Computer made trading patterns

ZEN archer

Experienced member
Messages
1,528
Likes
241
Apparently lots of trades are done by computers, so it would be nice/profitable to find the way how to exploit the situation. There could be some computer driven patterns worth trading.

Have you found any?
 
sorry Z...I was just wondering why computer driven patterns are different to any other pattern ?

I suppose you mean the depth and complexity of such patterns unrecognised by the human Trading eye ?

N
 
sorry Z...I was just wondering why computer driven patterns are different to any other pattern ?

I suppose you mean the depth and complexity of such patterns unrecognised by the human Trading eye ?

N

Maybe there is no such thing - most likely they just produce random moves with no patterns that can be traded profitably?
 
My personal favorite for identifying patterns is a neural network. Doing a google search for Neural Network Pattern Recognition would yield much information on the subject. That said, ANN trading isn't new, and since it isn't the defacto standard for beating the markets, it isn't really that reliable. On the other hand, real neural networks can make some of the best traders out there - which means that the network just needs to be constructed correctly.

That said, consider this. Trading software evolves. The market makers, the home ATM trades, all of it. A strategy that works now may not work tomorrow, or yesterday (or next year / the year before, time scale is moot)

The point I'm getting at is the micro-patterns you'd be looking for are burried in a huge mass of macro-patterns. You'd need to be able to identify trades by origin, to identify their patterns. Looking at "the market", all you can do is pattern the market. And since the market evolves, by the time you isolate the pattern, it is different than it was.
 
My personal favorite for identifying patterns is a neural network. Doing a google search for Neural Network Pattern Recognition would yield much information on the subject. That said, ANN trading isn't new, and since it isn't the defacto standard for beating the markets, it isn't really that reliable. On the other hand, real neural networks can make some of the best traders out there - which means that the network just needs to be constructed correctly.

That said, consider this. Trading software evolves. The market makers, the home ATM trades, all of it. A strategy that works now may not work tomorrow, or yesterday (or next year / the year before, time scale is moot)

The point I'm getting at is the micro-patterns you'd be looking for are burried in a huge mass of macro-patterns. You'd need to be able to identify trades by origin, to identify their patterns. Looking at "the market", all you can do is pattern the market. And since the market evolves, by the time you isolate the pattern, it is different than it was.

Thanks AngryErik - I'll have a look. I've tried it before, but gave up (too complicated, time consuming etc.)
 
But there is. I can even tell you what algorithm they use:

1. Sell at price N
2. Move price to N - m
3. Buy back at price N - p.

Armed with this info, you should be able to beat them now.

That's very kind of you BJ (y)

It looks impressive, so I'll go all in from tomorrow:cool:
 
Don't assume corporate computer trading is always profitable all of the time. Knight Capital, who is a NYSE Market Maker, lost the more than equivalent of an entire year's profit in just a few minutes due to their computer trading. In fact they almost because insolvent because of it.
Knight Capital Says Trading Glitch Cost It $440 Million - NYTimes.com

Funny how when they lose it's call a "glitch".

Peter
 
Top