Edit.. i mine on slush so i feel that the only reason they are disproportional is because they are recovering from a bad streak over the last couple days..
There's quite a few days in that graph
I was referring to the disproportional nature compared to other pools.
If all pools in the graph stay at the same hash rate, then in that graph of mine they should all basically have horizontal lines; they can vary up and down a bit, but in the large averages (wider lines) this should smooth out pretty well.
If all pools go up by 25%, then that wouldn't change - they would still all be horizontal.
So when you see one (or in this case, two) graph line(s) go up quite a bit while all the other ones go slightly down, then whoever's behind those lines are adjusting their hash rate disproportionately compared to the others. You can see this in the more extreme in a weekly graph:
You can see BitFury's sudden rise quite clearly, while BTCChina's is more of a slow but steady growth. You can also see BTCGuild shutting down ~10 weeks ago by its sudden plummet to 0 blocks.
And a statistic based on 7 days is probably a good thing.
There's several parties on Twitter that point to blockchain.info or blocktrail.com's graphs for the past 24 hours and saying "look! we're #1".. sure, for that day.. sometimes only just a few hours into a day. That's just misleading for the overall picture. So while having day-to-day stats or even block-to-block estimates is a fun exercise, it's not very indicative.
By the way: if you look at that first graph I posted - that's analogous to organofcorti's "Diachronic hashrate distribution". The same side notes apply there: if all pools stay at the same relative hash rate, then the 'bands' in his graph would keep the same width throughout the time span. Instead we can see that some of them have gotten wider (implying increased hash rate) while some have gotten more narrow (implying either same or decreased hash rate).
