Indeed, it's rare because there is no worse!
But the solution is not in changing my progression, there isn't better progression to handle such negative outcomes.
Let's take a step further and establish some intervals every 100 outcomes...
From 25 wins out of 100 results we are arriving on the next interval of 66 wins out of 200 outcomes.
Let's assume that the first 100 and the next 100 are from the same session, 25/100 is 25% while 66/200 is 33%.
As you see from the first checkpoint to the next there is an increasement of 8% (33-25=8).
We could expect a similar increasement (8%) within the next set of 100 results which means approximately 40% of 300 outcomes or 120 wins VS 180 losses.
It's becoming apparently clear that regression towards the mean takes place after extreme deviations, so by expanding the event's horizon we're moving closer towards the mean and in return help us to recover smoothly.
This is where the solution lays and could be applicable in practical terms (within limits of time and money)