Random Walk Theory Proved, once and for all.

Justin Mamis had a very good argument as to why markets aren't random.

It has to do with the reaction time and confirmation demands of verious market particpants.

By trading's very nature, it can't be random. It isn't as predictable as many people think...but it is predictable enough to make money, provided you are using proper risk managemement.
 
how about creating a "walk forward" chart of random numbers within the framework of normal stock/future movements and presenting it to several TA experts one bar at a time on the hard right edge of the chart, then have the experts make decisions based on what they see to determine if any "edge" exists in TA????

this will quickly prove the point!


surf


ps. great thead by the way!
 
I think this is a pretty good idea by OP.
I've run this exercise many moons ago, and it is very useful towards ungluing yourself from false beliefs in many so called "objective" TA dogma, and understanding that TA is a reference to be built around, not followed as law.

For posters that say show me a stock that goes 12, -15, +90, -10, it's not that simple to characterize a random system. If you start by assuming normal random (although markets are not purely normal) behavior, you will find stocks also have 1, 2, and 3 sigma bounds. Run any distribution of a typical stock, and you'll find the 1sigma distribution is much tighter than the sequence shown above, (meaning ~70% of the time you won't have a variance like +150, -40, +569).
You have to be able to characterize all of the properties of a normal distribution before stating that it is or is not normal.

As stated earlier, markets are not normal, but they are CLOSE enough to normal to use it as a model, and then take into consideration fat tails and 10 sigma events DO HAPPEN more than the simple model would predict, although they are highly unlikely.

There is a simple reason to understanding why markets don't go sideways, nor for that matter downward forever. Run a normal distribution with a mean of zero, you'll see the long sideways random behavior you're familiar with.

Now change the mean to be offset by some amount, say 0.5%. What happens to the signal over the long run? Take a look at the distribution of daily changes on the markets, they don't have a mean of zero. That's the key to understanding long term upward drift and some proof of an edge.

Again, great exercise to the original poster and those that understand it.
 
Agree, the market is non-stationary so Statistical Analysis is very difficult. How do you define your data set? It's a lot of trial and error. Find something that works and be ready to tweek it when it stops working. Trading is just another form of business and all businesses have to make adjustments to adapt to the ever-changing external environment.

Quote from matador04:

Also take into account discrete vs continuous movements
More...
 
TheDarkness,
These are 9 consecutive charts produced by the formulas in Excel worksheet.
RAND() RAND() INT(B1*2)-1 (C1-0.5)*2 D1*A1
RAND() RAND() INT(B2*2)-1 (C2-0.5)*2 D2*A2 SUM(FIRST:E2) SUM(SECOND:D2)
Name E1 as "FIRST"
Name D1 as "SECOND"

You can prove it yourself. No need to cheery picking.
Mu.
 

Attachments

Quote from dtrader98:


For posters that say show me a stock that goes 12, -15, +90, -10, it's not that simple to characterize a random system. If you start by assuming normal random (although markets are not purely normal) behavior, you will find stocks also have 1, 2, and 3 sigma bounds. Run any distribution of a typical stock, and you'll find the 1sigma distribution is much tighter than the sequence shown above, (meaning ~70% of the time you won't have a variance like +150, -40, +569).
You have to be able to characterize all of the properties of a normal distribution before stating that it is or is not normal.
More...

Dog chasing its tail.

Did the original poster measure the sigma variations in a "typical" stock?

Did the original poster incorporate those sigma variations into the model?

No and no.

So what if he did? So what.

The algorithm would punch out the same price variances and occurrences it was modeled after.

In other words, the algorithm made A COPY of the price sequence it was trained on.

Any reasonable observer understands this proves (or disproves) nothing.
 
Quote from mu200411:

TheDarkness,
These are 9 consecutive charts produced by the formulas in Excel worksheet.
RAND() RAND() INT(B1*2)-1 (C1-0.5)*2 D1*A1
RAND() RAND() INT(B2*2)-1 (C2-0.5)*2 D2*A2 SUM(FIRST:E2) SUM(SECOND:D2)
Name E1 as "FIRST"
Name D1 as "SECOND"

You can prove it yourself. No need to cheery picking.
Mu.
More...

Again.

What exactly does this "prove"?

That a second data set can be made to imitate the first?

WOW. EARTHSHATTERING!
 
Quote from mu200411:

In this thread Random Walk Theory will be proved.
There will be no discussion about TA.

If I can draw charts using Random number series that look like stock charts, will it be a proof that Random Walk Theory is true or a proof that Random number series also follow the Nature law of Elliott?:confused:

Anyway let's produce some charts first.:D

N.B. Please move this thread to an appropriate forum.
More...

If you can draw charts using Random number series that look like stock charts...
All that it proves is that you can draw charts using Random number series that look like stock charts

Does not prove anything about the nature of the markets at all....

I agree a lot of the TA talk by the gurus ( many of them on this forum ) are unsubstantiated trash... but what you are doing here is of no use either...
 
The thing is, most price charts are not actually the reality of what happened in the market's double auction, so the ability to reproduce a price chart that looks like a price chart doesn't prove, well, anything.

One time, when I was just starting out (not long ago), watching the price ladder on the YM, prices went kinda crazy. It must have been an FOMC day or something like it, but I didn't know that at the time. Price spiked 100 points in a couple seconds. I didn't know what was going on. I thought there was something wrong with my data feed. But then, price was jumping everywhere up and down as orders came into the market way above the ask and way below the bid. After that first minute, price started moving sequentially, like it normally does. Within a couple minutes, it was up 150 points. Over a period of 15 minutes or so, it retraced most of that price move. Then, it went up again, one tick at a time.

I've seen gaps up and down in the ER2 as well in the middle of the day. Price is just cooking along sequentially, and all of a sudden, price is jumping down, and touching everywhere in the gap.

It would to be quite a feat to try and imitate these events on one of your randomly generated price charts. I guess I would start believing a little if you could successfully randomly generate a complete double auction with however many thousands of orders being matched and rare price action occurances like these ones I have described.

Has anyone else seen something like this, or have I slipped over the edge finally?

EDIT: by the way, isn't there something about computer random number generators not being truly random, but they can only simulate randomness? Doesn't that throw a monkey wrench into this, er, study?
 
You must log in or register to reply here.
Back
Top