The recent update of Dogecoin to version 1.6 consisted of a change to stable block mining rewards and the adoptation of the DigiShield difficulty algorithm. The changes became effective as of block 145,000. Dogeconomist performed an analysis on block data before and after the update, provided by somegeekintn, to determine how this has affected Dogecoin mining.
Data and methods
The dataset consists of Dogecoin block 137,501 to block 152,500, 7,500 blocks before and 7,500 blocks after the changes became effective at block 145,000. The datafile contains information on each block’s difficulty and the time required to solve the block (derived from the block’s timestamp). As the average block target time is set to one minute, all observations are divided over 10-second intervals from zero to 120 seconds to solve. A paired sample (before/after) test is applied to all observed differences within these intervals. This is required to determine if differences can be attributed to chance, or the changed variable (the replacing of Dogecoin 1.5 by 1.6 at block 145,000). In all cases, the null hypothesis is that two compared subsets won’t show any change in the proportion of blocks in a certain time interval (other than some measurement error caused by chance).
The biggest change in Dogecoin version 1.6 was the new DigiShield difficulty algorithm: “this allows for rapid block retarget times, and rapid recovery from multipools suddenly hitting the network; meaning that our miners will no longer suffer from extremely long block times if a Multipool does decide to jump on board.” The following graph shows how this change seems to be effective, as the relation between block times and difficulty are strongly reduced.
Time to solve
The second question would be how Dogecoin 1.6 has affected the time required to solve a block. The following histogram shows the distribution of the number of blocks per amount of time required to solve the blocks.
It can be noted that the distribution represents a so-called log-normal distribution. This is due to the fact that a block normally cannot be solved in less than zero seconds, but there can be extremely long times required to solve a block. It can be observed that there are still many blocks that require a time to solve of more than 120 seconds. The latter, doesn’t seem to have changed much comparing Dogecoin 1.5 to 1.6. The biggest moves take place at the short end of the chart.
As stated, blocks normally won’t require less than zero seconds to solve. There is, however, no centralized determination of the block timestamp. It could therefore be possible to tamper with block timestamps in an attempt to manipulate the difficulty. The histogram below shows the number of blocks with an implied time to solve of less than zero seconds, in sequence, starting at block 137,750 with 250 blocks increments.
It can be seen that the number of blocks increases strongly after the update to version 1.6. Even though the pattern already reveals that this change isn’t due to chance alone, this is also confirmed by statistical testing. The number of blocks with less than zero seconds required to solve increases significantly after the update (p-value <1%). If the sample is split further in subsets of 3,750 blocks (half of both the 1.5 and 1.6 sample) it can be seen that there is also a significant decrease from block 137,500-141,250 to 141,250-145,000 (p-value <1%). In the second half there are no significant changes.
The decrease in the first half can, however, be caused by the volatile period following the announcement of Dogecoin 1.6. The timestamp of block 137,500 is 2014-03-12 14:21:05 UTC, which happens to be during a volatility peak as shown below. This could interfere with the consistency of the data. The volatility remained stable during the remainder of the sample. Despite these market conditions, the conclusion remains the same even if different sampling periods are taken into consideration; there is very strong evidence that Dogecoin 1.6 led to an increase of negative times to solve a block. This seems to reveal a group that started to tamper with block timestamps as of Dogecoin 1.6.
Update March 30, 2014: The first observable increase in negative times to solve takes place on the interval 146,251-146,500. This is about 20 hours after block 145,000. It can be shown that even though the variation in difficulty increases immediately in the following 1,250 blocks, there is no significant change in the number of blocks with a negative times to solve in the same subset. The variation of the difficulty actually drops a little on the interval 146,251-147,500, compared to the previous 1,250 blocks. The sample is big enough to establish a significant change in negative times to solve from blocks 145,000-146,250 to 146,251-147,500. This makes it seem like the change isn’t a natural one.