Bitcoin Miners Fried in Game of Chicken | Hacker News Reader

Bitcoin Miners Fried in Game of Chicken

bloomberg.com

43 pointsadventured67 comments

It's a little bit simpler than the psychological drama described.

If your marginal/flexible operating costs (i.e. electricity, temp staff, etc) are higher than your expected revenue, then you shut off your miners temporarily. It's a simple rational decision: if you want bitcoins, they're cheaper to buy on an exchange.

If your total operating operating costs (building rent, etc) are higher than your expected revenue, then you lose money due to paying down liabilities. You might try to trim long-term operating costs (laying off staff, move to smaller building, move to place with cheaper electricity), or you might try to exit (sell mining hardware).

There's very little psychological drama in all this: the market for hashing power is very predictable, and it will adjust over time to provide narrow profit margins to the most efficient operations. The profitable periods were only ever going to be temporary aberrations in response to cryptocoin price fluctuations.

If your operation is not efficient, your own balance sheet will tell you when it's time to pack up.

"and it will adjust over time to provide narrow profit margins to the most efficient operations"

Unlike actual mining operations where costs are dominated by sunk capital costs. They can stay hugely irrational: once you've invested a few billion in a highly automated mine those billions are gone and you keep the mine operating even if the market collapses since your marginal + fixed costs are still lower than the price.

Or worse: sometimes a mine operates in the red since that's cheaper than the cost of decommissioning it...

The market for hashing power is only as predictable as the price of bitcoin.

(it's really the aggregate profitability of mining all cybercurrencies that have competitive rewards, but the price of bitcoin is close enough)

This is true, but only in the short term.

When prices go up, mining becomes very profitable. Given a little time (measured in hour/days/weeks or a couple months at most) people build out mining operations, and the mining returns adjust to the previous baseline. When prices go down, it's largely the same in reverse.

All that is to say: if one were to build out mining operations in expectation of a future increase in coin price, be forewarned that your market advantage will be short-lived and will only last as long as it takes others to fill in the gap.

Bitcoin is approaching the next difficulty retarget (in approximately 10 hours) and the difficulty will increase by around 10%.

If bitcoin price collapses to say ~$3000 we could see a significant portion of miners give up. This would increase the block time, which in turn increases the adjustment period (which happens every 2016 blocks, ~2 weeks at 10min/block).

An extended period of time with very long block times could cause the price to collapse even further, causing even more miners to leave...

I think there will be some determined miners willing to mine at a huge loss (due to having large bitcoin reserves) so the difficulty will eventually drop down, but there could be months of barely moving bitcoin which on top of the huge crash in the market right now could kill off bitcoin for good

I can only hope.

I don't care about Bitcoin, but they've been preventing me from finishing this system build. If it had been useful, I wouldn't mind so much.

This chicken game won't do anything to dampen GPU prices. All Bitcoin mining is done on ASICs.

Bitcoin not so much, a drop in Ethereum miners etc will have an effect.

Blockchain currencies have correlated price movements. Bitcoin rose and so did the others; bitcoin is falling and so are the others.

Other asset classes show similar behaviour. When the share market dives, the individual fundamentals of companies suddenly have much less import. During a surge, ditto.

Why not blame the companies that makes the GPUs?

Because GPU manufacturers don't want to play chicken with Eth prices either. If they double manufacturing capacity and the market collapsed they would be screwed.

So the customer is responsible for metering their usage of said products?

Where is this assumption coming from that there is even a problem? Demand goes up, prices go up. Sorry you can't buy a GPU as cheap as you could a few months ago, that's life.

I can't get the GPU I need. That's a problem for me.

You can argue that it's also the market working correctly, but if you drop the 'market' abstraction and look at what people are actually doing, they're burning a lot of power and resources for no actual useful purpose and causing minor problems like this in the process.

At best, this demonstrates a coordination problem. If we're so bad at coordinating that we can't build a functional economy even with the minimal central control implied by fractional reserve banking and central banks, but have to do stupidly inefficient things instead of having even a single trusted party, then humanity sucks rather more than expected.

At worst, well. The above isn't actually true. The existing, fiat-based economy works just fine, and Bitcoin doesn't. It's inherently deflationary, it's a prototype -- all the things so many people have already mentioned. So why are we doing it?

> I can't get the GPU I need

You mean you can't get the GPU you want at a price that fits your expectation.

> then humanity sucks rather more than expected

Interestingly I think it may be beneficial to adjust your expectations in this area too.

> The existing, fiat-based economy works just fine, and Bitcoin doesn't.

It works fine for you. Also bitcoin isn't the reason GPU prices are high, it is other cryptocurrencies and interestingly they actually work because they aren't being intentionally destroyed from the inside out.

> It's inherently deflationary

Doesn't matter, and guess what, when people have a choice in their currency, they aren't going to choose something that inflates.

> it's a prototype

According to who and what definition?

Bitcoin isn't. Bitcoin mining on GPUs has been dead for a long time.

The more likely cause is that the GPUs you want are ending up in deep learning projects or some niche altcoin miners.

That said,altcoins have always tended to go up and down together with bitcoin

And people who drive cars tend to experience sudden deaths. It's not the driving that kills people, it's the car crashes that do.

What I'm trying to say is that this is just correlation not causation. It's not Bitcoin going up that rises your GPU prices, it's Ethereum and the fact that it uses GPU-based mining. If no cryptocurrency used GPU-mining, then your GPU prices wouldn't rise, even if the the prices of cryptocurrencies went up by 100x .

Do you see how correlation is different than causation?

It's not Bitcoin doing that. TheVerge (or whoever told you it's Bitcoin's fault) are idiots for calling it the "Bitcoin mania" recently as the reason for your high GPU costs. Bitcoin has nothing to do with GPU prices as no miner uses GPUs. It's Ethereum and other GPU-mining cryptocurrencies.

Keeping an eye on Craigslist for a good GTX deal.

When miners give up, difficulty decreases, miners come back in. Checks and balances.

If difficulty decreases substantially, it may convince some to re-enter, but no where near as many as exited; otherwise they’ve wouldnt have exited? Still a huge net loss.

Also, more of a natural equilibrium than checks and balances.

I discussed that in my comment... But difficulty doesn't adjust fluidly, it's every 2016 blocks. A sudden decrease in hash power causes a sudden increase in block times, which means an increase in the time until the next adjustment

If it is true that some Chinese miners are using Bitcoin to funnel money out of China, then I could also see them continue mining at a loss.

Let's say half the miners stop mining. This would make it twice as likely that the remaining miners will mine each new block, right? If so, then this acts a bit like a price doubling from their perspective which would counteract the disincentives you mention. Perhaps a new equilibrium will emerge that's not quite so extreme because of this effect.

> This would make it twice as likely that the remaining miners will mine each new block, right?

Not immediately. The "difficulty" of each block adjusts every 2016 blocks, the algorithm takes the average time between blocks and changes the difficulty accordingly so that the expected time becomes 10 minutes.

If half the miners stop mining, it will take around a month to mine 2016 blocks to trigger the difficulty adjustment, and that's if none of those miners stop mining before the adjustment happens. If the price keeps dropping and miners keep stopping, it could be 2 or 3 months before the adjustment happens

Yes, but even if the block time is longer, the probability that I will "win" the block goes up. Someone has to mine the block and if half of the competition goes away my chances go up. This incentive will counter the temptation to stop mining, dampening the effect you mention.

Miners don't compete directly to "win" blocks. If everyone except a single miner stopped mining right now, that last miner would mine 100% of the blocks. However, they'd still make exactly the same amount of bitcoins as if everyone else was still mining - until the next difficulty adjustment

I see. Thanks

Is there a good reason for the difficulty adjustment period being 2016 blocks? What would be the consequences of reducing this number?

I'm not sure what the reasoning is, but to be honest I don't think it's a good system. Block time averages have been about 9m30 since bitcoin's inception, so it's clearly not good enough to properly target 10min/block

When bitcoin cash forked they changed the period to 144 blocks (along with some other changes i believe?), so they have a much smoother difficulty graph.

I just cannot trust bloomberg on cryptocurrency reporting. They conveniently cut off the price graph (just below "The Big Dip") in Nov 28. Why? Because at the beginning of October 9 2017, the price would have been at 4500$ - then an all-time high, and yet below the current price! It would have destroyed their narrative.

Not sure how that would have destroyed the narrative? The narrative seems to be pretty irrelevant to whatever happened to the price before november? The hashrate is consistently climbing ignoring the actual USD price with very few dips.

You're completely ignoring the increase in difficulty since it was at $4500 last.

Yes, and I believe the actual mining breakeven is closer to $1,250 based on a tweet from Adam Back today.

If the price drops too precipitously, and too many miners drop off at once, bitcoin enters a death spiral as the hashrate drops, and it takes so long to confirm blocks that the difficulty adjustment downwards just never happens, freezing the whole network.

Can you elaborate on why the difficulty would not adjust downwards as transaction rates slow? That seems like quite a flaw in the design.

The difficulty would adjust down, but only when the next milestone is reached. The argument seems to be that if enough miners drop out, the next milestone might never be reached.

That seems unlikely to me.

Let's see how the analysis was derived:

> Note: Assumes China wholesale price ($0.06/KWh), using listed specifications for Antminer S9 (13.50 TH/s), adds 30% cooling & operational costs, assumes retail is 30% markup from wholesale hardware price, 52-week depreciation schedule (Bitmain offers 180-day warranty). No transaction fee or pool fee. Hash rate and computational difficulty as of Feb. 6, 2018.

The biggest sunk cost is on the hardware price. Assuming 30% markup that too in China on a China build product seems too high for me.

Anyone based in China can confirm the retail markup on Antminer S9?

If a miner does not make money then they wont mine, if there's no miners cos no one can make money... who verifies transactions then? Or am I just really stupid and don't understand this stuff...

A variable number of miners are competing for a fixed reward. So everytime a miner calls it quits, all other miners become a tiny bit more profitable.

Isn't the BTC / hash fixed until the difficulty changes? Like, for a given difficulty you should expect N hashes calculated until a block is mined, so if you kill off half the miners (in terms of hash rate) the time until block mined just becomes twice as long (so now there are half as many miners chasing half as much reward).

BTC Miners make most of their money from transaction fees not block rewards. Fewer blocks = reduced supply = increased costs to get into blocks = higher reward for the same hash power.

Ahhhh, this is good to know!

I see, thanks.

That's true in the general case, but it only happens after a difficulty drop. Until then, miners dropping out will raise the block time (again, just a tiny bit). For cryptocoins with more dynamic difficulty adjustment, this process happens a lot more quickly.

This worries me about BTC, because it may get to the point where for the drop in difficulty takes too long that it will no longer be profitable, in this case crypto currencies where the reward is higher and the increase in reward vs difficult increases makes it more profitable means more miners will ditch BTC and go to those... That means transaction verification times just keep raising and then BTC eventually becomes useless.

That's my understanding anyway...

As long as the currency maintains some value, there's an equilibrium at which it will be profitable for some to mine.

To paraphrase Yogi Barra, "nobody mines anymore, the market is too crowded"

Are there mining algos that also help to solve real world problems (say gene mapping)? It seems that you could leverage some of the gamification happening here to better some social facet of society.

A simple search would find several of these.

No. The problem in that suggestion is, who is going to provide the problems? It would be a central actor who can opt to pre-work on the problems and use those solutions to issue themselves "money".

No, it would simply require solving a function dependent on the previous solution. Preferably more complex than double SHA.

FoldingCoin is doing something like that, but I'm not involved enough to speak authoritatively about specifics.

Mining this stuff is a funny thing, that is a devisive feature even in the crypto space.

That being said, this isn’t the first time where mining has plummeted in profitability. Early last year similar articles started popping up and in fact GPUs on eBay we’re being sold at dirt cheap prices because crypto was “plummeting”.

I wouldn’t bet on crypto going anywhere, not in this crazy world we live in.

"Fried" as in

> If the price doesn't rise, then the average miner is set to lose $3 per week at current levels.

(Per week and rig I assume though.)

Props for the title.

This is not a very big problem for all the big name miners though, as they're all betting on prices rising soon. This has happened before, and will happen again until either everyone adapts some alternative to proof of work.

Everyone seems to just be mining altcoins and hodling for a rise.

According to this web site, all you need is 11th/s to not lose money, that's an investiment of $2.5k in mining hardware https://www.cryptocompare.com/mining/calculator/btc?HashingP...

How can any miner lose money unless prices reach ~$3k?

Edit: why am I being downvoted? Let me know if I'm mistaken.

The calculation you did on that website shows an annual profit of $413.25 which is still a lot less than the initial investment of $2.5k. At today's price ($ 6,976.14), you'd have to be mining for over 5 years before getting that investment back again. I'm not sure how you come to the conclusion "how can any miner lose money".

For $2.5k you can get 13th/s, that's $935 annually, that's a 40% gain annually. If you know of any investment that can give you a passive 40% yearly return, I'm all ears. And that's only if bitcoin doesn't go up. Again, I fail to see how this is a bad investment. Plus, with the latest all time high, most existing miners already paid for their hardware and are just all profits now.

https://www.cryptocompare.com/mining/calculator/btc?HashingP...

At ~$7,500 we're roughly back to mid-November price levels, so it seems like it would affect only newish entrants.

Looking forward to the first deep learning framework that puts all these energy-burning-machines to good use.

Miners use specialised hardware (ASIC - application-specific integrated circuit) to mine. They are only useful to calculate hashs efficiently.

is there no way to make 'good' use of them? whoever comes up with an idea might become a billionair.

Proof of work is intentionally useless to the miner to increase the cost of attacking the network. There are initiatives like curecoin and foldingcoin which integrate folding@home into their proof of work, but the output of that is still "useless" to the miner and relies on the folding@home team not manipulating the process, which becomes a bigger risk once hard rewards for folding are introduced.

Thousands of companies go bust every day in our "great" economy; retailers, startups, etc.. Now the price of crypto-currencies has dropped and it's called a "Game of Chicken".

There seems to be a tendency to laugh and write bad about miners, almost like bullying. Why is this? Mining is business as usual, nothing different than many other businesses. There is investment, risk, profit and loss.

Investment in mining hardware whose value is tied to a speculative asset without fundamental value beyond utility.

None of the current implementations can scale, so even their utility is questionable. There is a way to make a scalable trust-less digital currency by using a mutable leder, but none of the current coins that I know of are heading that way. The main difficulty will be how to distribute ledger space in a way that cannot be monopolized by the richest. That is a simple technical problem to solve though, because storage capacity will eventually surpass the ratio of richest to average wealth. Then you can allocate ledger space in exchange for an affordable deposit, redeemable when the balance is emptied.

Does anything have fundamental value beyond utility?

Reply on news.ycombinator.com