EMA-Z Difficulty Algorithm about difficulty-algorithms HOT 8 OPEN

hello,
I'd like to test your daa, but getting confused with these two lines:
ST = max(10T,min(-9T,ST))
how does -9 * t come? in cpp from my understanding, expr above always returns 10t, then why do we
need this?
prev_target * ((TN)^2-T^2N+k(k/2+(TN)^2-T^2N/2) / (TN)^2
should the missing bracket be placed like below?
prev_target * ((TN)^2-T^2N+k(k/2+(TN)^2-T^2N/2)) / (T*N)^2

Looking forward to your reply.

from difficulty-algorithms.

from difficulty-algorithms.

@zawy12 Loved your efforts here. Are you in San Francisco and would you be interested in giving a 1-hour presentation/talk on difficulty Algorithms?

from difficulty-algorithms.

I think for this level of awesome, we should pay you to fly out and come speak! I'll send you an email direct once I touch base with the rest of folks with a proposal.

from difficulty-algorithms.

Will there be some live streams? It will be awesome too if anyone interested can view your talks via live streams. @taariq

from difficulty-algorithms.

I should do a video that summarizes everything I've learned.

from difficulty-algorithms.

(weighted) averages tend to be sensitive to corrupt data, a problem weighted medians do not have. This should presumably give a measure of current hash rate that is less sensitive to manipulation. Have you run any experiments either with a linearly weighted moving median or an exponential moving median (if so, can you make my life easy and link the comparisons?)

from difficulty-algorithms.

Medians interject a delay that is proportional to the amount of protection. The problem is that delay is in every block instead of just the bad ones. There is very little manipulation that can be done in the newest l w m a algorithm thanks to the stricter limits. The best thing to do is to just accept all the time stamps as honest even if they go back in time. The honest timestamps immediately cancel the negatives of the bad time stamps. So the only manipulation that can occur for more than one block is only when the attacker has more than 50%. With a stricter limits maximum benefit he can get is about 5% lower difficulty for only as many blocks as his multiple of the Baseline difficulty. For example if he has 10 times the base line difficulty he has a 50% chance of getting 10 blocks at about 5% lower difficulty

from difficulty-algorithms.