If all the pipes had the same difficulty and all the players had the same ability, then the only factor to pass a pipe is luck and a x% of the players die just before each pipe and the graphic is a row of equal mountains.

But some of the pipes are more difficult or easy, so they will have a bigger % or smaller % of deaths, and higher or lower mountains.

And some players will have more ability. If, for examples noob, have a 1/3 chance of diying per pipe and pros has a 1/20 chance of dying per pipe, then the first 3 or 6 mountings will be big and full of noobs carcasses, and the other mountains will be much smaller. (In real word there is not a binary noob-pro classification, but it's easy to explain with this toy model.)

They were never seen in "real time" anyway. Communicating more facts that could have been derived from a smaller set of different facts does not make it any less of a "recording" of facts or more "real time". Fewer facts to communicate may enable lower latency in their delivery, which is a better definition of "real time". Would it be any less of a "recording" to send the rendered pixels of the fish? I would argue the opposite.

Perhaps the point was that much more time was passing before elements of this list in the new format were sent, is that true?