It’s a question of proof density. No one is claiming the future block challenge can be predicted.
Isn’t proof density completely dependent on the challenge itself? One challenge could produce lots of proofs in a plot while another one would produce zero. On average and on a long enough timeline, all plots would produce the same number of proofs per challenge. So again, there is no use in attempting to measure “proof density” because that density would be for a specific challenge. Yes?
No, as mentioned in the paper, the average can be greater than one.
1 Like
Ah, I think I see what you are saying now that I read this bit in the paper that you reference:
Some plots will by random noise have greater per-byte effectiveness. The amount of this is proportional to the square root of the size of the plot with a small constant factor so the potential gains aren’t high. Furtermore, this requires evaluating the entire forward propagatio phase. If a farmer has multiple plots, and wants to delete one of them, the least effective one can be reclaimed first.
I think what this is saying, in layman’s terms, is that some plots just simply have more “answers” to choose from than others, just based on randomness and how the plot was originally created. And then that last sentence is definitely intriguing - it hints that there is a way to find these plots.
I’m all caught up now, thank you for the discussion and teaching me something new!
2 Likes
I wonder, is this related to the number of “entries” that are written to the plot? Here are several Phase 3 messages from my plotting logs - notice that the 2nd one has ~150k more entries than the first one. Based on the paper, could we simply consider number of entries for “plot quality?”
Phase 3 took 768.185 sec, wrote 21,877,147,145 entries to final plot
Phase 3 took 775.683 sec, wrote 21,877,322,509 entries to final plot
Phase 3 took 796.42 sec, wrote 21,877,223,886 entries to final plot
1 Like
IMOP the luck you seek splits hairs so fine that the value of the time spent searching far outweighs any added value you might find.
2 Likes
Are you telling me you aren’t gonna chase a 0.0006% advantage?!? 
*Note I don’t know if I even got that math right, or if my hypothesis based on “entries” is right, but yeah assuming it is: probably not even worth the electricity to replot.
2 Likes
I’m interested in:
#1 Answering the original question of the thread.
#2 Measuring the difference out of curiosity.
2 Likes
I’ll gladly take 0.0006% of a 2009’s pizza value in BTC!
1 Like
It’s worth it if you have to replot anyway! Eg. k33
That is a false equivalency. Sure, having just that 0.0006% would produce great results; however, in this case that percentage applies to you already owning 1XCH, as such it would even not show in your account (still could be worth Lamborghini 
1 Like
It would be worth ~700 dollars, it was meant as a joke 
As was my Lamborghini slight_smile:
Quoting from the link above, another kind of plot improvement:
Hellman Attacks usage
There is an experimental implementation which implements some of the Hellman Attacks that can provide significant space savings for the final file.
./HellmanAttacks -k 18 -f “plot.dat” -m “0x1234” create ./HellmanAttacks -f “plot.dat” check
1 Like
They said it would be implement probably by third parties first as they don’t intend to spend engineering time on this optimization. Their prediction is in the range of ~10% space saving iirc from their last ama to date.