By now, bitcoin enthusiasts and professionals alike are aware of the tremendous growth of the bitcoin network so far this year. As millions of dollars (and bitcoin) are invested in increasingly powerful mining equipment, the cumulative hash rate of the network has continually blown by the periodic difficulty adjustments.
Earlier this year we explained the long-term implications of this, but the near term effects are being felt as well. Since that piece was published, the network has continued to grow, as has the growth rate, creating profound near-term effects. In particular, the average time between blocks has fallen to approximately five minutes on average, half the time bitcoin was originally programmed for.
Why this is happening
The bitcoin protocol is set to adjust the difficulty of finding a block such that the average discovery time is approximately 10 minutes on aggregate. An adjustment is made in the difficulty level every 2016 blocks (roughly two weeks if following the expected schedule) based on the average time required to find the previous 2016 blocks.
Since network speed has increased both dramatically and persistently so far in 2013, the difficulty adjustment has consistently lagged behind the actual hash rate, reducing the time between blocks. This has happened now for so long that the time between blocks has been reduced by 50% to the five-minute block time currently observed.
Greater inflation, faster confirmations
The implications of this are important to note, but should not be perceived as definitively positive or negative for bitcoin. For one, more bitcoin are being issued than initially planned. Since each block still carries 25 new BTC and time between blocks is cut in half, we’re at effectively double the expected monetary inflation rate. Remarkably, the market has largely withstood the inflationary pressure, as illustrated by our piece on recent low volatility on exchanges.
Additionally, the confirmation time required for the same level of confidence that a transaction is not fraudulent has also been cut in half. Confirmations are created when subsequent blocks are added to the block chain, indicating a high probability of the transaction being included on the main chain. Satoshi described this in his initial white paper (page 8), illustrating how the probability of a double-spend attack decreases with each additional block. It’s become commonly accepted practice to accept six blocks, as this would give someone with 10% of the total network <0.1% probability of success in an attack.
A common misconception is that transactional confidence is based on time – this is only partly true. The probability of a double-spend is reduced with each subsequent block confirmation, regardless of time between blocks, so long as the time distance between those blocks is sufficiently greater than the time required to propagate newly discovered blocks to the network. For a detailed statistical explanation of why that is, we recommend the following:
Analysis of hash rate-based double-spending by Meni Rosenfeld
A recent study from Microsoft Research showed that while the median time to receive a block is 6.5 seconds, after 40 seconds there are still 5% of nodes that have not yet received a block. This means that even five minute block times are still comfortably greater than required, but the comfort margin is thinning. You can see the real-life, non-impact of this by observing how there have not been any orphaned chains with more than a single block since March. If block propagation were an issue then there would be significantly longer orphaned chains as confusion persists regarding which blocks arrived first.
Given this reality, one might question the value of 10-minute planned block times, as many alt-coins already have. While there is certainly a case to be made for decreased confirmation periods, there are a number of reasons 10 minutes is unlikely to change.
In reality, five minutes is effectively the same as 10 minutes when waiting for a confirmation of an in-person transaction – either is well beyond the point of expected convenience. If you’re not prepared to accept a zero-confirmation transaction or off-chain equivalent, you should clear your schedule after a transaction occurs.
There is also the possibility that significantly larger block sizes could increase network propagation time. In that case, the margin between propagation period and confirmation period would decrease, potentially below levels of comfort. The study cited above from Microsoft Research also quantifies delay cost, or the time delay each kilobyte of block size causes in the dissemination of a block. As bitcoin’s popularity grows, maintaining a comfortable margin above propagation time may require longer average confirmation periods than it does currently.
Even if all else were equal, the incredible expense associated with updating a global network may simply outweigh the potential benefits. Ultimately, the reduced confirmation period should normalize on a longer-term time frame, but remains an issue that any bitcoin investor should be aware of.