What network-statistics will be available?

Recently I have read here that it will be impossible to know the exact number of safecoins issued by the network over time. That will make it obviously impossible to determine a marketcap for safecoin. For many folks used to bitcoin etc, that’s gonna look rather sketchy given that for other cryptos the number of units issued is exactly known.

More broadly speaking: given the anonymous and private nature of the SAFEnetwork (which is obviously what is great about it), it seems like it will be hard to grasp certain data about it, which would be good to have though for statistics such as:

  • number of vaults online
  • rank distribution of vaults
  • geographic location of vaults
  • get/put requests over time
  • total storage available
  • total number of safecoins issued/recycled
  • distribution of safecoins over vaults
  • computational capability of the network
  • average cost/revenue to be expected per GB of storage used/provided

etc….

It would be useful to have these data to monitor the growth/health of the network over time and see it evolve. Clear statistics would also further adoption through transparency (people like to look at things, seeing is believing).

So my questions is: what statistics will be available for this network?

Happy holidays, everyone :slight_smile:

1 Like

I’d like to know this as well. The issue is of course that every node connected to SAFE can only see a tiny part of the network; the other nodes that are close to it in the XOR address space.

As far as I know most (if not all) of the listed statistics cannot be exactly known, but accurate estimates can be made through statistical analysis. For example the collision probability when trying to mine a SafeCoin can be extrapolated to get quite close to the total supply of SafeCoin. If you take the probability stats of several big vaults over a period of time, so you have a sample of many thousands of farming attempts, the margin of error should be very low. Since a collision will occur if the SafeCoin to be farmed already exists, a collision probability of 0.5 indicates that 50% of all 2^32 SafeCoins currently exists.

In a similar way, the actual distances of addresses of a close group can be used to make a good guess about the number of nodes on the network. The smaller the address range of a closed group is, the larger the network must be.

I think that all statistics need to be estimated in such a manner (using data of your close group, or for a larger sample, the close groups of your close group nodes), for the simple fact that there is nothing in SAFE that knows the entire network, and polling all nodes is practically impossible once the network has matured. It is possible to make use of a blockchain-like approach to gather exact data, but if that would be mandatory core SAFE functionality, I think that would impose scalability limits on a network that currently doesn’t have any. Which, merely for the sake of exact stats, is not worth it.

2 Likes

I agree this would be great and am still interested in building or helping build this into an app - I posted an idea for one about farming setup stats a while ago.

I think there will be stats that can be accurately known too - don’t have details, but think the are techniques that aren’t just statistical.

I for one like how things are anonymous and unknown by us.

Keeps things more secure

This would be important for maintaining an accurate position coinmarketcap too.

I might suggest a general answer. The original internet design is decentralized and for example the ‘exact’ number of connected computers to the internet is an unknown quantity. In many ways this will be similar for the ‘exact’ number of connected vaults.

However, an exact number is produced by a counting exercise. A statistical estimation is produced by sampling the network and calculating the expected values for the whole from those samples - as we do in every science. So it should be rather straight-forward to get an estimation of the number of vaults, for example. Just connect a large enough number of vaults, in parallel or sequentially under the assumption that the total number of vaults will vary insignificantly during the measurement period), and you will see the vaults in your direct vicinity. Measure how far and spread out they are, i.e. calculate a density, and because we know how big the address space is (2^512) the total number of connected vaults can be calculated with a margin of error that can be made arbitrarily small by increasing the number of samples taken.

Hope that can help the discussion

2 Likes

We could actually consider making a SAFE app that users can voluntarily run to sample the network and gather the raw data for producing open statistics about the SAFE network.

4 Likes

I think that’s a good idea, especially to arrive at a total supply figure to derive the market capital. An API can then be provided for websites such as coinmarketcap.

Btw, now I wonder how the SafeCoin PUT price is determined? Is this also done per close group(s)? In that case the PUT price coud vary between different groups, which seems odd to me if the difference is big enough.

1 Like

That is a question that I would also like the answer to.

There is a lot to it, but in short. Here is an answer to a question similar

1: It is a sigmoid curve, all vaults earn. Less fast to start, increases up to average, levels of at 20% above average. So huge vaults massively over average are not linearly earning more, vaults are incetivised to get just over average, but not more.

2: Rate is 2 step,
a: The rate per X Gets, X is calculated by network to increase safecoin when there is a space deficiency and decreases this X on over abundance of space is available.
b: Rate above is regulated by the Sigmoid curve

3: More than network average is to ensure there is always space to store, bit not too much.

3 Likes

Thanks, interesting question and answer, but it addresses the GET rewarding, not the PUT prices?

The PUT price is balanced with the farming rewards. So the cost to put is balanced with supply, oversupply makes puts cheap and under-supply more expensive. On the other side farmers get less when oversupply and more when under-supply.

So it is the bad part is that it is very hard to estimate, the good part is that it is hard to estimate. (good luck to ben bernankie and co-horts this is gonna be different :wink: )

For me this is the missing link in crypto (and economics in general), how does the supply of the engine oil happen, as the engine is getting dry we need to create some (safecoin costs to store are high, this makes the farmers provide that oil) and when the oil reservoir is nearing capacity non of us want to buy oil and the price goes down (safecoin buys a ton of space when the supply is excessive).

It feels more true to a market than the manipulated versions we currently have with excessive variables when money is printed or produced without consideration of anything except an algorithm and stopwatch or similar.

I find this better than linear inflation, fixed supply derived from time (10 minute blocks etc. even 12 second blocks) or proof of stake etc. I think these can all suffer from the rich centralising. I doubt the rich will be the only suppliers of the worlds data (information and knowledge) and as we spread education though access to all info this naturally decentralises the supply/demand across the planet (hopefully).

Then people say we are tying data to currency, but we are not, we are tying supply rate to demand for data services (storage initially and hopefully computation later, by computation I mean being able to ask the network questions (see opencyc for some ideas on where this could go)). I find this interesting we tie knowledge to supply rate and as we crave knowledge then it gets better for farmers (more expensive), Knowledge base will always grow and the thirst is not being satisfied today, I hope to help change that. No more burning libraries and killing scientists, this time it will be different.

Sounds weird, OK now dive further. We pay once for data to be stored forever, so nobody does not have access to everything, but they start adding to the pool and as that happens we need more space (get the farmers rewarded for their good deeds. So information becomes our yardstick for supply balancing the ‘economy’ if you like. As we move forward we will add to this data in massive amounts (historically we do) but with better and cheaper storage, so this balances as well. Computation will become very important as time goes by and this will add to the model so not only silly websites, but information hubs so not search but question, question the network for qualified answers not search and sift and certainly not tailored to your buying habits search.

If we get this right then much of this will all start to play out and explaining it will get easier. @nicklambert is writing some up now, which is no easy task at all, but he is the man for this one. I have some much further reaching ideas and observations on much of this, but not enough time to explain it with anything like the detail it deserves. Soon, though soon.

This is another reason we work very hard on churn (smaller account transfers, faster routing messages), I want farmers to be all of us and at no extra costs other than the current computers we have, then we amortise this wealth and with it the knowledge base.

5 Likes

I should add, if the external price shoots up, then it makes no difference, if farmers are already rewarded with tiny bits of safecoin (as the oil is very rich and we need less) then that is cool. The common misconception is that external influences affect this. They do not, if farmers are not rewarded supply goes down, the network does not care what safecoin is worth in $ or similar it just measures it needs more space so makes that happen by farmer rewards. It does that by charging more for PUT’s (we are under-supplied so there is a need for space by the people)

The supply/demand remains the same. If farmers are not rewarded costs go up for storage and the farmers are rewarded more. The number of safecoin does not matter, only the effect of the number supplied to farmers and available space being maintained.

Again it balances.

4 Likes

Thanks for the reply! I fully agree with your views on this market mechanism, I already knew most of the general idea behind it and have been enchanted with it since I first heard about it.

Thanks, that’s what I wanted to know. According to the whitepaper the farming rewards are based on the stats of the close group:

group_average is average healthy_space among the close group the vault belongs to

So the PUT price will be close group-based as well, which means that different groups can have somewhat different costs and rewards. Is it perhaps possible to use the close group’s close groups to calculate the group_average figure? That would increase the sample size exponentially (instead of 32 vaults 1024, or instead of 64 vaults 4096), and would drastically reduce the deviation from the network-average (for example instead of a 12% margin of error a 1.5% margin).

2 Likes

Yes it is the full 64 nodes (routing table size) minimum (there is an opportunity for better by using group heuristics). We have also considered some further mechanisms (I doubt we will need them, though) As rank is also amortised and groups rearranged on rank (expel nodes with rank values far from the group average and they need to find another group where they bring it closer to the average) then this ‘estimate’ is much more accurate than it would at first appear. Average rank across groups should always converge to a network wide average. In essence we get the whole groups average and then improve it.

6 Likes

We indeed have several mechanism (not yet worked out in detail) to increase the sample size. It is important, as @Seneca, points out to increase the sample size, because smaller samples are inevitably more volatile. It is important to suppress the urge to externally from the network decide on a ‘global’ fixed price for PUT.

Like David said, the PUT price will be a statistical average, but we can force it close to a global network-constant, dynamic (not constant in time though) value. When the (calculable) error-margins on a given groups estimate of the current PUT price cross a certain threshold, one possible mechanism is for it to poll random locations in the network for their current estimate of PUT price.

The actual mechanisms when implemented will no doubt be different, but one of the important features is that the network will always average out fluctuations and hence push towards a smooth price curve. This is contrary to market forces where the incentive for profit always seeks out to amplify price differences.

2 Likes

Will we see a real-time “Network Average” statistic displayed when creating/running a vault? This gives farmers an indication if they are providing too much space (resources), or need to add more. Finding the “sweet spot” (100% to 120% of NA) makes farming more… interesting, competitive, desirable, fun!

2 Likes

Neat idea, would also encourage nodes (farmers) to split when appropriate as well. I will add to the list of to look at, hopefully in testnet3 we can add these things to try and see how they pan out.

5 Likes