Quote:
|
Originally Posted by question
TraderABC, here's an exercise in statistics if the markets are truly random:
1. Let's accept that news events and new information can be considered random to the outside viewer in the fact that it can either be good or bad.
2. Let's assume that the random security we're talking about is EUR/USD which is currently trading at 1 EUR to 1 USD.
3. If the market movements are totally random, after a sufficiently large number of iterations, the EUR/USD will be trading at or very near to 1 EUR to 1 USD.(why?)
4. The reason for this is that the standard deviation of an event decreases after each successive iteration.
5. Toss a coin. Assume heads equal 1 and tails equal 0. After one toss, it is impossible to have an average value of 0.5. But after 2 tosses, you have a 50% chance of having a value of 0.5. Toss the coin more and more, and you will see that more and more of your average value possibilities will hover around 0.5.
6. This whole exercise proves there is some non-random component to the markets.
|
I don't quite agree with that.
Firstly, in the coin example, you need at least 100 samples to form the basic normal distribution to have statistically significant probability of 50:50 between heads or tail.
Secondly, the random walk theory simply states that the market follow an autoregressive order (1). Which means that the best prediction for what the market will do at t+1 is t-1. In otherwords, only the most recent time period matters.
Thirdly, the standard deviation of regression errors (i.e. difference between the mean regression line and actual prices) actually remains constant. Because in a random walk situation, the only thing that matters is the most recent price data. So regardless of the number of previous iteration, the standard deviation of errors that counts is only the most recent one. That's why prices don't converge to 1 EURUSD.
I hope it makes sense
