Based on feedback from this list and further simulations, here is a new algorithm for Block75:
new max blocksize = x + (x * (AVERAGE_CAPACITY - TARGET_CAPACITY)
TARGET_CAPACITY = 0.75
AVERAGE_CAPACITY = average percentage full of the last 2016 blocks, as a decimal
x = current max block size
Please note that this algorithm actually tries to keep blocks 75% full, unlike the old one that unnecessarily capped growth at 250KB. While this would theoretically allow a maximum increase of 25% over the previous max block size, in practice it's not possible to get that high.
This would be calculated every 2016 blocks along with difficulty.
Block75 should maintain transaction fees at about the level they were in May/June 2016 when blocks started hitting 75% full somewhat consistently.
Thoughts? For any predictions as to how this would behave, please provide the numbers used to arrive at any conclusions.
Other questions:
1. How do we make Block75 play nice with SegWit?
2. Is there any need for a minimum max blocksize? Block75 allows for decreasing the size as well as increasing it.
Activation:
To help negate some of the risk associated with a hard fork and to prevent a single relatively small mining pool from blocking Block75's adoption, activation would occur once 900 of the last 1,000 blocks mined signaled support, with a grace period of 4,032 blocks.
Thank you again to all those who commented on the previous Block75 thread. Together, we can make 2017 the year the block size debate ends (hopefully forever).
Happy New Year!
- t.k.