and for 99.95 in the back of my car i have a dvd trading system that has a 100% win rate and 4000% annual growth rate
Why do I see "Trends" in Randomly Generated Data?
- Thread starter Rahula
- Start date
and for 99.95 in the back of my car i have a dvd trading system that has a 100% win rate and 4000% annual growth rate
Very interesting thread.
Why do I see "Trends" in Randomly Generated Data?
In random data you can see everything, including shapes of aminals, people, trends, chaos, lines, waves, see, and so on.
But it does not mean that these things does not exist.
This thread is just a thinking error.
Why do I see "Trends" in Randomly Generated Data?
In random data you can see everything, including shapes of aminals, people, trends, chaos, lines, waves, see, and so on.
But it does not mean that these things does not exist.
This thread is just a thinking error.
Quote from epetrov:
Very interesting thread.
Why do I see "Trends" in Randomly Generated Data?
In random data you can see everything, including shapes of aminals, people, trends, chaos, lines, waves, see, and so on.
But it does not mean that these things does not exist.
This thread is just a thinking error.
Or does it?
The only question that has practical value is "can a computer without human intervention be used to trade the market more profitably than human judgement?"
If it can the potential is trancendental. No sitting in front of multiple screens jumping around like a gerbil looking for the right combination of cute colored lines to trade.
Instead let the computers do the work.
The answer is Yes.
I took 10 minutes to write a simple program that I called "market generator".
It generates market data. The random process is this:
200 OHLC bars are generated.
For each bar a random number from 100 to 300 is selected (uniform distribution). This number is "number of ticks in the bar".
With the above number I then generate up and down ticks with 50/50 probability. I.e. chance of up tick = 0.5 and down tick = 0.5, so no bias.
Starting price is 100. One tick = 1.
In the attachment you can see a single simulation. This is random data. You can analyze, use TA and data-mine your ass off with it, but it is completely random and all patterns you will come up with will be meaningless bull crap.
It generates market data. The random process is this:
200 OHLC bars are generated.
For each bar a random number from 100 to 300 is selected (uniform distribution). This number is "number of ticks in the bar".
With the above number I then generate up and down ticks with 50/50 probability. I.e. chance of up tick = 0.5 and down tick = 0.5, so no bias.
Starting price is 100. One tick = 1.
In the attachment you can see a single simulation. This is random data. You can analyze, use TA and data-mine your ass off with it, but it is completely random and all patterns you will come up with will be meaningless bull crap.
Here I'm introducing "market generator" with gaps. The process is the same as above with a new feature of "gaps".
The process for generating gaps is simple:
It generates number of ticks from 0 to 15 from uniform distribution.
Then the number of ticks are generated to simulate "night session". That's it, this is a gap.
I love this chart. You can see "range-bound" market in like first 100 bars, then it "breaks out" to the upside. Oooh looky here, cup and handle pattern somewhere around bars 127-165
Then the "crash" and the "smooth up trend"
The process for generating gaps is simple:
It generates number of ticks from 0 to 15 from uniform distribution.
Then the number of ticks are generated to simulate "night session". That's it, this is a gap.
I love this chart. You can see "range-bound" market in like first 100 bars, then it "breaks out" to the upside. Oooh looky here, cup and handle pattern somewhere around bars 127-165
Then the "crash" and the "smooth up trend" In case you want to feed the OHLC data of the above chart into some hardcore analysis tools, here's the data in the attachment.
Try fibonacci numbers, moving averages, oscillators, ACD, support and resistance lines, fitting ARCH model, try something more sophisticated, like neural networks and genetic algorithms.
Try fibonacci numbers, moving averages, oscillators, ACD, support and resistance lines, fitting ARCH model, try something more sophisticated, like neural networks and genetic algorithms.
I'm stepping in to this thread a little late and this post is probably quite old. But it is quite interesting.
I agree with the asserion outlined above. But I also do not follow that assertion by then assuming that the markets are therefore not random.
I would expect that most would agree that the market is a complex and often highly chaotic system. So chaotic that noone up to now has been able to accurately model it - therefore it could be random). The issue I have with technical analysis tools (chart reading tools particularly) is that they are laughably simple methods of trying to model such a complex system. Every empirical study I'm aware of has shown that they do not work.
Consequently, I try to go with the algorithmic approach. Measurability and repeatability is what guides me rather than notional lines drawn on a chart. Also, I'm the kind of second guessing myself so automation helps save me from "me" - but that's a different story!
Thx
D
Or is it?
