Correct, that is the one! Nice work finding that.
What do you like about mempool.space in particular? The low/medium/high priority suggestions or something else?
I am aware that each node has its own mempool, but it's interesting to compare the info mempool.space shows when it comes to the total mempool size and the number of unconfirmed transactions, and the data jochen-hoenicke has. While writing this, mempool.space shows a mempool size of 18MB and over 26.000 unconfirmed transactions. At the same time, jochen-hoenicke only has a mempool size of 14MB but about the same number of unconfirmed bitcoin transactions - +26.000. Why the difference?
I prefer using the 8 hour graph, as seeing the trend over more than 2 hours can be relevant. If the mempool was empty 8 hours ago and has been steadily filling up since, then I would aim near the tip to get confirmed within the next couple of blocks. Conversely, if the mempool was 30MB larger 8 hours ago and has been steadily emptying since, then I know I can aim a good bit below the tip and still be likely to confirm in a reasonable time.
That is reasonable thinking, unless a sudden spike of transactions occurs just around the time when you broadcasted your transaction with lower fees when you saw the mempool emptying during the last 8 hours and it keeps going upwards. But looking at the 2 hours graph wouldn't help in such a situation either.
The least ambiguous way of discussing it is to say that a block can take 4 million weight units.
That now brings in a new actor in the game. So we have MB, vMB, and now weight units. That might become overwhelming for the average joe.