Topic: Monte Carlo Spreadsheet Comparing Buy vs Mine BTC (Read 572 times)

@jabowery - Thanks for doing this!

What aspect of this is using Monte Carlo simulations?

I had found the same spreadsheet and made my own modifications as well, and would be interested in seeing Monte Carlo projections of, Price, Difficulty and Miner cost.

I have been using a difficulty increase of 4.5% per month, since that seems to be the historical average (excluding recent China events). But I would really like to run Monte Carlo analysis to determine the likelihood of certain ROIs (positive or negative)

 Depending upon the gear and circumstances  I can have 100% certain profit with mining.

Buy and hodl never offers that.

Recent increases in advancement in technology have resulted in the number of transistors per circuit doubling in more than two years. I think further increases in the advancement of technology will continue to slow.

As new generations of miners become less frequent, the ILC will allow the latest generation of miners to be produced more efficiently (over time), but this does not mean that miners will be sold by manufacturers for less. If the market will bear a high price when the cost to manufacture declines, the manufacturer will continue selling at a high price. The ILC plus the decline of Moore's Law means that producing miners will become more profitable in the future, not that buying miners will be more profitable.

The above suggests that the rate at which difficulty increases will decline moving forward. This does not answer the question of if it makes more sense to buy bitcoin or buy bitcoin miners because it does not predict the rate at which the price of bitcoin will increase.

S
Do you also believe the industrial learning curve is "bullshit" because that's what I built my argument on, not Moore's Law.

Pointing out that an argument built on a fallacy is pointless.

I'm not sure what implications you draw from that for the future that differ from my implication that we're entering into an era when industrial learning curve is more likely to dominate capital cost declines in hash rates.  Are you simply saying I shouldn't have mentioned Moore's Law?

Moore's "Law" is bullshit.
In all cases it's a tiny sub-section of a very large curve that doesn't follow the theory.

The two curves I see are, in order of magnitude of importance:

1) Moore's Law
2) Industrial Learning Curve

Moore's Law capital cost decrease per hash is proportional to inverse exponential with time.
The ILC capital cost decrease per hash is proportional to inverse log units manufactured.

There is good reason to believe (knock on wood) that Moore's Law is coming to an end -- and not just because speed of light constrains the radius of control in CPU clock speeds (as has been the case for almost a decade now).  Hash ASICs are immune to the radius of control problem but not the absolute limit on feature dimensions as vertical features are now the last resort in the 7nm and below fabs.

As a general rule, if you expect the difficulty to increase faster than the price of bitcoin, it will make more sense to buy bitcoin, and if you believe the difficulty will increase slower than the price of bitcoin, it will make more sense to buy mining equipment. This assumes the market is 100% efficient, which may not be a valid assumption. The timeframes are also over the estimated useful period the miner will last.

When you buy bitcoin, you are long bitcoin and that is your only position. When you are mining, you are both long bitcoin, and are short difficulty.

The above rule is valid even if you can find a miner for sale selling at below-market rates (and there are no caviots), as if this is the case, you should purchase the miner, and subsequently try to sell the miner at market rates quickly (you can mine with the miner while you have it listed for sale).

The only reason why the above rule may not be valid is if you assume that tx fees will skyrocket, in which case it may be acceptable for difficulty to increase somewhat more than price.

You can use a spreadsheet to calculate if you are paying an appropriate price for a particular miner, given a particular set of assumptions.

It is hard to make use of that particular stat because a miner mining bitcoin in 2011 (likely a GPU), would be useless mining bitcoin today, and due to advancements in GPU technology, it would probably be useless in mining most altcoins today. The cost of both mining and bitcoin were very close to zero in 2011, which changed the decisions for many people. There were also major advancements in technology, such as the ASIC for bitcoin mining that is not going to be repeated in the next 10 years.

Absent efficiency increases, difficulty should increase as additional miners are manufactured by mining manufacturers. In general, manufactures will create the most efficient miners, while as difficulty increases the least efficient miners will be taken offline as they become unprofitable to operate (the latter may not always be the case as operating costs vary throughout the world).

you are missing the factor of efficiency to get profitably

so 1,533,333,233%

needs to be divided by a large number

250 watts would do .5gh
250 watts does 10 th

so
2x = 1 gh
20x = 10gh
200x = 100gh
2000x = 1 th
20000x = 10th

so shrink the 1,533,333,233 by 20,000

that is the profitability number.
the price number does go up so that 1,533,333,233 fits the diff

Could you elaborate on the feedback loop whereby a change in difficulty influences BTC price? I can easily understand and describe how BTC price influences difficulty over time, but the reverse is hard for me to visualize. I know you stated it was low, but I am just not seeing it.

Ok, that's fair enough, but it's based on the assumption that difficulty only goes up if gears became more effienct, which isn't entirely true, the difficulty will go up as long as there is profit to be made by someone, somewhere, it's even worse as many miners don't make profits and keep on adding gears.


I agree.


By the way, I by no means think that difficulty will keep going up 100% every year, just the same as I don't think price will go up every year, I also don't claim that this model is accurate, but the general idea of difficulty doubling while btc price only goes up 50% isn't at all a bad model to use when doing the math before investing in mining.

What most people do is underestimate the difficulty and price effect on their ROI, and given the fact that once your gear is no longer profitable it will have little to no resale value, ROI should be the main focus, so the average joe goes to Whattomine (or even worse Nicehash that displays wrong income) and work with the numbers as they are, and they see 6 months ROI, compare that to any other business on planet earth and it's a no brainer, so they go all in, some even borrow money hopping to return it in 6 months.

So by applying a simple model like this one, if the calculator shows 100$ a month, you should assume that by next month it will be 5$ less as a result of the difficulty alone, and assuming the price isn't going to go anywhere, and then you have to compare that against buying bitcoin, because if you invest $600 in a gear that makes

$100 in month 1
$95   in month 2
$90   in month 3
$85   in month 4
$80   in month 5
$75   in month 6
------------------
** 5% increase in difficulty won't yield these numbers, but this is a rough estimate using numbers that are easy to work with.

At this stage, you only have $525 + all the time and efforts wasted on running the gear, and as the difficulty keeps going up, it will eventually catch you with your pants down and gets you to the point where you can't profitably mine with that gear, this assumes the difficulty gets you before the chip thermal paste or the PSU does. 

So then, for those who like the fiat related math, they should actually not use the 50% increase in price which OP projects, just assume that your income will drop by at least 5% a month, the price will not go in either direction and your gear will die in 12-16 months, if using these figures doesn't bring up an interesting result then you are likely better off buying the coin directly or saving your money under the pillow.

Common, that's too easy  
~500 J/GH from a video card compared to 30 J/TH, let's cut this from those numbers and we're  457,042% to 91,999 %.
I doubt we'll be able to get even an order of magnitude in gear efficiency in the coming years..


Hm....his scenario?    
I see 100k possible even this year but 1 million I think is just not doable without hyperinflation.


Of course, I'm with you on this especially since I know we're close to what we had in difficulty in November, so at least for a small period, it's clearly in a normal world the difficulty growth should outperform the price, only another chip shortage or some other major mining ban or god know what could prevent this. Although...we should get used to things that make no sense.

Considering those numbers of course we could go to 84T and 150k, since we had 14T and 10K just a while ago, but his scenario was going long-term, if he would have said 10% and 20% yeah, it was possible, 30% with 50% also, but no way 50 with 100%.

Alas since there is absolutely no linear relationship between price and difficulty, since there are many other factors that effect both, making linear comparisons as an argument seems to be a rather pointless argument.

Also, of point to note: the force of price on difficulty is relatively high.
The force of difficulty on price is relatively low.

Anything can happen in the future, I sincerely doubt that we will reach a consensus to touch dormant addresses, but even if that happened, the effect of burning or reusing 1M coins won't change the fact of "diminishing returns".

I am not making this up, it's in the chart:

Bitcoin went up from low to high as follows:

55617% (2010-2013)
12700%  (2014-2017)
1700%   (2018-2021)

For "diminishing returns" theory to be invalidated bitcoin will need to up with at least one of those cycles, meaning, it has to go to $16,000,000 this cycle to stay in line with the 2010-2013 cycle or at least goes to 400k to keep up with 2014-2017 cycle, the first one is out of the question, I hope that nobody expects bitcoin to go to 16M by the end of this cycle, you might debate the 400k is doable but I would bet the house against it, and then even it magically happened, for the next cycle it needs to be above 10M, I honestly don't think that its remotely possible by any means.

So as it stands right now based on the facts that we have before us, the price growth will slow down, the difficulty will also slow down but most likely go up, mining profitability, in the long run, is going to go down, I don't present to know the future, but this has been the case for 10 years and I see no reason as to why it has to change.



-----------------------


Ok, if the period I picked didn't work for you, how about 10 years' worth of data?  

10 years ago difficulty was 1.5M and the price was $14, 10 years later price is $64,000 and difficulty is 23T, to put that in number digits it's  23,000,000,000,000 now vs 1,500,000 then.

Price went up by 457,042% while difficulty went up by 1,533,333,233% , and thus difficulty outperformed price by 335,390%, you can make of that what you want.


No one, but I should ask you the same question, what will take bitcoin price 50x in the next 10 years?

The price peak was 64k and difficulty was 23T, right? so OP is talking about the difficulty of 46T and price of about 100k, so I think that is possible? very possible, and then in 2023, difficulty sits at 84T and price at 150k? also possible, do I think this can go on forever? of course not and I have said that I agree with your point, but because I know that bitcoin won't be putting 50% up every year, this difficulty doesn't have to put in 100% for OP's projection to be accurate.

If you look at the past data in the right way, I believe these numbers have actually been the case or very close, you can't look at 3k vs 60k and say oh well price went 2000%, the 60k was so brief, even the 19k and the 1.5k before it, you need to work with something like the average yearly price by taking the daily close of all candles within that year and divide it by the number of days in the year.

For example, the average yearly price for the 2017 peak was 2.8k and not 19k, the average of the previous year was 0.5k (( I used 365 daily moving average on trading view to get that), so realistically speaking, the gains of 2017 was merely 460%, you need to compare that against the average difficulty (can't find an easy way to get but it's doable if someone is willing to).

Of course, this can't go on forever, both price and difficulty growth will have to slow down, I can't tell for sure if it's going to be 50% vs 100% or any other ratio, but in the long run, the difficulty will outperform "profitability" assuming price here is somehow irrelevant when another halving hit the place because once the rewards are halved for the next 4 years, the difficulty doesn't need to go up by 100% to keep OP's numbers in check.

Yes, I need to add a conditional in the spreadsheet to check for that as it _is_ too pessimistic in that regard now.  It will be interesting to see how much that impacts the Monte Carlo average for mining:

UPDATE

Mean:
0.2
Range:
-0.17 – 1.57

So it is quite a bit better, but still not competitive with buy and hold BTC under the other distributions given to the Monte Carlo sampling.

If we pick random dates yeah, of course, we can do it but if we spread that period with 3 months in each direction what will happen?
You picked a peak in price and if we move two months the price goes down to 1/3 , the hashrate simply didn't had time at that point to adjust with the ever-growing price, remember that at the start of the year the price was 1k, dif at 0.3? So to December, it went 20x times in price and 6 times in difficulty.


I've never said it will not go up, but that 50% vs 100% won't last forever on multiple time plots.
One year, of course, one other freak accident yeah, maybe but difficulty going 1000 times up with the price only 57 times? A 1/20 compared to what we have now? Something like the current difficulty at 1.5k per BTC?
Nope, I made a lot of shitty predictions and got a long wrong but this one I can bet a lot won't happen.


Of course, but who will produce 1000x of the current gear in the next 10 years?


Yeah, that was exactly my theory when I said the difficulty will double by last December compared to the price at that date!
I had the exact prediction like this, the difficulty will outpace the price by at least a factor of 2, how did it go, I said the same thing, people will buy miners like mad, and ....it didn't happen  

If we pick a time not affected by the whole China mess we need 40k and 180Th/s. In three years 135k and 1440Th/s, do you see that coming?

 

Maybe maybe not.

A lot depends on the un crack ability of wallets.

And if 'lost' coins are in the millions.

A simple rule to free lost coins is. 40 years of zero withdrawls  from and address means coins go on forfeit list.

At year 50 they go back to rewards.

You have time to move them to a fresh address. If you do not by by.

This keeps us at 21 mill max but allows a rewards refresh for miners.

We are in year 12 but I do expect to hear this idea more and more and more as rewards shrink and stale addresses grow.

It does make sense, it makes sense for the price to go down 50% and difficulty to double, the difficulty will only go up as it has done for a decade, ignoring the china ban and what happened in the past a few weeks, you can easily look at the difficulty chart and see that it only goes up, it does cool down once in a while but noting like the price does.

on top of my head would be the 2017 peak, diff was 1.6 and price was 19.5k, a year later price was at 3.3k and difficulty was approaching 6T, that is -83% decline in price vs 275% gain for difficulty, you might think that was me cherry-picking periods, but if you spread those periods across the years, OP's projections won't sound as crazy as you may think.

I do agree with the bolded part you mention, but it isn't a direct coloration, you can't assume that if price doesn't go up so won't the difficulty, this theory has been proven to be wrong over the past decade, we have seen difficulty go up right after the halving and after major declines in price this is because the mining profitability has not gotten to it's max.

Think of miners as fishers who go to the lake to fish, you can't assume that because the number of fish is going down, less fishers will come to the lake, because for one you don't know how much fish will keep every fisher happy, and two, just because it costs you too much to get to the lake and you bought an expensive fishing rod so everyone is in the same boat, yes to you and to some other people if you can't bring home 10kg of fish a day you rather sit at home, to others, they will happy with a small fish once a week.

The same goes for mining, many people have access to free power or super cheap power sources, those can just keep adding inefficient cheap gears to infinity, they will squeeze every bit of the profit and cause the difficulty to go up regardless of the price.

You should also not forget the fact that not every miner is going to profit, many people mine at a loss because they don't know what they are doing, if every miner was a smart person who does the math, the difficulty wouldn't be going up as it does, but remember that many people paid 15k for an S19 pro, they will most likely lose money in the long run but they will keep the difficulty up, and once those go broke other losers will join along and cause the difficulty to go up, and this is where the confusion comes in, assuming that all miners are making profit leads you to the idea that " if profits aren't great, the difficulty will go down or at least not go up", but that isn't the case, sadly.

I personally know a few folks who mine at a loss as I type, their thesis is that bitcoin will go up in price so they will just hodl the mined bitcoin and sell it when it hits 100k or 200k or whatever everyone's target is, I am sure we all know a person or two that does that.

One thing to mention is that price decline or price not going up does slow down the difficulty growth, but it does not stop it, on the other hand, the larger the difficulty is the harder for it to go up, and this has been my theory forever, it's why in our past predictions I have always mentioned that doubling the hashrate by year-end was not possible even if the China ban didn't occur, sometimes prices go up way to fast, the difficulty can't keep up with it, it's a simple fact, but in the long run, the difficulty will go up and price won't keep up with it.

I should also mention that the increase in price will also slow down, "diminishing returns" is real, the evidence is all over the chart, those who think bitcoin will go above 100k and will stay above it forever or even for a few months will be very disappointed.


 

try
 December 2017 price   19.9k     diff 1.87 t
 October    2018 price    6.4k      diff 7.45 t

So the diff went way up and the price dropped way down.

Ok,
July 2019 Difficulty 8T, price 11k
January 2021 Difficulty 18T, price 30k.
Growth in difficulty 225%, growth in price 300%

July 2019 Difficulty 8T, price 11k
July 2021 Difficulty 15T, price 32k
Growth in difficulty 90%, growth in price 300%

January 2019 Difficulty 5T, price 3k
January 2021 Difficulty 18T, price 30k.
Growth in difficulty 350%, growth in price 1000%.

So again, when did the difficulty growth outpaced the price 100% to 50%?

I extensively modified and expanded a minimalist such spreadsheet from 2017.

I made a video explaining things in some detail, including why the numbers you see are so conservative.

The Monte Carlo search goes through a wide variety of scenarios including some very optimistic ones.  If you don't like the inputs to the model, change them to suit yourself.


Look at what happened from July 2019 to January 2021

Really curious where you got that from, but other than that...what's with those numbers?
You put the difficulty at the maximum at 25T and the price at 17k, and you assumed for the miner a 75% uptime?
Whoever plans to do such a thing will probably not last all the 365 days of a period.

But, that aside, it makes no sense for bitcoin price to grow 50% year while the difficulty to double within the same period and this continuously over a long period of time, the moment an S19 miner with 3 cents power doesn't turn on profit who will add one more time the hashpower?
According to that in 10 years the difficulty would go up by 1E3, even with an increase in hardware efficiency of 5 times you would still need all the power in the world to feed that hashrate. Not going to happen!

For a better comparison, you should look at past data, not random projections.

I found an old (2017) Buy vs Mine spreadsheet, fleshed it out a bit and populated it with variable ranges in a Monte Carlo add-on (Add-ons -> Causal Scenarios -> Launch).

Here's a video explaining the spreadsheet and rational for the numbers.

https://i.imgur.com/WCIh46a.png

Maybe I have a bad model and/or maybe I have the wrong distributions of uncertainty among the input variables, but it seems to me the easiest and most profitable thing to do is just plow your money into BTC -- by far.