Block Production analysis July 2013

I’ve temporarily forsaken column graphs for a line chart which best shows the improvement in Feathercoin of block production in July over June. I’m assuming that’s due to 40% difference change restriction implementation and possibly despite some attacks.
Blue  Feathercoin Block production per day Specification level,
Dark Green  Block per day  1 Month Average per day  July 2013
Light Green, Actual Blocks per produced per Day July 2013
Yellow  Accumulating Average Feathercoin Blocks per day for July 2013
[attachment deleted by admin]

That yellow line has me a bit concerned, and leaves me with little doubt as to why the network is constantly attacked: There’s no smoothing of difficulty. It’s constantly too high, or too low, never just right. When it’s too high, miners leave, making the problem worse. When it dips again, they come back and push it up again.
Can we get a comparison of other coins and their block generation rate? Specifically one with a fixed retarget rate, and one with a constant retarget rate?

Here’s an interesting update of the Feathercoin Block production chart, showing the first 3 days of August. It appears the Hash rate must be up, as the difficulty is increasing. Its not the usual weekend hash increase (I see on the p2pool), although it could be a trend of more miners…
The days of concern can be seen clearly where the green daily, block production, line is excessively high.
I’ve included another chart that is interesting, it shows the averages daily block production per month in a column format. Feathercoin Block production was down in July, it has now gone up, despite the difficulty going up. My only current explanation is a gradually increasing hashrate. Unfortunate I haven’t got daily average hash rates to do a correlation
[attachment deleted by admin]

Wrapper, could you produce a graph for PPC in July? They retarget every block looking one week back exponentially for difficulty calculations.

Hi Ghostlander,
I’ll have a look into it, I am (Physically) limited tho’ and am doing a bit too much for Feathercoin. I think its in a good cause.
Do you have any stats I could import?

I suggested that the Difficulty should be set on 3 averages 100, 500 and 1000 blocks. This is because that is self aligning over 3 fractal dimensions.
At the moment the difficulty is adjusted on the time to complete ~ 500 blocks  Average time for block = Time to do 500 block / 500. About 500 or so block should be made a day (roughly)
The new average would be Average time per block = (Avg over 100 + Avg over 500 + Avg over 1000) / 3
You can see this equation automatically compensates for short duration changes whilst maintaining the long term goal.
It is based on a system I was involved in inventing, which set the number of averages depending on the signal characteristics.
Obviously, The PCC difficulty calculation, you describe is more fixed in its logarithmic, and over too short a period. Whereas the method I describe associates 3 ranges of interest and could apply a weighting based on need to further refine the tool.
We should look at the times scales over which the Feathercoin block rate is variable and adjust the parameters accordingly.
At the moment Feathercoin is taking 2 weeks to damp excessive variability of the block rate, that would indicate we should add ( 500 * 14 ) to the equation.
Calculated Block average time for Difficulty calculation = ( Average for 100 Blocks + Average for 500 Blocks + Avg over 1000 Blocks + Average over 7000 Blocks ) / 4
or adding a weighting to compensate for the current Feathercoin block underproduction, by multiplying the long term average and the then dividing the total time per block by 5 to get the new average time per block.
Calculated Average Time per block = ( Average for 100 Blocks + Average for 500 Blocks + Avg over 1000 Blocks + 2 * (Average over 7000 Blocks ) / 5

[quote name=“wrapper0feather” post=“24576” timestamp=“1375731195”]
Hi Ghostlander, have you got a link to PCC?I suggested that the Difficulty should be set on 3 averages 100, 500 and 1000 blocks. This is because that is self aligning over 3 fractal dimensions.
At the moment the difficulty is adjusted on the time to complete ~ 500 blocks  Average time for block = Time to do 500 block / 500. About 500 or so block should be made a day (roughly)
The new average would be Average time per block = (Avg over 100 + Avg over 500 + Avg over 1000) / 3
You can see this equation automatically compensates for short duration changes whilst maintaining the long term goal.
It is based on a system I was involved in inventing, which set the number of averages depending on the signal characteristics.
Obviously, The PCC difficulty calculation, you describe is more fixed in its logarithmic, and over too short a period. Whereas the method I describe associates 3 ranges of interest and could apply a weighting based on need to further refine the tool.
We should look at the times scales over which the Feathercoin block rate is variable and adjust the parameters accordingly.
At the moment Feathercoin is taking 2 weeks to damp excessive variability of the block rate, that would indicate we should add ( 500 * 14 ) to the equation.
Calculated Block average time for Difficulty calculation = ( Average for 100 Blocks + Average for 500 Blocks + Avg over 1000 Blocks + Average over 7000 Blocks ) / 4
or adding a weighting to compensate for the current Feathercoin block underproduction, by multiplying the long term average and the then dividing the total time per block by 5 to get the new average time per block.
Calculated Average Time per block = ( Average for 100 Blocks + Average for 500 Blocks + Avg over 1000 Blocks + 2 * (Average over 7000 Blocks ) / 5
[/quote]I like this idea quite a bit for diff retargets. I thought of another metric that could be added to that averages equation, get the Average time per block of all of the blocks from the start. This would give another average to pull us toward our long term goal and help stay on target for distributed coins over time.

Cheers, for that mate. I’ve updated the charts to correct a couple of minor errors (They started in April not May).
I’ve aligned some of the markers to Sunday, shown in red. I’ve been doing the chart averages over various time spans, to see which is most at the scale of the variability.
[attachment deleted by admin]

their is one thing that you should consider we retarget every 21h in theory. usually high diff are longer so we get lower diff after. but low diff are shorter and we get higher diff after so in many case you don’t see the low diff as they are ~12h.
i don’t know if you can make it an hourly points on a continuous line with same x scale, the graph would be pretty different. I think would looks mostly a wave cure

Hi Groll, I agree who does the Forum site? It would be easy to add some hourly and Daily graphs, just need to compile a reading every minute or hour.
Unfortunately, I’m just sampling these manually per day so don’t have live information.
If this were a job, I’d have got my programmers to knock up some stuff to automate my best / most useful graphs. If the Database sampled every hour that would be enough resolution, to show security, without being excessively large? It would help with compliance, showing transparency.

[quote name=“wrapper0feather” post=“24553” timestamp=“1375721495”]
Hi Ghostlander,I’ll have a look into it, I am (Physically) limited tho’ and am doing a bit too much for Feathercoin. I think its in a good cause.
Do you have any stats I could import?
[/quote]Never mind, I’ve got the answers already.
http://cryptometer.org/ppcoin_104_week_charts.html
PPC generates about 20% PoW and 80% PoS blocks now. They are into real problems which can lead to market collapse. PoS is killing PPC basically.
[quote author=groll link=topic=3128.msg24613#msg24613 date=1375752569]
their is one thing that you should consider we retarget every 21h in theory. usually high diff are longer so we get lower diff after. but low diff are shorter and we get higher diff after so in many case you don’t see the low diff as they are ~12h.
[/quote]We retarget every 504 blocks, but look back 21 hour for difficulty calculations. 504 blocks = 21 hour in theory, but it’s different in reality. That’s why I suggest to retarget faster while keeping 21 hour retarget window.