

One would think if they had really cracked it mathematicians would be billionaires? They do say this is a new subject maybe 50 years old so there is lots of room to improve as they look for 'laws'.

If someone found a system that was 60% accurate all the time, yes they'd be rolling in it. However, what about a system that's 51% accurate? 50.1%? There's a lot of us working on this problem, and as time goes by the scope for diffentiating yourself from the rest of the technical analysts becomes smaller. Even accuracy isn't a complete measure; I've got a system that's about 66% accurate, but achieves that by hanging onto losing trades for far too long. By the end, it loses so much on the losing trades that it manages only a tiny profit per trade over time.

the indicator that, according to some, seems to have the greatest predictive value is the direction of net volume. Sort of like if the massive volume is equally balanced then all you need is a finger push to give it momentum in a direction assuming a frictionless environment. i suppose if one had a brokerage then one could see the data of buying and selling then develop models from that and use those to predict? But who has that advantage?

Volume is nice for showing the significance of an event (so, if the price jumps on a tiny trade, that can probably be ignored, while if it jumps on a huge trade it probably means the market is about to rocket off in one direction). I'm working with currencies, so not readily available, but nevermind :-/

 

[Jaydee]

Given automated or algorithmic trading is now up to 80% of the action in some markets and the others are catching up does it mean the end of the retail day trader?

A large proportion of this 80% are just liquidity aids, where a bot just slices and dices a large order into lots of smaller ones and feeds these into the market gradually so as not to move the price too much.

The number of prop trading robots (black boxes) is a much smaller percentage and these are often programmed based on some sort of stat arb between two or more products. Most black boxes are not based on indicators or any form of technical analysis, they mainly work at the order book level.

Regarding their impact on trading, well the liquidity bots are annoying because if you are a local on the LSE, you get charged per fill and these bleeding things give you a heck of a lot of fills if your passive order is caught (filled) by them.

As yet, I don't think there has been that much of an impact with the Black Box robots and there probably won't be until the machines actually become intelligent and can think for themselves (if you believe that will happen in the future).

 

[Pat494]

sooner or later bots are going to get intelligent and collaborate into taking over the world's finances. Lets hope someone sees the danger in time and pulls the plug.

 

[£10kLoser]

I've been sent back from the future to kill several EA creators...