Recent content by nfactorial

The purpose of this thread is to stimulate thinking. Who cares how long the thread gets... or whether market are random walks. Do you think index components exhibit flocking behaviour (i.e. the 4 flocking characteristics) around the index?

This paper might be of interest: "On the distribution of stock market data" http://lit.jinr.ru/Reports/annual-report05/ShortLITReports2004_StockMarket-163.pdf Quote: "[...] for most stocks the distribution of the closing prices normalized by traded volumes fits well the log-normal...

In recent years trading has become an activity for engineers.

I've backtested both strategies, there is no edge. That's why you're not consistent.

Replace "true" with "profitable" then. Same problem.

When you study a stochastic process you can't draw conclusions from observing a single run (realization)... Let's say you're studying the coin toss process (A=head, B=tail). Your run is AAAAABBBBB. Can you conclude that trend-following works on coin tossing? Let's say you have a different...

False. However each member of the herd keeps constant distance with its immediate neighbours and follows the center of gravity of the herd. That's well enough for a profitable strategy.

The path where those patterns were successful was one among many possible paths. No one says you can't find successful patterns... I think you need better understanding of "randomness".

:D Off topic, Why would a company prop trade except for scalability issues?

What for do you need a partner?

I see.. I guess v could be measured by # of unique price changes? So you'd measure kinetic energy on 5m,15m,30m,4h,1d and when all measures are above expected value (random walk deviation) it highlights price has gone too far too fast (flocking behaviour) At that time you can play...

Maestro, I have a question for you. Assuming one can predict with significant accuracy changes in price return, how can one develop a positive expectancy system that trades price?

We can track the flock's center of gravity with piece-wise interpolation. However we'd still need to know when the synchronization is taking place. How would you go to measure that? Can we assume that prices move exclusively when spontaneous synch occurs? We'd therefore interpolate price (or...

EC2 might be an option although I wouldn't trust Amazon with sensitive information

You can buy (lease) a small server (dell, hp, IBM,..) and put it in a datacenter. The datacenter will provide redundant power supply and internet access. You can use Ninjatrader (C#) or Marketcetera (java) for instance to run your algo on the server. Or if it's simple enough you can run the...