It was made to quickly answer many questions that could only be done with the tables of numbers from accessible version.
Say something happened. You want to know how many hours ago was there a big move? How is it holding after that? How does it compares to yesterday? To the week before? To the month before?
To do that, even if we have data, the resolution must lowered. Because you want to a quick eyeballing, and if the bucket size vary too much, it is much harder.
Take XRP for example, go http://cryptomarketplot.com/timeline/ripple.pdf
[actually, I must have crashed the PDF part last week, so the numbers are from Dec 2. Sorry. Let's imagine it is today. We are working on it right now]
You see that the peak was 6 days ago, it had a large jump compared to the other DAYS. And the left, you see that on the last 4 months XRP/BTC has kept growing.
If you just look at the graph very quickly, you see a patter. The shape.
You can visually correlate the pattern to the other coins, and see if they all responded in the same way or not to a market event.
I am liking the non-linear x-axis on the slow zoom graphs. I had never seen those before. I just wonder why the resolution in the month and week part of the graphs is not so good? You should have many more data points than what you are rendering, shouldn't you?
I guess that you are preserving min/max in a period despite the lower sample rate?