Google Doc with exponential fit, lead time, and linear rate target

[jspielberg]

I went ahead and made a new Google Spreadsheet with the latest difficulty data in wolfram:

https://docs.google.com/spreadsheet/ccc?key=0Auya3iRE6az1dDc0UVgwMU52YVpTazVjSHByOGNiWHc&usp=sharing

This time I didn't make any changes to the exponential factor spit out by wolfram.  It seems little too optmistic, so I don't think it will be accurate for more than 1 or 2 retaigets.

Here is the wolfram link:
http://www.wolframalpha.com/input/?i=exponential+fit+%7B%7B13%2C+3651012%7D%2C%7B37%2C+4847647%7D%2C%7B96%2C+11187257%7D%2C+%7B120%2C+15605633%7D%2C%7B156%2C26162876%7D%2C%7B167%2C31256960%7D%2C%28189%2C50810340%7D%7D

Again... feel free to tweak the values.

Edit: Updated 7/13 with latest difficulty numbers: base = 2.83643 exponential=0.0142844
Edit: Updated 8/16 with latest difficulty numbers: base = 1.89041 exponential=0.0172202
Edit: Updated 8/31 with projected difficulty values: base = 1.05869 exponential=0.0207254 !!

===== Original =======
Just for fun while waiting for Avalon to get around to shipping B3 units, I created a google spreadsheet to project the estimated BTC generated for a given hashrate received at a particular time during the year.  Is it pretty easy to use... just put in your expected delivery date in the yellow box, and your hash rate in the green one.  Feel free to try it out!

https://docs.google.com/spreadsheet/ccc?key=0Auya3iRE6az1dG9fRkpXdFJtT1dNX0VCU1F0VFFUX3c&usp=sharing

To get the projected hashrate I used historical some data points starting in mid January through this last for difficulty adjustment (6/16/2013) and used and exponential fit in wolfram alpha.

http://www.wolframalpha.com/input/?i=exponential+fit+%7B28.19%2C2968775%7D%2C%7B40.88%2C3275465%7D%2C%7B65%2C4847647%7D%2C%7B87.9%2C6695826%7D%2C%7B112.11%2C8974296%7D%2C%7B124.59%2C10076293%7D%2C%7B137.2%2C11187257%7D%2C%7B161%2C15605633%7D%2C%7B172%2C19339258%7D


One caveat:
Wolfram alpha's best exponential curve seemed a bit optimistic on growth rate. I upped it a bit to what I think would be more conservative (but who knows?!).
These projections put the difficulty at 100M+ at the end of september, and 1B+ at the beginning of March 2014.

I am not sure how to allow you guys to make your own edits without saving... so if you want to edit, you will have to make a copy I'm afraid.

[suryc]

Interesting predictions and thank you.

Why did you choose to have difficulty changing every day when, in fact, it only changes every 2016 blocks (currently about 10 days).

[rammy2k2]

thats scary 

[TheSwede75]

thats scary 

Sure makes you think when it comes to ASIC profitability past Sept/Oct that's for damn sure..

[jspielberg]

Interesting predictions and thank you.

Why did you choose to have difficulty changing every day when, in fact, it only changes every 2016 blocks (currently about 10 days).

I did daily for a couple of reasons:

1) I had wolfram alpha calculate the exponential regression on a day time scale rather than block (though I did do a block spreadsheet but found it less interesting)
2) I didn't want to try figure out the early completion of the 2016 blocks (I'm lazy  )

[jspielberg]

Would y'all prefer if I made the document editable?
The risk is a griefer would mess around with the formulas (and break the spreadsheet -- or inserts something weird in it).

I guess it isn't that big a deal for everyone to make a copy if they want to change the values.

It is a shame they don't offer per-cell locking as I could just open up date and hashrate cells.

[dben428]

Thank you for this, very interesting.

I think a document like this is best left locked. Even without griefers, multiple people trying to edit the same couple of cells can be pretty annoying when you are trying to use it.

[odolvlobo]

Just for fun while waiting for Avalon to get around to shipping B3 units, I created a google spreadsheet to project the estimated BTC generated for a given hashrate received at a particular time during the year.  Is it pretty easy to use... just put in your expected delivery date in the yellow box, and your hash rate in the green one.  Feel free to try it out!

https://docs.google.com/spreadsheet/ccc?key=0Auya3iRE6az1dG9fRkpXdFJtT1dNX0VCU1F0VFFUX3c&usp=sharing

To get the projected hashrate I used historical some data points starting in mid January through this last for difficulty adjustment (6/16/2013) and used and exponential fit in wolfram alpha.
...
One caveat:
Wolfram alpha's best exponential curve seemed a bit optimistic on growth rate. I upped it a bit to what I think would be more conservative (but who knows?!).
These projections put the difficulty at 100M+ at the end of september, and 1B+ at the beginning of March 2014.

I am not sure how to allow you guys to make your own edits without saving... so if you want to edit, you will have to make a copy I'm afraid.

It is impossible to get an accurate prediction with just extrapolation. There are 4 more pieces of information you could add to make the predictions much more accurate:

1. How much hashing power is expected to ship over the next year or so (and when).
2. The difficulty rate of change at which an ASIC costs more than the amount it will ever mine.
3. The difficulty at which electricity costs make it unprofitable to mine with an ASIC.
4. The difficulty at which electricity costs make it unprofitable to mine with an GPU.

[yantis]

+1 Thanks +)

[jspielberg]

Correct... This is pure speculation with a little math thrown in to keep it somewhat sane.  The exponential phase we are in can't go on forever.  However I thought this had a good balance of optimism and pessimism.  It also tracks to the last 5 months which include the asic deployments by avalon and asicminer fairly closely. 

My theory is that the exponential growth phase requires asics to be deployed to continue the trend.  The amount of growth required won't be plausible with FPGA and GPU mining. At some point that growth won't be sustainable. This model doesn't try to predict that point.

[ScaryHash]

Interesting spreadsheet.

A couple of interesting points are that difficulty, network hash rate, AND price actually went down in March-April of 2012, and also in October of 2011.

While it is difficult to do statistics on something like that, one could assume because of the size of the network that a lot of people made some calculation that it was not profitable for them to mine, so they stopped, then at some point started back up again.

This is supported by the subsequent spike in difficulty in the months immediately following those months.

The same thing happened in December of 2012, there was an actual decrease in difficulty, followed by consistent increases in difficulty.

Thus, we can conclude that we will probably see huge jumps in difficulty (over 20%) per every 10 days over the next months.

Pretty interesting stuff.

[dwdoc]

I am using your spreadsheet to evaluate various WTS offers. Pretty interesting...

[mustyoshi]

400M by the end of the year you say?
2.8PH/s would be an interesting mark.

[superresistant]

We are the 6/26/2013 and you predicted 24,817,097 but it is only 19,339,258
So a difference of 28.32%

Thx for making us scare.

[lazydna]

We are the 6/26/2013 and you predicted 24,817,097 but it is only 19,339,258
So a difference of 28.32%

Thx for making us scare.

next change is 21mil so far.
it's just that bitcoin difficulty adjusts within 7-14 days.

he's not off by much. considering ASICminer randomly turned off their machines for a few days this cycle and avalon fucked up there batch2-3 shipments.

[superresistant]

We are the 6/26/2013 and you predicted 24,817,097 but it is only 19,339,258
So a difference of 28.32%

Thx for making us scare.

next change is 21mil so far.
it's just that bitcoin difficulty adjusts within 7-14 days.

he's not off by much. considering ASICminer randomly turned off their machines for a few days this cycle and avalon fucked up there batch2-3 shipments.

It's not a big difference for now but it increase with time.
Even taking 21mil that's 18.2% diff

[lazydna]

We are the 6/26/2013 and you predicted 24,817,097 but it is only 19,339,258
So a difference of 28.32%

Thx for making us scare.

next change is 21mil so far.
it's just that bitcoin difficulty adjusts within 7-14 days.

he's not off by much. considering ASICminer randomly turned off their machines for a few days this cycle and avalon fucked up there batch2-3 shipments.

It's not a big difference for now but it increase with time.
Even taking 21mil that's 18.2% diff

http://www.asicminercharts.com/

lost 17TH for 7 days.
i'm too tired to do the math. but that's over 10% of the total bitcoin network lost.
throw in those numbers and the margin of error will be much slimmer.

123.76 blocks missed.

[jspielberg]

In all fairness, I did state a caveat that I made the chart a bit more pessimistic than Wolfram was plotting.

Wolfram came up with a least squares best fit of 2026090*e^(0.0129097x)
Where x is the day in the year.  The spreadsheet uses 0.014825.

My exponent is 15% larger than what was calculated because I do firmly believe we are in the calm before the storm.

I tweaked the exponential factor because I knew the pace of Asic deployment was going to be pretty aggressive in the second half of the year.  I guess I see these charts as a sort of as a balance between the "worst case" and "it won't be too bad" scenarios.  That is why I shared it.  I figured if it were somewhat in the middle, it might be realistic.

I have actually moved onto an even more pessimistic projection for my planning purposes (0.01525) as I believe that to be even more accurate later (even if it is less accurate now).

[JstnPwll]

Well, we are currently 4,000,000 behind the estimate for today.  That's moderately encouraging, I suppose.

[lazydna]

well, asicminer predicts they'll push out another 800-1000TH by the end of this year.