Topic: [ANN] Noirbits Update required Improved algos !!!! - page 4. (Read 74427 times)

If a block takes too long to find, diff. will automatically drop on next drop. If you look at the previous sequences I posted, the 30min. or more case will yield the following :

120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-1800 = 2.93 mn block average

Which is way above the 10% limit, triggering a diff. drop on next block.

What I'm gonna do this time is extract actual block times from the NRB chain, and use those as test cases for the new diff. algo, so we'll have concrete data to monitor the suggested change's impact on difficulty.

So per block target it is, but sometimes diff may rise to the point that 30 min go by without a block solved....should we leave it like that or adjust the time to force diff dwn after a certain time period. ?

Not too fond of the idea. I think evaluating retarget at every block is a smoother move : it gives the block chain the opportunity to recover from hashrates spikes as soon as they are gone, and inversely, to raise difficulty as soon as hashrate spikes (like, when a multipool kicks in). Retargetting every 5 blocks is too short if you only look back 5 blocks, you need to take variance into account.

The core idea is not to lower the count of blocks used to evaluate difficulty, but rather to evaluate retarget at every block, by looking back at the last 30 blocks and the time it took to find them. So when a hashrate spike comes along, the next block is going to be found way too fast. Difficulty might rise just a little bit, or not at all. But on second block after spike starts, it will rise just a bit more. And so on...

Assuming we have a stable 2 minute block time before a spike that brings block time down to 10s, you'll have this sequence going on for block times :

120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120 = 2mn avg
120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-10 = 1.93mn avg
120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-10-10 = 1.87mn avg
120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-120-10-10-10 = 1.81mn avg
....
And so on... if you don't retarget.

But if you retarget everytime the 30 block average time is off, by say +/-10%, you'll be forcing retarget by 3 block. Diff won't rise as much, but it will rise. The same will be true the other way around, but we won't reach situations where spikes leave diff. way too high. The previous difficulty fix had a bug, but it still proved useful. Difficulty was left by a spike @ 10 at one point. By next block, which took more than four hours to find, diff. dropped back to 6. I didn't monitor what went on afterwards, and don't remember the block height, but it shows that this is a good starting point.

What we really need to avoid is massive difficulty variations, they penalize loyal miners and encourage coin hoppers, who only come back when diff. has dropped back to looow values thanks to the backing of loyal miners.

agreed, what do you think of the 5 block retarget?

Gonna work on that as soon as I can...

Also, on another issue, I think fees are way too high. If you look at Bitcoin, min. fees are 0.001 BTC for non-free transactions. With Noirdice, we're at 0.1 NRB for non-free transaction. So basically, NRB transaction are 100x more expensive than BTC. I think we should lower them to 0.001 NRB to encourage NRB transactions.

The change could be included with the upcoming difficulty rules to ensure everyone gets those changes at the same time...

For a start it's great , i'm thinking of applying new rules @ block 33333. any ideas?

Yeah, I will increase them once I'm sure everything rolls smoothly. I've ran a lot of tests, but you can never be too sure.

PS. Sorry for the design, but it's *really* not my area of expertise...

Nice job! We're gonna need a higher "max bet" cap, but I'm guessing it is only this low for beta purposes. 

great work, now jus need to adjust the diff, i was thinking of either a per block adj or else a 5 block adj with immediate rises to match hash rate.

Well, finally ! Noirdice is up & running, running bets on 0 confirmations, so you get instant results

You can give it a try @ http://www.noirdice.com

Please note the site is still in beta phase, I will be monitoring bugs & so on... Any and all feedback is welcome.

Noirbank is stagnant due to waiting on paperwork. As for the retarget i have decided to set a block target 3 cycles away from when i push. What i am trying to get right is the quick drop in diff.

I've got a question about the upcoming update. Will we fork to the new version soon after the release of the updated client or will we wait another week before doing so?

Also, any news about NoirDice or Noirbank? 

This morning there was a 0.4 diff.
Then it went on top of CoinChoose, 5 seconds later we have a 10.11 diff which is lasting hours and automatic pools guys gone far away xD.

True that! 

The beauty of which has allowed others to buy up cheap nrb

Hurray! 

The multimine pool has kinda been killing the coin recently...

Algo update incoming, give me two days. I am thinking that a per block re-target is best instant high and instant low, and in the event of 20min no block solved, it drops to previous diff. it forces you to either stay or leave. The multi mine pools will leave us alone after they notice this.

 

mine is synced, but i'm guessing this is due to there being very few mining. I'll be looking at the diff code tonight and offer my idea later.

It's been an hour and my transaction still has 0 confirmation...