Summary of Difficulty Algorithms about difficulty-algorithms HOT 8 OPEN

I see you're keeping up with things. So far I have everything on that in my TSA article that I keep updating. Any of the above can use solvetimes that are shifted forward 1 more block to be Real-Time-Target algos. This article is to give an overview of DA's prior to RTTs.

from difficulty-algorithms.

good to know! BTW, what about RTT?

from difficulty-algorithms.

fwiw, if you're interested in completeness, you may want to include Cryddit/Ray Dillinger's MIDAS: http://dillingers.com/blog/2015/04/21/altcoin-difficulty-adjustment-with-midas/

from difficulty-algorithms.

I see Shield and Verge switched from Midas to LWMA. It looks like maybe 1 or 2 coins are using it. There is another one I know of "OSS" that 1 or 2 coins use. I just wanted to cover the big ones and ASERT and TSA as they are the only better options. Midas's motivations and ideas are good but there are at least 6 different parameters it uses (9/10, the 5 to 17 choice, 5/8, 2/3, and 6/5) where all the other algos have only 1 or 2 parameter because their math basis is clear. Digishield has 2 parameters (17 and 4/3) and KGW has 2 (0.7084 and 1.228). It would be nice to reduce it. It uses two ideas I never tried. One is like ASERT in that it makes a more aggressive correction if the long-term average solvetime is incorrect. Unlike ASERT, the method looks like it could be manipulated in on-off mining and result in oscillations. It's also interesting in the way it requires both a long and short term avg solvetime to be a little fast or slow before it makes a correction. But the way it does it could be manipulated in on-off mining, causing oscillations. For example, an on-off mine could be 3 blocks on and 3 block off and get up to 33% more blocks than target without causing any adjustment in difficulty.

from difficulty-algorithms.

Thanks for taking the time to analyze MIDAS' position and status within the current context of development and thanks also for being explicit about the parameter criteria, I found that to be both illuminating and useful.

from difficulty-algorithms.

https://eprint.iacr.org/2017/731.pdf

from difficulty-algorithms.

from difficulty-algorithms.

I tried various methods of correcting a simple moving average for the slope which is like applying a least squares, but I never tried a full-blown least squares. I could never get it to work out better than a simple moving average.

How can the least squares method in the paper work without using any timestamps? I can see how it can work for the where the slope is constant, but not if the slope should change over time which will be determined by timestamps which tell us what the hashrate is.

from difficulty-algorithms.