Can I predict the value of BTC/USD?

[macsga]

This is a long post and I beg you to bear with me. I have one question for you:

Can I predict the value of BTC/USD?

Lets please talk about a bit of theory here:
Is there a mathematical model to describe a chaotic system? Don't force it. The question is a trap. We must first define what IS a chaotic system. In science, a simple chaotic system can be described by a double pendulum. If one sets it in motion from an unknown angle, the trajectory will be totally different than when you start it from another one. The complexity of the system is enormous even after a few seconds; you can take your peek here (If it isn't visible click below to start it - at the time this was written forum wouldn't let gifs animate):

http://en.wikipedia.org/wiki/File:Double-compound-pendulum.gif

Now; lets make things simpler. Say we can set with a very good precision the initial conditions. We're now talking about a SIMPLE dynamic system here with fully determined initial conditions. Now pay attention: Even if you can set the initial conditions on such a system, small differences have a totally different outcome; so as the time goes by; your predictions are getting enormously difficult to be true, thus rendering the predictions useless. This system's behaviour is called "deterministic chaos" and was first observed by Edward Lorenz who defined it like this:

Chaos: When the present determines the future, but the approximate present does not approximately determine the future. ¹

Now; I need to get your attention a bit more:

Imagine an (easy) prediction of the USD/BTC trade value on MtGox within the next 5 minutes. This is a bit tricky as well. Can this model be defined as a "deterministic chaos" model? Do I see hands raising?

-Yes miss tell us:
-No it can't.
-Why do you believe so?
-Because I can read my MACD and RSI and EMAs and can predict the value.

OK. The above is a total BS and I can scientifically certify this. This is simply NOT happening. All the above BS are ways to eliminate errors (read: normalization) of an abnormal value trajectory. As much as you can read those and take the proper decision whether to buy or sell, there's also a great possibility of an average monkey can beat your odds and win more than you by randomly pushing buttons. ²

Do I see another hand raising?

-Yes sir, tell us:
-Yes it can.
-Why do you believe so?
-Because I know certain individuals that can control the market and at a specific timeframe I can control the initial conditions. THEN AND ONLY THEN I can predict the value for the next 5 mins without looking at anything and make a move on selling or buying.

True story. A deterministic chaos system (DCS for short) can be "settled" only under the above conditions. This means you have your pendulum and you have the ability to move it at will WHENEVER you feel comfortable. Then your system can be predictable. After that you can make money and be glad about you, be in a beautiful position of knowing such great movement "factors" of the "system".

The above are my scientific conclusions. I'm not claiming the non-false. If you think otherwise, please tell me your thoughts. I initially wanted to post it on the Wall Observer BTC/USD - Bitcoin price movement tracking & discussion but I thought it was a nice topic to start and maybe, it'll give people the chance to gather their odds before blindly jumping into trading.

Cheers.

¹Danforth, Christopher M. (April 2013). "Chaos in an Atmosphere Hanging on a Wall". Mathematics of Planet Earth 2013. Retrieved 4 April 2013.
²Article: http://www.ft.com/cms/s/0/abd15744-9793-11e2-b7ef-00144feabdc0.html

[Nemo1024]

This should be a compulsory reading for any wanna-be trader who knows what the future holds. Would save them a lot of money.

[ElectricMucus]

The lowest points the double pendulum can reach is the same as a single pendulum of the same length.
The highest point is at the origin of the pendulum plus the length of the lower part of the arm.
The upper part of the arm always swings back and forth and can not flip over.

So while the overall system may be chaotic it can be predicted with specific constraints.

[kurious]

Great and erudite analysis of why all TA is utterly useless, unless you have inside knowledge of what the big movers are about to do or know news before others do.

Which is why financial markets have a preponderance of insider knowledge cheats and market manipulators, it's the only thing that gives you an edge.

Perhaps it is possible to learn about sentiments, meaning sometimes it is possible to have a good idea of how others are likely to react - but this is again at best marginal and can of course go horribly wrong.

[ElectricMucus]

Markets always bust after they boom.

[macsga]

The lowest points the double pendulum can reach is the same as a single pendulum of the same length.
The highest point is at the origin of the pendulum plus the length of the lower part of the arm.
The upper part of the arm always swings back and forth and can not flip over.

So while the overall system may be chaotic it can be predicted with specific constraints.

What you're saying can be simplified to this:

The lowest value of USD/BTC is zero
The highest value of USD/BTC is the total capitalization/BTCs produced
The value can only go up and down (well, it can also be static for a while)

Again; the timeframe is of the essence. You lose any chance of prediction within a very short timeframe. Let me explain. Say you are an observer of a tiny piece of steam flowing into a sealed box. If you observe the trajectory for 2 sec you can predict with a relatively low possibility of being wrong where the H2O steam particle will be the next fraction of the second... After 2-3 seconds you may forget it.

[oda.krell]

Who said you'd have to predict the price with absolute certainty? It would be enough to predict it often enough s.t., given the amounts you chose to trade with, you are in total profitable.

I believe predicting price (with a substantial error rate, true) is possible because it is mainly a function of human (group) behavior, and there is nothing in principle that makes predicting such a system impossible, difficult as it may be.

You on the other hand seem to think it is inherently impossible to model markets, did I get that right? I hope you don't take your linked to article as evidence for that, because it's not about modeling markets, but that a random stock selection beats a capitalization weighted index (couldn't find the actual paper either, I'm going to guess it's not peer reviewed, but feel free to correct me if I'm wrong about that).

And your little Platonic dialogue about chaotic systems is rather useless in its generality: you could equally well ask "Can I predict the weather conditions 5 minutes from now". And we know the answer to that one: not perfectly, but given enough information to run the best models we have, with high probability you will be able to do so.

[macsga]

As evidenced by the first line, the entire OP is an invitation to join a nudist colony.

If you had said, ".... bear with me ..." everyone would have sold.

If you had said, "... bull with me ..." everyone would have started an ASIC pre-order scam.

LOL! 

[macsga]

Who said you'd have to predict the price with absolute certainty? It would be enough to predict it often enough s.t., given the amounts you chose to trade with, you are in total profitable.

I believe predicting price (with a substantial error rate, true) is possible because it is mainly a function of human (group) behavior, and there is nothing in principle that makes predicting such a system impossible, difficult as it may be.

You on the other hand seem to think it is inherently impossible to model markets, did I get that right? I hope you don't take your linked to article as evidence for that, because it's not about modeling markets, but that a random stock selection beats a capitalization weighted index (couldn't find the actual paper either, I'm going to guess it's not peer reviewed, but feel free to correct me if I'm wrong about that).

And your little Platonic dialogue about chaotic systems is rather useless in its generality: you could equally well ask "Can I predict the weather conditions 5 minutes from now". And we know the answer to that one: not perfectly, but given enough information to run the best models we have, with high probability you will be able to do so.

1. There's no prediction with absolute certainty. That's what I wrote the article for.
2. We're saying the same thing, only in a different manner; as it may suit you best.
3. No you didn't get that right. Read again the 1st & last paragraph and come back.
4. I like Plato; he was my favorite Philosopher, so please accept my sincere apologies.

Interesting point of view though; thank you kindly for your comments.

[NewLiberty]

Markets always bust after they boom.

The bust defines the boom.
tautology

[macsga]

Markets always bust after they boom.

The bust defines the boom.
tautology

As the pendulum has high and lows. The same thing.

[NewLiberty]

Great and erudite analysis of why all TA is utterly useless, unless you have inside knowledge of what the big movers are about to do or know news before others do.

Which is why financial markets have a preponderance of insider knowledge cheats and market manipulators, it's the only thing that gives you an edge.

Perhaps it is possible to learn about sentiments, meaning sometimes it is possible to have a good idea of how others are likely to react - but this is again at best marginal and can of course go horribly wrong.

This thread is burning a straw man.

Having a p value of less than 100% is different from being useless.
TA is less about being always right, and more about how much confidence you put in being right at any particular point.
How much you make when you are right, vs how much you lose when you aren't is what makes it valuable or not.
For example:
RP gave himself an 80% p value for that prediction.  He may be right or not, but either way it isn't going to be indicative of whether TA is useful of not, just whether that application of it worked well or didn't.

What you have here is an analogy and a theory.  Science uses data, so from that point RP's TA is more scientific than your criticism of it.
Markets aren't pendulums.

[Vycid]

Great and erudite analysis of why all TA is utterly useless, unless you have inside knowledge of what the big movers are about to do or know news before others do.

Which is why financial markets have a preponderance of insider knowledge cheats and market manipulators, it's the only thing that gives you an edge.

Perhaps it is possible to learn about sentiments, meaning sometimes it is possible to have a good idea of how others are likely to react - but this is again at best marginal and can of course go horribly wrong.

This thread is burning a straw man.

Having a p value of less than 100% is different from being useless.
TA is less about being always right, and more about how much confidence you put in being right at any particular point.
How much you make when you are right, vs how much you lose when you aren't is what makes it valuable or not.
For example:
RP gave himself an 80% p value for that prediction.  He may be right or not, but either way it isn't going to be indicative of whether TA is useful of not, just whether that application of it worked well or didn't.

What you have here is an analogy and a theory.  Science uses data, so from that point RP's TA is more scientific than your criticism of it.
Markets aren't pendulums.

In any efficient market TA should yield zero edge over random chance.

That said, Bitcoin markets have demonstrated, time and again, startling examples of inefficiency.

The scientific way to settle this argument is to create 30 TA algos that backtest well, simulate them them for a year, and find out if there's statistical significance.

[macsga]

This thread is burning a straw man.

Having a p value of less than 100% is different from being useless.
TA is less about being always right, and more about how much confidence you put in being right at any particular point.
How much you make when you are right, vs how much you lose when you aren't is what makes it valuable or not.
For example:
RP gave himself an 80% p value for that prediction.  He may be right or not, but either way it isn't going to be indicative of whether TA is useful of not, just whether that application of it worked well or didn't.

What you have here is an analogy and a theory.  Science uses data, so from that point RP's TA is more scientific than your criticism of it.
Markets aren't pendulums.

Thanks for your criticism. Let me please answer:

This is not a game of words and surely has no intention of being an informal fallacy based on misrepresentation of an opponent's position. If that's what you made out of it I'm truly very sorry. This was not my intention.

This article is based on a chaotic system modeling in order to define the possibilities to predict WITH CERTAINTY what will happen after a specific time frame. This thing cannot be happened and it's proven. It's a theory with scientific proof and thus (until someone presents something different that kills its validity) still valid.

Markets are looking a lot like pendulums. In fact, I'd add, a lot like triple or quadruple pendulums which in their very essence can be A LOT more chaotic than a double pendulum which I above mentioned. Therefore there cannot be a certain prediction about their next timeframe of existence.

If you can provide me with scientific data that defies what I wrote; I will be glad to rephrase and/or change my thesis. Until then... me and mr. Edward Lorenz here we stand.

I'd indulge you to take a look here:
http://en.wikipedia.org/wiki/Chaos_theory

Hint: You're aware of the urban myth that physicists are one of the most wanted jobs in Wall Street... aren't you?

[macsga]

In any efficient market TA should yield zero edge over random chance.

That said, Bitcoin markets have demonstrated, time and again, startling examples of inefficiency.

The scientific way to settle this argument is to create 30 TA algos that backtest well, simulate them them for a year, and find out if there's statistical significance.

Exactly! Statistically, this would offer a much better model, in order to make valid predictions. It's like narrowing down the specific movement factors of the pendulum and predicting timeframe by timeframe what will  happen. But this NEEDS HISTORICAL DATA (with certain similarities to the current conditions)! I do have a couple of ideas about it, but are a bit too crazy to be honest for me to start posting them on a public forum on the net...

[Altoidnerd]

While chaotic systems are, well, chaotic, their trajectories often tend to evolve toward one of a number of attractors, at least temporarily.

While it is not in principle feasible to determine a trajectory of some dynamical variable p (say, price) as a function of time p(t) as you can for simpler systems, the case of chaos and attractors is unique in that the phase space is of a dimension that is non-integer.

A double pendulum for example:  You might not be able to get the position q(t) or momentum q'(t) as a function of time, but plotting the set of points {q(t),q'(t)} measured when a damped, driven double pendulum is set in motion will yield fractals.  The plot is known as a pioncare section.  The 2-D plot may have dimension 1.578887 - the trajectory is drawn on the 2-D plane, but the curves with mathematical certainty do not live in a 2-D space.  They have less room than all of R² to move.

pioncare section pic:


The fractal nature of the trajectories suggests that despite inability to reliably produce a trajectory, there are always pairs of values ( q(t), q'(t) } which are never allowed!  That lends predictive power, albeit in a different sense than what we're used to.




And we have a tulip fractal:

[sublime5447]

Great and erudite analysis of why all TA is utterly useless, unless you have inside knowledge of what the big movers are about to do or know news before others do.

Which is why financial markets have a preponderance of insider knowledge cheats and market manipulators, it's the only thing that gives you an edge.

Perhaps it is possible to learn about sentiments, meaning sometimes it is possible to have a good idea of how others are likely to react - but this is again at best marginal and can of course go horribly wrong.

This thread is burning a straw man.

Having a p value of less than 100% is different from being useless.
TA is less about being always right, and more about how much confidence you put in being right at any particular point.
How much you make when you are right, vs how much you lose when you aren't is what makes it valuable or not.
For example:
RP gave himself an 80% p value for that prediction.  He may be right or not, but either way it isn't going to be indicative of whether TA is useful of not, just whether that application of it worked well or didn't.

What you have here is an analogy and a theory.  Science uses data, so from that point RP's TA is more scientific than your criticism of it.
Markets aren't pendulums.

In any efficient market TA should yield zero edge over random chance.

That said, Bitcoin markets have demonstrated, time and again, startling examples of inefficiency.

The scientific way to settle this argument is to create 30 TA algos that backtest well, simulate them them for a year, and find out if there's statistical significance.

Just the fact that other traders use TA makes it viable as a predictive indicator.

[macsga]

While chaotic systems are, well, chaotic, their trajectories often tend to evolve toward one of a number of attractors, at least temporarily.

While it is not in principle feasible to determine a trajectory of some dynamical variable p (say, price) as a function of time p(t) as you can for simpler systems, the case of chaos and attractors is unique in that the phase space is of a dimension that is non-integer.

A double pendulum for example:  You might not be able to get the position q(t) or momentum q'(t) as a function of time, but plotting the set of points {q(t),q'(t)} measured when damped, driven a double pendulum is set going will give you fractals.  The plot is known as a pioncare section.  The 2-D plot may have dimension 1.578887 - the trajectory is drawn on the 2-D plane, but the curves with mathematical certainty do not live in a 2-D space.  They have less room than all of R² to move.

The fractal nature of the trajectories suggests that despite inability to reliably produce a trajectory, there are always pairs of values ( q(t), q'(t) } which are never allowed!  That lends predictive power, albeit in a different sense than what we're used to.

A fellow Physicist!!! WE ARE MANY!!!

Thanks for the comments - you've wrote some of the crazy ideas I wanted to post, mentioned on the above comment... BTW: It's Poincare section.
http://www.mwit.ac.th/~physicslab/applet_04/fun@learning/JAVA/pendchao/pendchao.html

[Altoidnerd]

A fellow Physicist!!! WE ARE MANY!!!

Thanks for the comments - you've wrote some of the crazy ideas I wanted to post, mentioned on the above comment... BTW: It's Poincare section.
http://www.mwit.ac.th/~physicslab/applet_04/fun@learning/JAVA/pendchao/pendchao.html

Coolness, I need help developing the bitcoin RF idea.  :-P  I sounded like a loon describing it but its real!

[KFR]

Excellent thread.  My compliments Mac.