Will we have a computer trading in the future? What is the present state of art in machine-learning? Will it help create the future automated trader? I suspect that many on the Software thread would be able to shed more light on the subject and enlighten many of us ...... the seeds of the subject are on:Bankers make more money than Quants

This is popularly known in academia as the "ICMAUS versus real-mouse" debate.MP

Estimate in next 5 years. Some limited form going now, but in the world ofOTC derivatives, the first step has to be to be for each side to understand exactly what a deal is... and at the moment, this is not possible.Seeing FpML rolled out over the next year should eliminate understandingthe deal bit, but I suspect actually automating the flow from the receivedtrade to pricing/risk management will take another few years. Sadly, thecurrent agenda is to replace those fax machines, sack a pile of people to save some money (always brings a smile to a banker), and build someconfidence that it might be possible to do something that's actuallyinteresting.

In terms of equity trading, these people are effecively computer automated trading. If you apply the same techniques to critique a trade you have by definition a 'system'- if you repeatedly apply the same technqiues to critique trades over and over you end up effectively with an automated system.Many people in proprietory equity trading don't recognie they are infact applying a computer automated system. If a trader responds to a programmed signal he himself is only adding a layer of interfacing between his computer system and that of the order execution layer.If a trader overrides a set of repeatedly adminitered signals, what feed forward information does that trader possess ? Effecively this where discretionary trading adds value above an automated system, irrespective of how this automation is acheived.Many trading book authors have criticied automated trading. They basicaly state that it is the 'system' which is fault despite themsleves then going on and explaining a set of trading signals which they go apply in an ''auto-mated' fashion.Many equity trading book authors have knocked the idea of computer based automated trading systems on the basis they are are reliable - but that's only because they are based on mis-specified technical analysis.There is little point for a system to be based on no-more than simple averaging process, and then blame the system - it's a matter of what analytical techniques the system built on. If these are mis-specified or unreliable then the 'system' itself will be as well.For example - assess an equity time series. Does it possess a unit root at levels. Okay it's probably 1st difference stationary. Now make a trading system by applying some average and a stochastic oscillator on levels just like technical analysis. Ask yourself how relaible will this system be. It can't be very reliable. The average and the stochastic outputs are as random as the equity price time series itself because to simply feed a price time series at levels into some techncial analysis can't create 'order' from 'chaos' - no wonder many argue against systems - but it's not the 'system' which is at fault - its what they are built on.GIGO - garbage in - garbage out !!

SORRY GUYSthis should readMany equity trading book authors have knocked the idea of computer based automated trading systems on the basis they are are UNRELIABLE - but that's only because they are based on mis-specified technical analysis.thanksadam