Speaking of that, I always wondered whether CryptoNote coins hashrate is any accurate. Monero's one jumps up and down from 7 mHash/s to 15 mHash/s within 24 hours (based on ChainRadar).
Obviously, hashrate is calculated as difficulty / block_time. And difficulty here is the number of hashes roughly required to find the last block. Isn't it better to evaluate CN hashrates not with this simple formula, but rather include a series based calculation?
Difficulty is calculated over a window of 720 blocks, so that already takes into account some period of time (12 hours). In reality it seems that the hash rate really does fluctuate a lot, even if the estimate is necessarily imperfect. This probably has something to do with time-of-day as it always seems to have been synchronized to a 24 hour cycle even on coins with slightly different parameters. Maybe computers that are idle at night in some geographic area and being used to mine, lower off-peak power costs, etc. I don't think it is possible to really know for sure.
I know about 720-block window. And it should also cut outliers, in theory. Maybe 720 blocks is not enough?
It indeed looks like there's 24 hour cycle, but I don't believe that it's due to certain PCs being idle (well, unless botnets again?). It can't constitute 50% of the network? And how does one even assess mining profitability if it fluctuates by the factor of two on a daily basis?
Well when the block time changes to 2 minutes, it will then be 24 hours, since we didn't change the 720 block window. That will be back to the original parameters as designed by Cryptonote/Bytecoin. Perhaps the behavior will be different then. We also have some research ongoing for better ways to adjust difficulty.