Block finding is purely random - you don't build up to finding a block.
It's random in exactly the same way as rolling an un-weighted dice is random
(though the block hash is pseudo-random coz it's a mathematical calculation with enough complexity to consider it random)
You double sha256-hash an 80 bytes value, check it's pseudo-random result and see if it's less than a value calculation of the network difficulty.
If it is, then it's a block.
So yep there is absolutely no history effect on finding blocks, and you can NEVER say when the next block will be found by any pool or even the whole network.
The block length distribution is a poisson distribution.
https://en.wikipedia.org/wiki/Poisson_distributionMeaning you expect them to average at 100% difficulty, but the length between blocks can be 1 hash or infinite hashes (never)
The CDF table also explains that - it's not just for the pool, it's for all of Bitcoin.
0.39346934028737 50.000% 1 in 1.6
0.63212055882856 100.000% 1 in 2.7
0.86466471676339 200.000% 1 in 7.4
0.95021293163214 300.000% 1 in 20.1
0.98168436111127 400.000% 1 in 54.6
0.99326205300091 500.000% 1 in 148.4
0.99752124782333 600.000% 1 in 403.4
0.99908811803445 700.000% 1 in 1096.6
0.99966453737210 800.000% 1 in 2981.0
0.99987659019591 900.000% 1 in 8103.1
So using that table also for all Bitcoin network blocks, and assuming blocks are expected to average 10 minutes in length,
you can see that since 10 minutes would mean 144 blocks a day, then it would average, over time, a 50 minute block about every day -
this comes from "500.000% 1 in 148.4" i.e. "500%" of 10 minutes = 50 minutes, and "1 in 148.4" is a little over a day's expected number of blocks: 144.
(144 = 6 per hour times 24)
The most important part to understand is that it's an expected average, nothing more.
So when we say the network blocks should average once every 10 minutes, it doesn't mean they will be every 10 minutes.
Some will be a few seconds, some 50 minutes, some half an hour, some 1 minute etc etc etc
But the long term average of all those is expected to be 10 minutes.
Lastly, most if not all Difficulty graphs are interpreted incorrectly by most people who view them.
You cannot give an accurate estimate of the network hash rate from the graph.
Yep, you can't.
Why? Coz the time used to estimate/update the network difficulty is 2016 blocks, and even that is not very accurate.
The network hash rate changes every minute, every hour, every day, but a day's worth of block data is far from accurate enough to estimate the network hash rate.
As I implied, even ~14 days worth of data (2016 blocks) isn't very accurate, but it's necessary to be used to reduce the effect of a large hash rate drop.