Currently, I don't have the time to get into it too much. I'll post a more detailed explanation later on, tomorrow probably, on exactly where these things play into the algo vs ours. We are two versions behind because I saw that our algorithm had been stable for NERVA and Blur Network. Once they HF'ed instability was a major issue and they wound up with inadvertent 51% attacks happening as a result of many miners not downloading the new software... I'm not positive that was the reason... but it's what was the most obvious answer. of course, we won't ever know. For that reason, we have stayed back on our variant, since. I've also been watching the swings on their nethash lately (over 1MH swings on a 3MH nethash)... which serves to cause some concern, on my end. Their current v3 variant was a fork, correct me if I'm wrong here, due to expanding those layers for randomness a bit, and to regain stability. While they regained control of the network, stability doesn't support that 1MH swing.... thats a LOT of cpus...
For these reasons, I'm thinking about changing something much smaller, to make our algorithm dependent not on height of the block before it, but instead on a metric that is more intelligent and that can't be pre-determined just by looking at the algo and coding that into a miner. NERVA's implementation did just that, and that may be the cause of their diff swings, more than bad-actors. They added a random component that has a lot of possibilities, and I don't know how those factor into hashrate. However, their implementation shortened the window for diff swings to include (height) mod (64) as opposed to (height) mod (1024)... so that portion of the changes is not where that swing is coming from, because that would actually make the diff swing less. I'll write more when I'm more educated on the changes myself.
