Yes, but what I was trying to say is that bad actors do not happen at a regular 50% rate, they have their own distribution and that can accentuate extreme cases.
Masternode is up, everything OK - no payments.
- Thread starter pyc
- Start date
OK, I just sat down and ran a quick and dirty analysis using some broad assumptions, and I can't find any evidence of anything outside the ordinary. In fact, it looks very similar to the way the math says it should look. My simplifying assumptions were that 1) there were 400 masternodes active from 6/26th to today, and 2) about 50% of blocks made a payment during that time.
Analysis of historical:
According to those assumptions, you would expect to see about 30 masternodes with 17 or more payments, so finding an example of a masternode with 17 payments was not unexpected. If you've been running a masternode since June 26th, the most likely outcome is about 11 payments. On the low end, I would be surprised to see an example of a node that had been running that whole time and received 0 or 1 payments, but even finding one example with only 2 payments would not be unexpected. As I scan through the MN's with uptime going back to the 26th, I'm not finding any examples of 0-1 payments. On the high end, I would not be at all worried unless you could find an example of a masternode with at least 26 payments (the probability of seeing one is less than 1%). Even then, if a lot of those payments came in the first few days (when the number of active MN was much lower than 400), it would require further analysis to confirm anything out of the ordinary (and I'd need to know exactly how many MN's were active at any one point in time to do this more accurately than my simplifying assumptions allow).
Going forward, what to expect?
Now that the number of masternodes has stabilized at around 520, here is what you can expect going forward:
Over the course of the next two weeks, assuming 520 masternodes and 50% blocks with MN payments, we should only be able to find one or two examples of masternodes that don't get paid at all during this time. If we see more than about 6 examples of masternodes that operate continuously during that time, are configured correctly, etc. that haven't been paid, then I would start to suspect that we have a problem. The chances of that happening are indeed low... less than 1% chance. Obviously, if the ratio of blocks getting paid rises higher than the current ~50%, we could start to draw conclusions sooner than two weeks from simply observing that there are some non-paid MNs. At the other extreme, now that we have lots of MNs operating, the chances of someone getting 18 or more payments over the next two weeks is very low and could also illustrate we have a problem. Again, this assumes 50% of blocks paying, so if that ratio starts to increase, even finding nodes with more than 18 payments could be likely (I'd need to rerun the analysis based on the actual observed ratio).
Once payments are enforced:
Once all payments are enforced, you really need to wait a week before drawing any conclusions. If we can find an example of a MN doesn't get paid once in over a week, that could illustrate a problem. But again, this assumes 100% of blocks include a MN payment, which isn't the case today.
Bottom line:
Observing 2 days, 4 days, or even 14 days of no payments on a single masternode is not unusual when we have ~520 operating and only 50% of blocks getting paid. Truly random selection will result in some examples of extreme results across a population of 500+ MNs on both the high end (e.g., 20+ payments) and the low end (0 payments to date). With more data we could draw stronger conclusions, but thus far, there is no evidence of anything unexpected given a random selection. In fact, it would be unusual and concerning if we didn't observe some examples of zero payments or 20+ payments given the short length of time payments have been operating.
I hope that gives everyone a bit of comfort. The more I think about it, the more I like the idea of a "masternodedays" or "masternodeblocks" concept similar to coindays. It would help prevent extremes and make things more fair over shorter periods. I don't know how difficult that would be to incorporate, but it could help ensure greater fairness in the distribution of profits without creating a risky situation in which the next node could be known in advance (like with a turn-taking arrangement).
Analysis of historical:
According to those assumptions, you would expect to see about 30 masternodes with 17 or more payments, so finding an example of a masternode with 17 payments was not unexpected. If you've been running a masternode since June 26th, the most likely outcome is about 11 payments. On the low end, I would be surprised to see an example of a node that had been running that whole time and received 0 or 1 payments, but even finding one example with only 2 payments would not be unexpected. As I scan through the MN's with uptime going back to the 26th, I'm not finding any examples of 0-1 payments. On the high end, I would not be at all worried unless you could find an example of a masternode with at least 26 payments (the probability of seeing one is less than 1%). Even then, if a lot of those payments came in the first few days (when the number of active MN was much lower than 400), it would require further analysis to confirm anything out of the ordinary (and I'd need to know exactly how many MN's were active at any one point in time to do this more accurately than my simplifying assumptions allow).
Going forward, what to expect?
Now that the number of masternodes has stabilized at around 520, here is what you can expect going forward:
Over the course of the next two weeks, assuming 520 masternodes and 50% blocks with MN payments, we should only be able to find one or two examples of masternodes that don't get paid at all during this time. If we see more than about 6 examples of masternodes that operate continuously during that time, are configured correctly, etc. that haven't been paid, then I would start to suspect that we have a problem. The chances of that happening are indeed low... less than 1% chance. Obviously, if the ratio of blocks getting paid rises higher than the current ~50%, we could start to draw conclusions sooner than two weeks from simply observing that there are some non-paid MNs. At the other extreme, now that we have lots of MNs operating, the chances of someone getting 18 or more payments over the next two weeks is very low and could also illustrate we have a problem. Again, this assumes 50% of blocks paying, so if that ratio starts to increase, even finding nodes with more than 18 payments could be likely (I'd need to rerun the analysis based on the actual observed ratio).
Once payments are enforced:
Once all payments are enforced, you really need to wait a week before drawing any conclusions. If we can find an example of a MN doesn't get paid once in over a week, that could illustrate a problem. But again, this assumes 100% of blocks include a MN payment, which isn't the case today.
Bottom line:
Observing 2 days, 4 days, or even 14 days of no payments on a single masternode is not unusual when we have ~520 operating and only 50% of blocks getting paid. Truly random selection will result in some examples of extreme results across a population of 500+ MNs on both the high end (e.g., 20+ payments) and the low end (0 payments to date). With more data we could draw stronger conclusions, but thus far, there is no evidence of anything unexpected given a random selection. In fact, it would be unusual and concerning if we didn't observe some examples of zero payments or 20+ payments given the short length of time payments have been operating.
I hope that gives everyone a bit of comfort. The more I think about it, the more I like the idea of a "masternodedays" or "masternodeblocks" concept similar to coindays. It would help prevent extremes and make things more fair over shorter periods. I don't know how difficult that would be to incorporate, but it could help ensure greater fairness in the distribution of profits without creating a risky situation in which the next node could be known in advance (like with a turn-taking arrangement).
By the way, this appears to be the luckiest masternode so far...
XyRzfcdK94NmvYsRPQisNdjzpmuDstqooW
... and what do you know? It is sitting right at 26 payments! This is right at the "upper range" I predicted (though he got 4 payment on the 26th and 1 today... both days I excluded, but I was pretty close!).
XyRzfcdK94NmvYsRPQisNdjzpmuDstqooW
... and what do you know? It is sitting right at 26 payments! This is right at the "upper range" I predicted (though he got 4 payment on the 26th and 1 today... both days I excluded, but I was pretty close!).
flare
Well-known member
OK, now i got your question :smile:
Conceptually the vote system works the same as a dice throw, (currently a ~500-sided dice) with the payout not being immediately after the throw, but when the throw has been confirmed 5 times. This is different from a lottery where the balls are removed and first put back after 5 draws ("vote reset")
So, if a vote (dice throw) for a node is in the "confirmation queue" it is
a) not reset/deleted after a payout and
b) nodes which already got a vote are not removed from the candidate list and are available for further votes.
So there is nothing odd with a masternode being paid 2 or 3 times in a row. It maybe very unlikely (0.002^3 = 0.000000008 ==> 1:125,000,000) but not impossible :smile:
Hope this made it clearer :smile:
Thanks for this exhaustive analysis, good stuff! :smile: When it comes to questions re. the statistical details of Darkcoin in the future, i will refer to you.
Last edited by a moderator:
I watched again. The adress got 3 payments yesterday again: XrfFymhBuVrFhCcZZZ8PULhHeNf7iC9vXV
Sorry people, I know it's possible to win 2 times in the lottery in a row but the possibility is very very little.
Until you found a better random function may you block new payments for nodes for 1/2 of average payment time after getting one and don't give them any votes. It's just an idea.
Hows it's going now, it's not fine.
Sorry people, I know it's possible to win 2 times in the lottery in a row but the possibility is very very little.
Until you found a better random function may you block new payments for nodes for 1/2 of average payment time after getting one and don't give them any votes. It's just an idea.
Hows it's going now, it's not fine.
vertoe
Three of Nine
3 payments on Jul/4th and 0 payments on Jul/3rd - look pretty good.
Franz2, listen I understand that intuitively, it seems unusual that XrfFymhBuVrFhCcZZZ8PULhHeNf7iC9vXV should get paid so many times. But believe me, across 500 MNs, at least three MNs are bound to get that lucky over such a short period of time. Also, getting paid two blocks in a row shouldn't even be that unusual of an event. Consider this... if there were 500 MNs (about what we have), that means in about 1 of 500 blocks, the MN payee should be the SAME address twice in a row. Considering there are 576 blocks each day, that means someone getting paid two consecutive blocks in a row should be about a daily event. It is not evidence of unfairness.
And here's where I'll put my DRK where my mouth is and make things a little interesting / fun... I'll bet you - even money - any number of DRK you want up to 10 that XrfFymhBuVrFhCcZZZ8PULhHeNf7iC9vXV doesn't get paid over the 24 hour period from noon Monday EDT to noon Tuesday EDT (to give you a little time to reply). Gentlemen's honor (we could even get vertoe to serve as escrow if you don't trust me... I am pretty new to the forum). Keep in mind, MN payments are up to 58% of the time now over the last 24 hours, so your odds are better than when I did the analysis. If things stay the way they are, it's not a bad bet... I should win about 53% of the time. And if block MN payout % goes up even more, your odds improve. But if you believe you're right and there's a problem with the MN algo and therefore you think XrfFymhBuVrFhCcZZZ8PULhHeNf7iC9vXV is more likely to keep getting payments just because he's received them in the past, you should take the bet. Up for a little fun? :grin:
In all seriousness, this "fairness" thing will give people problems... even if it is fair, human tendency will be to look only at our own nodes and decide if it's fair based on something as inane as not getting paid for three consecutive days (which is going to happen a LOT). That's why people prefer mining in pools even though the expected payouts are actually slightly LOWER. Are we going to see MN pools develop? Probably.
... this is yet another reason I more and more think we need "BlocksSinceLastMNPayout" ratio multiplier incorporated into the MN selection algo that acts similar to "coindays" in PoS ensuring a more consistent payout frequency for each MN. You'd have few if any examples of payments to the same MN twice in a row and fewer examples of going days without a payout.
Last edited by a moderator:
darklord1981
Member
My MN got paid 3 times in 4 hours on the 2nd July! So far, been paid 8 times since 27 June, and missed out on 3.5 DRK in payments due to dodgy pools. Currently looking at 29% Annual ROI, with enforcement on, this ROI figure would be 44% which is nothing to sneeze at!
That is pitiful. I feel sorry for you (assuming your MN has been operational this whole time?). You've only been selected twice and one was a non-paying block. Rotten, rotten luck.
I got 7 payments so far myself, and it's only been up since July 2nd... I was out of the country until the 1st, so I got a late start, but got lucky on the first day with 3 payments.
vertoe
Three of Nine
You should consider using a masternode shares service to even out your earning