For those folks who aren't clear about the basic concepts, the following URL may help:
http://www.indexarb.com/
Best to all, Steve46
http://www.indexarb.com/
Best to all, Steve46
Program Trading
- Thread starter waggie945
- Start date
i completely agree to the content of this post and to the careful tone. the thing is this IMHO: if you use patternCreation, patternEvaluation and patternRecognition in combination, then you do not care what the input is. be it a bunch of indicators, a set of candle sticks, a combination of earnings forecasts and cointegrationstatespacehurstexponentialwhatever thing - if there is an inefficiency the machines will find it and exploit it. as long as it fades away. thus, whatever we are doing here in painful work is delegated to the machine.
if you think you are doing better than a machine try to find a systematic trading system without a computer. or optimise a strategy by hand. you see the point.
it might happen that computers will produce traces within their trading, exploitable by other computers, so i think the game will be between the machines. like real arbitrage, which still exists, but happens in micro seconds and requires absolute minimal trading cost.
stock arbitrage is probably subject to most intense use of quant models. bnp cooper neff, ren tech and some others have been the biggest players on the exchanges for years.
if this theory is valid, then flight to illiquidity will give just a short break. why? if it works - the machines will find it out too.
it is a question of time. but i just have to look across my desk and see what we are able to perform now and compare that with five years ago to know where the train is heading.
people should prepare. single trader trying to beat the world will be harder and harder. fewer and fewer will pass the test by efficiency.
peace
i completely DISAGREE to the content of this post. The movement of market is driven by emotions: fear and greed. Technology used in trading evolves, but human nature doesn't.
Here's my view of it:
Program trading will indeed get stronger in the future, competing in their strategies (e.g. index arbitrage, convertible arbitrage, fixed income arbitrage...) will get harder for humans in the shorter time frames, since these routines simply can be better accomplished by a computer than by a brain.
But with regards to strategical interactions, there will always be the "human factor", the uncertainty of behaviour. "All it takes to prove a trading strategy wrong is a single trader somewhere in the world" (Marc Douglas, "Trading in the Zone".)
This is the same trend we can see in chess:
in a very interesting analysis of the last match Kasparov had with Deep Blue, he explained that his strategy is to always prepare his attacks 8-9 steps in advance, because it is practically impossible to beat the computer in the shorter time frames (= less than 8-9 steps). Go ahead and play tic tac toe against a PC: I bet that after not more than 100 draws, you will freak out so the computer wins.
It's also the same with meteorology: it's easy to predict the weather for the next, say 5 minutes to 8 hours.
But everything more than 1 day becomes more and more speculative.
Garbage in, garbage out.
Program trading will indeed get stronger in the future, competing in their strategies (e.g. index arbitrage, convertible arbitrage, fixed income arbitrage...) will get harder for humans in the shorter time frames, since these routines simply can be better accomplished by a computer than by a brain.
But with regards to strategical interactions, there will always be the "human factor", the uncertainty of behaviour. "All it takes to prove a trading strategy wrong is a single trader somewhere in the world" (Marc Douglas, "Trading in the Zone".)
This is the same trend we can see in chess:
in a very interesting analysis of the last match Kasparov had with Deep Blue, he explained that his strategy is to always prepare his attacks 8-9 steps in advance, because it is practically impossible to beat the computer in the shorter time frames (= less than 8-9 steps). Go ahead and play tic tac toe against a PC: I bet that after not more than 100 draws, you will freak out so the computer wins.
It's also the same with meteorology: it's easy to predict the weather for the next, say 5 minutes to 8 hours.
But everything more than 1 day becomes more and more speculative.
Garbage in, garbage out.
IMHO, the most difficult part with program trading is chart pattern recognition. If the strategy has little to do with chart patterns, then it can be implemented by programs. Any idea on how to describe chart patterns in computer languages?
Exactly,
A program picking up 10k shares by refreshing a bid by 500 and supporting the pullbacks is very easy to spot. A smart and experienced institutional buyer is a bit tougher to pick out.
Dead volatility by buying programs makes money for noone. If two programs are trading against each other, one will have to blow out in order for the other to make smth. That would mean a HUGE spike in one direction.
1987 crash guys. Think that type of scenario. Or think Long Term Capital. Everything is nice and safe thanks to models and programs till something goes wrong.