Quote from: etblvu1 on February 02, 2014, 10:19:08 PM

Quote from: strelok369 on February 02, 2014, 10:02:51 PM

N = 36*10+1;
It's "signals" arrays length (don't like any dynamical memory allocation anywhere).
Those code representing 36 blocks sliding window and i simply want to see period as long as 10 windows.
N can be any value (not 36 dividible may cause some problems in this code, but it's only quick overwiev)

You should mention about astatism of diff recalc.
It's relative, so algorithm don't know what difficulty need to set for hashrate, it even don't know exactly hashrate.
Algorithm operates on how need to multiply current difficulty to get target block time with assumption(maybe aprove through probability theory),
that block time reverse proportional to hashrate with same difficulty and most important - that blocktime proportional to
difficulty with same hashrate. And all of coefficients of proportions is unknown and more, random (with distribution law
depending on the hash distribution law). And i use very symplified model of most probable average time (think it's "Expected value" in english terms).
Take it from LTC calculator as i remember.

If you take in account astatism, and tests show that blocktime stabilise on 10 mins after hashrate change and hold both ways, i'm calm.

And yes, catcoin never been anonimous, but i'm about logo design. My "serious cat" not a cake too.
I'm explain. Actual logo on my opinion has expected standart advertisment design. If you think about
food or traveling company logo it's what first come to mind - light offce mainstream style. And "old" anime-style
(not fan of anime, only like "ghost in the shell" for it's cyberpunk setting and not very simple plot) kitten was what you
never expect from currency, somehow original(in this place) and rememberable. It was cute and not serious,
it was attracting, something which "CATcoin" name can be accosiated. Posche - you imagine porsche 911 design. Bmw - you remember
double front grill and sign. Microsoft - you remember windows logo. Bitcoin - BTC. It's about associations.
If those gray kitten has no name and story behind it it's the best logo. Still offer gray kitten on mars surface.

Hi strelok369,

From a mathematical perspective, what do you think about the idea of breaking up the tendency for harmonics to develop, by replacing just the "36" portion of "SMA36LIM12" with a prime number pseudo-randomly at each block retarget, to be selected from a table? For example, if we take the least significant 4 bits of the hash of the last block, and used it to look up a table of 16 prime numbers consisting of {7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67}. Why this sequence of prime numbers? It's the closest consecutive list of prime numbers that approaches the average of 36 that we are already using (average of this list of primes is 34.875 and median is 31 or 37, so responsiveness to sudden changes in hash power would roughly match what we have now at 36, actually slightly better). And why prime numbers? By using prime numbers, we eliminate any unintended/unforeseen harmonics that could happen in relation to sampling sizes. Even if there is no danger of harmonics in this sense, it seems good to spread out the numbers a bit, and using prime numbers seems as good a way as any to distribute the numbers, and gives us a nice variability range from 7 to 67.

Thank you for your thoughts.

Etblvu1

p.s. If this approach is discussed it might need a name so, I'll give it the name SMA~36LIM12 to indicate the sample size is approximately 36, instead of exactly 36. And I guess "simple" moving average may not be the accurate name, but I don't know what else to call it.

Ok thought so, I was like that can't be a multiplication since it doesn't make since.

I have a question to everyone, I have my opinion on this but I want to hear yours, does the diff needs to be ruled symetrically ? how about making SMA36LIM12 when diff is increasing, but in case of decrease have it go down with any limite per block: how will this effect the coin well let me show you :
NB: The value I took are aproximatif and are for the sake of having a concrete example they do not represent reality (as hashrate is variable/s diff adjustement and so on)

Right now :

Explanation : SMA36LIM12 makes the diff change almost sysmetrical per 36 block and the point of sysmetry is the block where the hashrate drops significaly (or increase drasticaly for that matter) while this is supposed to be good, it is not because if we take in consideration block resolve time we see a huge disparity, because when the hashrate increase instantly, the block resolve time decrease by a huge marging, this makes block time resolve much shorter, the block resolve time increases with diff increase (12% a time) over time (over blocks) and gets closers to the supposed 600s target but once the Hashrate drop we are left with block with HUGE resolve time and due to the 12% limite, the next blocks will have a huge resolve time till we get close to the 600s resolve time again (it will take much longer to get diff down than to get up due to the limite and hashrate disparity, we are talking about an order of magnitude or two here)

So what if SMA36LIM12 is applied only for diff increase and the decrease is left limiteless:

Even if we suppose (absurd supposition) that profitability pools will leave the coin for the one block that the diff drops and comes back right away, we will have the diff retarget time decreasing overtime on each cycle (as these don't leave at the exact second before a new block, but it takes a couple of miliseconds to seconds to do so not to mention people who do this manually) which means with each cycle the chart is getting flat and and cycle it self it's getting shorter and less jerky, which will solve the issue.
But In reality independant miners that switch coins manually and profitability don't have instantnious reaction will make the coin level even faster and reach a sort of equilibrium over time on it own