do i get this right, you are suggesting some kind of... reverse hopping?
at first, hoppers switched to a pool when a new round started, knowing they would still get paid enough for a finished block in the old pool.
After scoring has been changed to prevent this, and shares submitted later in the round gain higher scores, its time for hopping 2.0!
The average round time could either be calculated via the n last rounds, or just estimated by pool hash-rate, and neo hoppers pop in to provide well scoring late shares.
Next evolutionary step will be hybrid-hopping, just find yourself one pool with proportional distribution to join early in the rounds, and one with slush-scoring to join late in rounds
I'm aftraid this reverse hopping wouln't work..
Long blocks are less frequent than short ones.
And you can't estimate current block duration..
Normal hopping gains from short (and very
short) blocks. Nothing like that is available in
"reverse hopping". I believe one can just lose.