The difficulty adjustment, and the ten-minute target

The adjustment is bounded. Difficulty may at most quadruple in a single retarget, and it may at most fall to one quarter of its previous value. The cap exists because the timestamps the rule reads are not trustworthy in the way the rest of the header is: they are asserted by miners, and the protocol only constrains them loosely, requiring each to exceed the median of the previous eleven and to stay within a window of the receiving node's own clock.

Without a cap, a miner who controlled a majority of hash rate could mine a retarget period's blocks with manipulated timestamps, drive the computed ratio to an extreme, and produce a difficulty so low that subsequent blocks became trivial to produce. The four-times ceiling and the one-quarter floor bound how much damage a single period of dishonest timestamps can do, at the cost of slowing the network's response to a genuine, very large change in hash rate.

The caps are rarely reached. In the ordinary course of events the adjustment is a few per cent in either direction, because hash rate changes gradually and the two-week sample is long. When they are reached, it is usually a sign of something unusual: a large mining operation going offline, a regulatory disruption in a major mining region, or the network recovering from a period in which blocks were found far faster than intended.

The ten-minute interval is not a promise about confirmation speed. Blocks are found at random intervals around that mean, so a wait of twenty minutes between blocks is unremarkable and a pair of blocks found seconds apart is equally normal. What the target provides is a predictable rate of issuance and a predictable rate of confirmation over any long stretch, which is what lets the supply schedule be stated decades in advance.

It also paces the propagation of blocks. Ten minutes is generous relative to the time it takes a block to reach most of the network, so miners are usually building on the same tip rather than racing on competing branches. Shorten the interval and stale blocks become common, wasting work and weakening the incentive to mine honestly. Lengthen it and confirmations become slow enough to be impractical for ordinary payments.

The adjustment is therefore the mechanism that keeps the supply schedule honest. Without it, a rising hash rate would compress the interval and issue bitcoin faster than the schedule intends; a falling hash rate would stretch it and issue slower. The supply schedule page shows the issuance the rule is protecting, and the hashrate page explains the quantity the adjustment is responding to. The proof of work page covers the puzzle the difficulty parameterises.