Bitcoin Forum

Bitcoin's difficulty always lags a little behind the actual hashpower of the network. If the network's hashpower massively decreases right after the adjustment of the difficulty, it will take considerably longer to confirm transactions in the blockchain (more than ten minutes per block), which compromises bitcoin's liquidity. Consequently, it will also take longer than the estimated two weeks until the next 2016 blocks are found, i.e., until the difficulty level will be adjusted again. At this point, I consider some potential for bitcoin to slip into an economic downward spiral due to the following reasons.

The compromised liquidity might cause a drop of exchange rates as bitcoin transactions just take a little longer than usual, which makes bitcoin a little less useful (probably also in comparison to other cryptocurrencies if they succeed). In addition, some people might fear that the network freezes (which will be the case if the downward spiral keeps going) and sell their bitcoins, which levels down the exchange rates. A drop in exchange rates, however, can be considered a disincentive for miners to continue mining. Mining may be just not as attractive when rates are falling. So, some miners might stop mining or switch over to mining other cryptocurrencies which they deem more profitable (which is not as easy for people who have invested into FPGAs or ASICs, I know). Hence, the remaining miners will have to continue mining at the too high difficulty level even longer, and they still have to invest the same amount of time (i.e., energy) into mining than they did at the time when the difficulty was last adjusted. However, the more miners leave, the more gets the liquidity compromised, the more might the rates drop and the more miners might be discouraged from mining. If this happens quickly, bitcoin freezes before the 2016th block is found.

How might a massive drop in hashpower happen:

• Support attack: An attacker with immense hashpower enters the network for a period of 2016 blocks to support the network, i.e., to participate in mining. Just when the difficulty is adjusted accordingly higher, he leaves the network. (It's a hypthetical attack. There is no reward for the attacker, but he had to invest a lot to achieve this.)
  
• Legal prohibition: Many miners in a country or a region could stop mining all at once due to a legal prohibition. (The share of miners of a single country might be too small for a massive drop in hashpower. Also, this scenario does not seem very likely.)
  
• Dropping rates: Dropping exchange rates may be a disincentive for miners to continue mining, just as rising rates are an incentive for people to start mining. However, it doesn't look like many miners have stopped mining after the drop on April 11th (see http://bitcoindifficulty.com (http://bitcoindifficulty.com)). Nevertheless, if other cryptocurrencies become popular, they will also compete for miners, i.e., dropping rates may be a greater disincentive for miners than today.
  

I have written a blog post (http://www.robin-klose.de/technology/how-bitcoin-might-slip-into-an-economic-downward-spiral/), which covers this topic in greater detail. Even though it might be very unlikely that bitcoin slips into this kind of downward spiral, I also think that a system that claims to store value should protect itself better against this kind of fault. The difficulty would have to be adjusted just more dynamically (which may bring other problems like orphan blocks, I know).

How much potential do you consider for entering an economic downward spiral as described above? Do you think this is critical?

In theory yes but anything less than 50% of global hashing power would result in a negligible (in the long term) change.  At this point there is no low cost method you could "rent" that kind of hashing power so you are talking about an outside entity building a massive ASIC hashing farm.  If they are going to do that and build out to more than 50% of Bitcoin's hashing power they have more effective attacks anyways.  Even a 75% reduction in hashing power wouldn't be fatal to Bitcoin.  The adjustment period would be 8 weeks which is punitive and average block time would be 40 minutes.  Most likely for many transactions people would start accepting just two confirmations (80 minutes).  It would be a hit but nothing like a sustained 51% attack.

The difficulty period is a compromise.  The longer it is the slower the network reacts to REAL changes in hashing power.  The shorter it is and the more the network INCORRECTLY adjusts to variance instead of actual changes in hashing power.  It is a potential issue but IMHO a minor one.  For other coins it is more of an issue.  Those with low hashing power or those where botnets or cloud computing is effective give an attacker the ability to "rent" enough hashing power instead of building out new computing resources.  That makes a "difficulty pump" attack more cost effective.

I think you're too optimistic. If the average block time was 40 minutes, this would be a huge disincentive for miners. They would earn only a fourth of what they should earn, which might cause many of them to quit mining, which in turn would make the situation even worse.



That's basically right, but the mechanism could still be more sophisticated. For instance, you could adjust the difficulty every 504 blocks but still take the average hashing rate of the preceding 2016 blocks into account to overcome the variance problem. In addition, you could use the derivative of the 2016-block average hashing rate to extrapolate the new difficulty more accurately into the future.

I do not see any strong reason why the hashpower drop can be reason for compromised liquidity.

The time between blocks won't matter to the miner.  The miners probability of finding a block in a given amount of time (and therefor their profitability) will be exactly the same as it was before the reduction in hash power.  If that probability was an issue for the miner, they already would have shut down their mining rig before the reduction in hash power.  If they are still mining when the reduction happens, they have already accepted mining at the current difficulty.

Furthermore, seeing the reduction in hash power, the miners will know that profitability is about to increase with the next difficulty change.  This could act as an incentive to leave mining rigs running so as to take advantage of the difficulty drop the moment it occurs.

Good point, you're absolutely right. It's better to look at the individual miner than the network as a whole. But consider that other factors (e.g., a drop in exchange rates) may still make miners quit during that period. They might accept the difficulty at the first instant but change their minds later. They won't accept a difficulty level forever, even if it was initially reasonable to do so. Hence, if hashpower and exchange rates fall at the same time and if things go quickly enough, bitcoin may possibly enter into such a downward spiral of evermore leaving miners. Miners would need an immediate incentive to continue or start mining at times when hashpower drops. But bitcoin simply lacks such a feature.



Or you can put it also like this: If miners consider a new difficulty level too high, they (or at least some of them) won't accept it and will immediately quit mining.
But this makes bitcoin susceptible to the difficulty pump attack described above. The attacker (with immense hashpower) joins mining for a period of 2016 blocks. All other miners will be absolutely happy during that period as they will earn just the same amount of bitcoins which they would have earned if the attacker hadn't even been there. However, the network as a whole will run a lot faster during that period, so the next difficulty level will be accordingly higher. After the adjustment, miners will have to decide if they continue mining. At this point in time, the network already runs somewhat slower (> 10 minutes / block) since the attacker has left the network and it will run even slower with each miner who leaves.



There is no incentive to leave the mining rigs running when you don't earn an appropriate amount of coins for doing the work. A miner could just as well decide to switch the rigs off and wait out the difficulty drop. Why should a single miner decide to do the work for others? The only reason I see is that a miner wants to keep the network going as he is heavily invested into bitcoins.

Hashpower will only drop slightly if many GPU miners throw in the towel due to the ASIC craze, but once those are on the market, the hashrate and difficulty will skyrocket, which can also make bitcoin slip a bit based on lack of interest.

We'll just have to see what happens.

If an attacker was able to gain 50% hash power, they could quite possibly get upwards of an entire cycle's worth of coins.
First they put that 50% hash power into the network gradually raising the difficulty (and mining at the same time), when it adjusts the difficulty, they pull their power.
They wait the 4 weeks, then when the difficulty which is now probably 25% slower, they start mining again, and they have 25% advantage still. Which means they net a bunch coins again.
The trick would be achieving this attack for under half a million USD which would be about what they got in coins after doing this.

If an attacker has 50% of hashing power they can achieve more by simply taking all the coins as long as they can maintain the monpoly.   It is very simple to.  Just ignore all blocks created by any other miner.