It's good if something is discussed to make things clear :) If s.o. beliefs I tell nonsens or err, please give feedback.
Let's start with some facts:
The doubling algorithm is more or less based on Janowski formulas. Therefore the value for double and ND is the value of the two curves (in fact straight line) at the current equity and the value shown as cubeful equities tells us the equity if I double now or double never.
                      |
|
e.g. green is better off about 0.1 if doubling now than by doubling never. But this values doesn't say anything what it costs to double later and it is nonsense to add up this value. Here waiting 1 roll costs nothing. IMHO it is difficult (to impossible?) to compute the cost of delaying the double for 1 roll in most non trivial situations.
let's take this position:
          |
|
The evaluation gives 0.16 for doubling now. A rollout gives about 0.081 for doubling now. The equity difference in the rollout reflects exactly (in the general limits of a rollout) the difference between doubling now and doubling one roll later. This value can and must be added, independent whether the situation reoccurs.
in one sentence: to get a reasonable ER value for an cube error you have to do a rollout.