Man vs Machine

[Photon]

There are certainly things that machine can do but human can't. But there are things that human can do but machines can't. It just doesn't have to be mutually exclusive.

But also factor in cost/benefit in equation. I'm sure its perfectly feasible to come up a bot that factors in all eventuality like delays, horse withdrawal, sweating, gamble, fixing etc. but at what cost? Surely the level of investment needed to achieve this goal would outweigh potential return from horse race betting after allowing for future uncertainty, premium charges, etc. There are machines that can beat greatest of chess masters but do you know how much investment is needed to achieve that?

So I say stop worrying about who/what is more superior but start using both to your advantage.

[Akindwurd]

The most advanced optical recognition technology is now able to distinguish between cats and dogs, although I don't think it is yet capable of spotting the difference between a horse and a nag.

However, it would take a brave trader to take a position on what computers can or can't do ten or twenty years from now.

[hgodden]

They can do the simple things now, for instance if you have a simple clear cut edge, but I wouldnt have thought (yet) they could replicate what a manual trader does - the markets are too alive and dynamic for that. Until the day they can programme robots to play football or surf better than humans then they wont be making hyper complicated judgements on the racing markets

[Iron]

Until the day they can programme robots to play football or surf better than humans then they wont be making hyper complicated judgements on the racing markets

I disagree.

In the financial markets, bots make literally billions of transactions per day, and you have bots trading every timescale from milliseconds up to the monthly charts.

I bet if you analysed your trading over a day, you'd find that it boiled down to using a few techniques. From there, getting a computer to recognise the patterns you utilise and act as you do probably isn't that much of a stretch. It's possible that you and other pros have some sixth sense of the market, and that your unconscious knows instantly stuff that a supercomputer couldn't calculate in a month of Sundays, but I doubt it.

Supercomputers can beat grandmasters at chess, so with greatest respect they are capable of performing the calculations that you perform when weighing up a market.

Jeff

[to75ne]

Until the day they can programme robots to play football or surf better than humans then they wont be making hyper complicated judgements on the racing markets

I disagree.

In the financial markets, bots make literally billions of transactions per day, and you have bots trading every timescale from milliseconds up to the monthly charts.

I bet if you analysed your trading over a day, you'd find that it boiled down to using a few techniques. From there, getting a computer to recognise the patterns you utilise and act as you do probably isn't that much of a stretch. It's possible that you and other pros have some sixth sense of the market, and that your unconscious knows instantly stuff that a supercomputer couldn't calculate in a month of Sundays, but I doubt it.

Supercomputers can beat grandmasters at chess, so with greatest respect they are capable of performing the calculations that you perform when weighing up a market.

Jeff

jeff,

you are possibly correct with regards a super computer but, how much does a a super computer cost. then you have to have an operating system more or less custom built,and on top of that your trading bot code. er few mill i would guess, then your going to need to house it somewhere suitable (power supply, cooling, ventilation/air flow control etc), which will add to your running costs of probably several hundred thousand er year.

at this moment in time, what with commission and pc, i doubt if it would be profitable

[Iron]

Tony

Possibly, although it wouldn't surprise me if there are people with desktop PCs and self-programmed bots who switch their bot on in the morning and come home to a nice profit! Peter has repeatedly said that to cross the threshold between near break even and profitability doesn't require you to get much right, and surely it can't be beyond an experienced trader and programmer to write a bot that avoids obvious mistakes.

The other week, a guy I correspond with sent me a video of a bot of his making something like £20 from a pre-off race, so it can be done. I don't know whether he manages that consistently, however.

Jeff

[to75ne]

Jeff,

i don,t doubt it can be done to a reasonable extent; and as i said earlier i have one that's reasonable on certain greyhound markets, but i would never let it loose on its own because, it can only blindly do what it as be told to do if certain conditions are meet. if those conditions are meet, and other factors are happening in the market, that it as no instructions to "notice" it will carry out its trade regardless, it cant pause for thought so to speak, just do as it was told.

i still believe that it is sometime off before a bot will exist that can trade more or less every race on weak raining monday afternoon, or a frantic saturday afternoon and trade as efficiently - profitably all class's/grades/distances/ flat/jumps/ delays etc as good a reasonable trader.

then again, if someone had one, they would be stupid to let everyone know, so who really knows?

personally i have seen no evidence of a "clever" bot in operation and i said earlier, i think someone would need to know how to quantify and then encode fuzzy things like gut feeling, intuition, emotions etc,

as an engineer i do believe (i am bound to believe) that at some point in the future it will be possible, but i think that will only happen when neuroscientists come up with a theory of what thought/memory/gut feeling etc actually are, and how to encode them. quite sometime off i believe.

tony

[pdupre1961]

in real life situation a computer will never outperform a human.

gutuami you are so naive, computers perform calculations much faster than humans. Humans only win when it comes to judgement.

As a chess player, currently ranked 57,787 in the world - I have been using chess playing computers for nearly 30 years and I can tell you that they now far outstrip their human opponents at speed chess. The program I used about 5 years ago was called Fritz 5 and in the end I gave it away as I could not beat it on 5 minutes per game mode. The difference is that computer programs have been around a long time.

I'm sure there is many bots out there making money for their programmers - for instance the UberBot.

I have been working on a trading bot for about 3 months, but only from a simulation point of view. I will succeed...

[gutuami]

thanks for compliment pdupre1961. But I think you missed the basic idea. An airplane can fly but that doesn't mean that it outperforms a human, does it?
I meant things than can not be programmed like emotions, humor and instincts.

[maycontainnuts]

That link to uber bot is a bit curious, the poster claims to have vast experiance but somehow missed that Betfair introduced cross matching AKA the 'uber bot' a couple of years before they wrote that post. Doh!

[James1st]

thanks for compliment pdupre1961. But I think you missed the basic idea. An airplane can fly but that doesn't mean that it outperforms a human, does it?
I meant things than can not be programmed like emotions, humor and instincts.

Gutuami, FAIL!

According to Boeing, "In the early days of flight, approximately 80 percent of accidents were caused by the machine and 20 percent were caused by human error. Today that statistic has reversed. Approximately 80 percent of airplane accidents are due to human error (pilots, air traffic controllers, mechanics, etc.) and 20 percent are due to machine (equipment) failures"

Of course a bot can be programmed to do anything a human is likely to do whilst trading (I have quite a few that work including that referenced by Ferru above). The last thing one needs when writing a bot is to include "human emotions, humour or instincts".

The market behaves in a certain manner fairly consistently and replicating a trading process or series of processes is not beyond the skills of any semi decent coder.

People can point to the odd events that happen in a market (withdrawn horses, mad bombers etc) but every known event will not necessarily impact on a well coded bot and as long as it is designed to failsafe, then there is no reason why they shouldn't be used.

[to75ne]

The last thing one needs when writing a bot is to include "human emotions, humour or instincts".

[/quote]

james,
that is something i have not considered. good point. time to re-think.

[gutuami]

1. human emotions can not be programmed period
2. man vs machine is actually man vs somebody who designed the algorithm of that machine. So in the end it comes up a man vs man competition. That's the idea.

Maybe in the future the man will design a humanoid robot with artificial intelligence. One day the robot will become interested in trading on betfair and will study and teach itself trading without any human intervention. After that it will compete with the best traders and outperform them all. At the moment there's no such robots.

[Iron]

1. human emotions can not be programmed period

Not sure anyone said otherwise.

2. man vs machine is actually man vs somebody who designed the algorithm of that machine. So in the end it comes up a man vs man competition. That's the idea.

True, but there is often a big gulf between someone's intentions when they set out trading and what they actually do in the heat of battle, when their fear and greed are running wild. Having a bot means that you can be sure that a solid methodology that is backed up by lots of evidence will be executed flawlessly. BTW, another thing about computers is that they aren't known to have lapses of concentration!

Maybe in the future the man will design a humanoid robot with artificial intelligence. One day the robot will become interested in trading on betfair and will study and teach itself trading without any human intervention.

Aren't we already there, with neural networks? Surely you can program a bot to make observations about the market and adapt its approach accordingly.

Jeff

[gutuami]

IMHO if you take the best bot and ask it to do 10 trades and then ask an average trader to do 10 trades on the same market the bot will lose. If that is false than very soon we will buy from Tesco bots on sale . Why do we need traders on the markets when we could have bots with proven record