Wall Observer BTC/USD - Bitcoin price movement tracking & discussion

[aminorex]

One of these plots is actual hourly closing price (USD/BTC) from one exchange, the other is a series of fake prices generated by a simple "log-Brownian" model.

Can you tell which is which, without looking at other charts?
If you think you can, what difference do you see?

Honestly, they both look very plausible. In fact I wouldn't be surprised if they're both BTC/USD, but from different exchanges.

The red line (Y) is Bitstamp's hourly closing price from 2014-01-05 to 2014-01-13.

The blue line (X) is generated by starting with

  X(0) = 840
  Z(0) = log_10(X(0)),

then, for i = 1,2,...

  Z(i) = Z(i-1) + 0.058*RND(),
  X(i)  = 10^Z(i)

where RND() is a random number with normal distribution, mean 0 and standard deviation 1.

In simple words, the simulated price X(i) increases or decreases at each step by a random percentage, without regard for the past history.

I had to try twice, with different random generator's seeds; on my first attempt the simulated price dropped too low at one point, it would have been a dead giveaway.

what about over a longer time period?

It's called a jump-diffusion.  Essentially all exchange-traded investments are modeled pretty well as jump-diffusions.  It is difficult but not impossible to discriminate the signal from the noise at finer scales.  Self-similarity argues that it is also difficult to discriminate the signal from the noise on larger scales; however, comparing BTC since 2011 to a log-normal process over the same number of points would not be remotely confusing.  The jumps have structure, but it is most not endogenous structure, so they are not markovian.

[Dragonkiller]

One of these plots is actual hourly closing price (USD/BTC) from one exchange, the other is a series of fake prices generated by a simple "log-Brownian" model.

Can you tell which is which, without looking at other charts?
If you think you can, what difference do you see?

Honestly, they both look very plausible. In fact I wouldn't be surprised if they're both BTC/USD, but from different exchanges.

The red line (Y) is Bitstamp's hourly closing price from 2014-01-05 to 2014-01-13.

The blue line (X) is generated by starting with

  X(0) = 840
  Z(0) = log_10(X(0)),

then, for i = 1,2,...

  Z(i) = Z(i-1) + 0.058*RND(),
  X(i)  = 10^Z(i)

where RND() is a random number with normal distribution, mean 0 and standard deviation 1.

In simple words, the simulated price X(i) increases or decreases at each step by a random percentage, without regard for the past history.

I had to try twice, with different random generator's seeds; on my first attempt the simulated price dropped too low at one point, it would have been a dead giveaway.

what about over a longer time period?

It's called a jump-diffusion.  Essentially all exchange-traded investments are modeled pretty well as jump-diffusions.  It is difficult but not impossible to discriminate the signal from the noise at finer scales.  Self-similarity argues that it is also difficult to discriminate the signal from the noise on larger scales; however, comparing BTC since 2011 to a log-normal process over the same number of points would not be remotely confusing.  The jumps have structure, but it is most not endogenous structure, so they are not markovian.

had to look up several things to understand that. thanks for the insight

[KFR]

http://www.independent.co.uk/arts-entertainment/music/news/lily-allen-calls-herself-an-idiot-for-declining-bitcoin-payment-now-worth-millions-9046211.html

posting cuz new to me!

i guess this is big news in england

Actually the consensus over here appears to be that she has mixed up bitcoin and Linden Dollars (L$), which makes sense given the timing and the virtual venue.

[Dragonkiller]

http://www.bbc.co.uk/news/technology-25727333

http://www.bbc.co.uk/news/technology-21032503

[ChartBuddy]

Explanation

[shmadz]

i panic sold once and i felt so dumb that i've adopted the strict HODL policy and it's been working out for me so far

you guys should try it out too

I've moved to a pure SODL strategy, since we're clearly heading to zero.

Why are you still here when you think we are going to zero?

Just as the Chinese government has recently warned citizens of the dangers of bitcoin, I am warning citizens of bitcointalk of the dangers of bitcoins, to help as many people as possible.  Only Satoshi and maybe 2 or 3 other people will make any money from bitcoin.  Everyone else will lose money, and probably destroy their life, as the Chinese government has explained.

Still just trying to catch up, but I just have to blurt out, this was epic!

Well done Proudhon. http://www.youtube.com/watch?v=A7TuFy0fcuw

[Walsoraj]

http://www.bbc.co.uk/news/technology-25727333

http://www.bbc.co.uk/news/technology-21032503

Irrelevant. Nobody buys bitcoin or any other cryptocurrency because they are superior payment systems. We buy for one and only one reason: to hodl and sodl our way to more fiat.

Please do not respond to my post with lies. thx.

 

[shmadz]

Actually bitcoin miners are fucked.
99% of the miners (paid in BTC) won't reach ROI ever !
They hoard the coins and wait for a price increase, otherwise they all lose money.

If the mining business collapses and 5000 coins are sold daily, there will be blood (candles)  

At the moment, only a few can win. Most can't cash out or they will lose money (market collapse). That's one of the main reason for the steady uptrend. It's not market adoption or increased utility. Transactions didn't increase very much the last 3 years. The price did with only a few percentage of bitcoins traded.

lol, just lol

*edit* I'm finally in the 1%!!! woohoo!

[notme]

One of these plots is actual hourly closing price (USD/BTC) from one exchange, the other is a series of fake prices generated by a simple "log-Brownian" model.

Can you tell which is which, without looking at other charts?
If you think you can, what difference do you see?

Honestly, they both look very plausible. In fact I wouldn't be surprised if they're both BTC/USD, but from different exchanges.

The red line (Y) is Bitstamp's hourly closing price from 2014-01-05 to 2014-01-13.

The blue line (X) is generated by starting with

  X(0) = 840
  Z(0) = log_10(X(0)),

then, for i = 1,2,...

  Z(i) = Z(i-1) + 0.058*RND(),
  X(i)  = 10^Z(i)

where RND() is a random number with normal distribution, mean 0 and standard deviation 1.

In simple words, the simulated price X(i) increases or decreases at each step by a random percentage, without regard for the past history.

I had to try twice, with different random generator's seeds; on my first attempt the simulated price dropped too low at one point, it would have been a dead giveaway.

Granted it is only one sample, I picked blue as the fake data right away.

[EFS]

yeah, but still it made enough headlines that a British friends told me about it:)  (he is not a bitcoiner)

I can't believe you have a friend who isn't a bitcoiner goat, after your success and reputation in community.

[JorgeStolfi]

Granted it is only one sample, I picked blue as the fake data right away.

Good. Can you tell what looked "wrong" with that one?

[joesmoe2012]

Pretty stagnant day since I woke up....

[notme]

Granted it is only one sample, I picked blue as the fake data right away.

Good. Can you tell what looked "wrong" with that one?

Brownian motion follows a normal distribution.  Price changes follow a power law distribution with clustering of high standard deviation data points.

[ChartBuddy]

Explanation

[byronbb]

lol news out of china, going north.

[ShroomsKit]

lol news out of china, going north.

And you're not going to post it?

[JorgeStolfi]

Granted it is only one sample, I picked blue as the fake data right away.

Good. Can you tell what looked "wrong" with that one?

Brownian motion follows a normal distribution.  Price changes follow a power law distribution with clustering of high standard deviation data points.

Indeed, in other time intervals and other scales it is quite visible that the amplitude of the "random" component (the price changes) varies over a scale of tens of hours, so that there are "steadier days" and "wilder days".  That should account for "clustering of high standard deviation data points".

In that interval, in particular, the red curve has larger deviations in the first half, smaller in the middle (from ~120 to ~150) and then intermediate ones near the end.

However, that effect seems to be rather subtle.  Perhaps it was the non-normal distribution of the changes that made the red curve look different?

[bobdude17]

lol news out of china, going north.

What news?

[vabtc]

I told him about BTC in 2011 but he never got involved. Hates himself for it. I know many people like this. Most of them still have not got in at all!

I can't tell you how many people I know like this. They keep asking me and I know they are still open to the idea, it's just a steep learning curve.

[byronbb]

lol news out of china, going north.

And you're not going to post it?

I think I got trolled by btce trollbox.