This is another graph of the PosDiff and the interval between the blocks, starting
from day 2015-09-09, as you see the values have decreased further.
In making this series of charts, I realized that the times of the blocks jump
back and forth too many seconds abnormally.
My hypothesis is that these jumps are the secondary reason that pushes down
the PosDif, and the interval between the blocks.
So I tried to figure out if these jumps were equally distributed between all addresses.
With some difficulty and with a little empirical methods, I tried to calculate the
average time between the previous block and that generated by a given address,
and between this and the next block.
These times are generally very random, but considering a large number of blocks,
the end values tend to stabilize, and should converge to the normal average time
between the blocks.
While in most of the addresses this values are practically symmetrical, there are
groups of addresses that instead show a strong preference to producing blocks
or in the past or in the future, in short they have a marked time offset.
This generates the anomalous jumps in time of blockchain of which I wrote earlier.
I thought about what "normal" causes could produce a time offset in the wallet,
for example, a drift of the system clock, or a excessive memory effect in the
mechanism of the median time of wallet.
To test these cause I asked to do that experiment, to close the wallet, check
the time and restart the wallet, and I took as a reference 12 addresses from
the top stacker that were representative of the various groups.
Among these are four addresses of group A (generating blocks at least 1 minute
in the future), three of the group C (values virtually symmetrical, just for
control), the other by Group B (back in time 20 or more seconds).
The graph it is also based on windows, wide 24 hours, then also a sudden event,
is presented as a gradual sliding, the width of which is represented by two
vertical dashed red line.
Although you may notice that after a few hours from the request, the
addresses of group C, traces green and magenta around 0, settle on to
new values, presumed effect of the request, the rest of the results
were disappointing.
Especially at global level. In fact, if a positive increase in the POS's
actually been, it was short, and quite after the request, and honestly can
not be said that it is an effect in the experiment.
Possible causes of the failure of the experiment:
1) A few user have read or complied with the experiment request.
2) The primary cause of the problem, the one that generates the offset,
resists to the restart of wallet.
3) I got it all wrong.
4) The problem does not arise from "natural" causes.
But if you look down at the track cyan, can be seen as an address that
belongs to group B (for unknown reasons) at some time has corrected
substantially its position to over zero, at least for some time.
This would leave thinking that something can be done to correct the offset.
All previous observations have been made by taking as the sole reference
the blockchain's time. Today I made an observation in real time between the
block-chain and a sure time reference, and the results with the inaccuracies
due to a smaller set of data, confirm the general data and offset.
I personally remain convinced of the initial hypothesis, despite everything.
General recommendation, try to reduce as possible the offsets effect.
We just have to find a way.
Bye Bye
