Updated after a closer examination of cost efficiency.
Table shows turns since the first turn this card is played.
The assumptions being that this is the only thing marking cards, and that the player uses its ability every turn.
The columns starting with 'E_' represent expected values. E.g. the average you would expect over many games.
I.e. E_Trigger is the expected value for the number of times the ability would trigger after a given number of turns after being played.
With multiple copies of this in play, the damage can ramp quite quickly. The upper limit maximum of course is +3 per turn / +4 per turn for unupped / upped respectively. Which puts the upped in line with a steam machine in terms of damage ramp in the case of having both decks with all cards marked.
(For anyone interested, I generated the table using Jupyter notebook with an R kernel and the dplyr package)