I'm going to disagree. The most reasonable reason is variance. The hashpower is unchanged, but the luck is a little bad for a few hours. The charts calculate hashpower based on blocks found, not based on actual hashpower. A decline in luck for AM can easily correspond with an increase in luck for a major pool.
Now, for the last time, can we stop discussing why hashpower changed over any period shorter than 3 days? If you think hashpower changed, ask yourself what time period the change is over. If it's less than 3 days, the MOST likely explanation is variance. If it's greater than 3 days, then you can start speculating about hardware being moved, planned slowdown before difficulty reset, etc.
It feels like we have this discussion every 7-10 pages in this thread.
Variance.
Again, variance.
Again, variance.
I think that attributing this to variance is a bit of a stretch. Something is going on, we just don't know what. With that said, I'm inclined to believe there is a valid reason behind the drop (hardware sales, etc.)
The probability of finding a block per hour based on a given difficulty is easy to figure out. Conversely, the probability of not finding a given block per hour (or any timeframe) is easy to figure as well. Therefore, determining hashpower over 24 or 48 hours should be fairly accurate. A lot of variance would have to happen for 24 or 48 samples to be incorrect.
Over 48 hours, AM is hashing at ~18.75 th/s. Using a larger sample size takes us before the dip, and the numbers are still lower than expected. 120 hours works out to 28.84, and that average is weighted heavily in the first couple of days. Tomorrow (and the next day) will provide us with good insight as to current hashpower over the past 5 days
Regardless, it's fun to talk about and look at the data.
