The introduction of randomness has an interesting effect:

On average, the expected result is known. For any given week, however, you don't know the distribution. So, you can't go balls to the wall because the next few days may also be. But you can't just play it safe, because the next few days may minima.

It incentivizes you, without even thinking about it, to calibrate correctly.