Price vs Difficulty Charts - indicators for buying or mining

[tripper22]

bitcoinbull. Thanks for providing these charts. I hope to see them more frequently. Keep up the good work.

[1714149106]

1714149106

[1714149106]

1714149106

[1714149106]

1714149106

[1714149106]

1714149106

[1714149106]

1714149106

[1714149106]

1714149106

[bitcoinBull]

Any chance you could do a log(price)/log(difficulty) (or maybe a (10+log(price))/log(difficulty)) chart?

The log(price)/log(difficulty) looks almost the same: https://i.imgur.com/sytsN.png


btw, the collection of charts is here:  http://bitcoinbull.imgur.com/all/

[bitcoinBull]

Why should we expect price to go up if difficulty increases?  Isn't it the other way around?

Price goes up when difficulty increases because as they become harder to get from mining, people buy.

Difficulty increases when price goes up because higher prices offset the increase in difficulty, making mining attractive again.

The question here is which is the leading indicator and which is the lagging.

Pretty obvious to me that price drives difficulty.  If it's super profitable to mine, then people start rigs.  That increases difficulty.  I don't see too many miners in the buying bitcoin arena (otherwise they would have never started mining, since it's less profitable).

Indeed, obviously price drives difficulty.  Less obvious other way around: difficulty driving the price. 

They are correlated, which isn't necessarily a causal relationship, but could be.

I think its causal in both directions (Two-Way Causality).

[k]

hi bitcoinBull,

great charts. Would it be better to use the actual difficulty for a particular 2016 blocks rather than the difficulty implied by the calculated hashing rate for 504 blocks?

[bitcoinBull]

hi bitcoinBull,

great charts. Would it be better to use the actual difficulty for a particular 2016 blocks rather than the difficulty implied by the calculated hashing rate for 504 blocks?

Nah.  We're trying measure the size/power of the network at the time, so we want to take the average hash rate over as small a window as possible.  We want to see intra-week fluctuations but not random fluctuations.  A 2014 block estimate would smooth things out more than necessary.

Doing 144-blocks (representing 6x24 hours, but more often a 12-hour estimate) is too small, producing a messy a graph of random fluctuations: https://i.imgur.com/unGHI.png

[bitcoinBull]

There are some other threads which chart difficulty against price.

In $/BTC—Difficulty Correlation chodpaba provided this graph:

And in Re: Will Bitcoin ever drop down below 2 USD? Enky1974 provided:

Also, back in November tcatm of bitcoincharts.com posted Price vs. Difficulty Graphs but the image files are missing now.  The price data I use for my charts is from bitcoincharts.com.

See http://bitcoin.sipa.be for good difficulty charts.

[acamus]

There are some other threads which chart difficulty against price.

In $/BTC—Difficulty Correlation chodpaba provided this graph:

http://chart.apis.google.com/chart?chxr=0,0,77000|1,0,1.1&chxt=y,x&chs=400x400&cht=s&chco=F1F12F|989866&chds=-15,77000,0,1.1&chd=t:917.831,1318.67,1378.028,2149.022,3091.737,4536.354,6866.899,8078.195,12252,14484.162,16307.421,18437.644,22012.381,25997.88,36459.887,55589.518,76192.619|0.064,0.061,0.06,0.062,0.095,0.099,0.19,0.243,0.23,0.23,0.239,0.267,0.3,0.4,0.417,0.897,1.04&chtt=%24%2FBTC%E2%80%94Difficulty+Correlation

And in Re: Will Bitcoin ever drop down below 2 USD? Enky1974 provided:

https://i.imgur.com/im3i4.jpg

Also, back in November tcatm of bitcoincharts.com posted Price vs. Difficulty Graphs but the image files are missing now.  The price data I use for my charts is from bitcoincharts.com.

See http://bitcoin.sipa.be for good difficulty charts.

man i really like that second graph, wonder if Enky1974 would be nice enough to update it for us. Also, it looks like log of price may reveal some spikes in the frequency domain. Anyone do fourier analysis on this beast yet? Nonlinear time series analysis?

[k]

hi bitcoinBull,

great charts. Would it be better to use the actual difficulty for a particular 2016 blocks rather than the difficulty implied by the calculated hashing rate for 504 blocks?

Nah.  We're trying measure the size/power of the network at the time, so we want to take the average hash rate over as small a window as possible.  We want to see intra-week fluctuations but not random fluctuations.  A 2014 block estimate would smooth things out more than necessary.

Doing 144-blocks (representing 6x24 hours, but more often a 12-hour estimate) is too small, producing a messy a graph of random fluctuations: https://i.imgur.com/unGHI.png

I was thinking to use the actual difficulty from each 2016 blocks (not a moving average of 2016 blocks, the actual difficulty) because that's the difficulty when people are doing the mining - the difficult is not constantly adjusting on a 504 block moving average . So if you are looking at the price/difficulty to decide the relative merits of buying over mining then would price/actual difficulty not be better?

[smooth]

Take the time series and compute the correlation between D(t+x) and P(t) as a function of x.   Graph the results.

Anybody going to do this?  If not, please send me the time series or a script to extract it from the block chain and I will post the results.

[k]

hi bitcoinBull,

great charts. Would it be better to use the actual difficulty for a particular 2016 blocks rather than the difficulty implied by the calculated hashing rate for 504 blocks?

Nah.  We're trying measure the size/power of the network at the time, so we want to take the average hash rate over as small a window as possible.  We want to see intra-week fluctuations but not random fluctuations.  A 2014 block estimate would smooth things out more than necessary.

Doing 144-blocks (representing 6x24 hours, but more often a 12-hour estimate) is too small, producing a messy a graph of random fluctuations: https://i.imgur.com/unGHI.png

I was thinking to use the actual difficulty from each 2016 blocks (not a moving average of 2016 blocks, the actual difficulty) because that's the difficulty when people are doing the mining - the difficult is not constantly adjusting on a 504 block moving average . So if you are looking at the price/difficulty to decide the relative merits of buying over mining then would price/actual difficulty not be better?

I went ahead and did what I was asking you myself. Hope you don't mind.



Plotted the daily average price from MtGox(*10^5) over the actual difficulty on that day (rather than the implied difficulty based on the hash rate). In fact it doesn't make that much difference. You can see the latest difficulty change has had an impact, making mining less attractive as would be expected.

[bitcoinBull]

hi bitcoinBull,

great charts. Would it be better to use the actual difficulty for a particular 2016 blocks rather than the difficulty implied by the calculated hashing rate for 504 blocks?

Nah.  We're trying measure the size/power of the network at the time, so we want to take the average hash rate over as small a window as possible.  We want to see intra-week fluctuations but not random fluctuations.  A 2014 block estimate would smooth things out more than necessary.

Doing 144-blocks (representing 6x24 hours, but more often a 12-hour estimate) is too small, producing a messy a graph of random fluctuations: https://i.imgur.com/unGHI.png

I was thinking to use the actual difficulty from each 2016 blocks (not a moving average of 2016 blocks, the actual difficulty) because that's the difficulty when people are doing the mining - the difficult is not constantly adjusting on a 504 block moving average . So if you are looking at the price/difficulty to decide the relative merits of buying over mining then would price/actual difficulty not be better?

Ah, good point.

Thanks for the graph!  It actually doesn't look much different

On the left up until October the ratio drops in about five steps.  I wonder if those are difficulty readjustments upward.

Could plot this as a shadow line against the ratio using estimated difficulty (or vise-versa).

[smooth]

Take the time series and compute the correlation between D(t+x) and P(t) as a function of x.   Graph the results.

Anybody going to do this?  If not, please send me the time series or a script to extract it from the block chain and I will post the results.

[k]

Take the time series and compute the correlation between D(t+x) and P(t) as a function of x.   Graph the results.

Anybody going to do this?  If not, please send me the time series or a script to extract it from the block chain and I will post the results.

Hi smooth,

i put the data I have here
https://spreadsheets0.google.com/spreadsheet/pub?key=0Arbegqx_2nKadEl5czMxZWZqazlFUzBiQkkzMFYtVEE&gid=0

Price data is available from MtGox in csv form and difficulty etc. from http://blockexplorer.com/q/nethash you can use /number at the end of the blockexplorer query to change the block size, e.g. http://blockexplorer.com/q/nethash/2016 will give data for ever 2016 blocks)

 Edit:

sorry I said above that I got price data from MtGox. In fact I got MtGox from here http://bitcoincharts.com/about/markets-api/

[acamus]

oh man, this is pretty great. time to break out TISEAN!

[Findeton]

This post is very interesting.

[Enky1974]

man i really like that second graph, wonder if Enky1974 would be nice enough to update it for us. Also, it looks like log of price may reveal some spikes in the frequency domain. Anyone do fourier analysis on this beast yet? Nonlinear time series analysis?

I'll do it, give me time:)
That chart is a non linear estimation of price/network speed using an exponential growth model.
I tried also fourier, and even ARIMA model with statistica 7 by statsoft.com, analyzing detrended daily data, there is a vague 12 days cycle not so strong,nothing exceptional apart a 90 days cycle on volume, every 3 months there is a new wave of investors buying, it seems so.

[smooth]

The 12 day cycle is probably due to the difficulty adjustments.  They're supposed to be 14 days, but when the network is growing they happen faster.

I haven't had time to do any real work with the data yet, unfortunately.

[Enky1974]

done it

to justify 17$ with this model we need a network speed of 6 terahash/Sec, is it overpriced?

[Findeton]

done it

to justify 17$ with this model we need a network speed of 6 terahash/Sec, is it overpriced?

Yeah it may well be. Too much publicity may get things out of control for a time, but that's good because it will test the system. Perhaps we'll see a market correction down to about $4-6, perhaps we'll get up to 25 or more before a correction takes place. Perhaps there's no need to correct the actual market (but I highly doubt it).

In my opinion, the best thing that can happen right now is for price to stand still at $17 until we get more TH/s.

[Enky1974]

Perhaps we'll see a market correction down to about $4-6, perhaps we'll get up to 25 or more before a correction takes place. Perhaps there's no need to correct the actual market (but I highly doubt it).

As you can see, a drop now at 4$ at current networkspeed would be a big outlier, unlikely that it will happen. It might drop to 8$ , a fast drop under 8$, without a network shrink, would be a great buying opportunity.