We need some mechanism to initially distribute our new tokens to the participants of the fathom network. The classic way to do this is through a token sale, simultaneously raising funds for a network and distributing tokens to those who value them.
However, it is difficult to accurately price tokens (so that they go to those who value them most) and it is harder to ensure that those who maximally value the token will be most in line with the long term goals of the network, especially because they may have speculative goals.
It is also potentially dangerous to immediately distribute all the tokens in an economy in a "drop" event. It makes the network less responsive as it grows, cutting off valuable opportunities for regulating token mechanics.
We want a continous distribution mechanism that allocates newly minted tokens through a randomized process, weighted by participation in the network.
A good metric for participation is the amount of tokens a user has spent or risked in assessments. These are staked schelling game where current participants who have credentials validate those who want to prove they do. If you want to learn more about it you can read our whitepaper
In order to distribute the tokens we hold a lottery every Y blocks (an epoch) for X tokens. At the start of each epoch a target hash is set as the hash of the first block. In the next Y blocks any user can submit a "bid", a hash of:
Their address
The address of an assessment they partipated in, to completion, within the epoch.
An integer <= the amount of tokens they spent or risked in that assessment.
The salt of that assessment (explained below).
All these values are fixed for a given assessment, but an individual can vary the value of the integer they submit. This gives them a potential hash for every token they use in an assessment.
The salt is to ensure that bids are not deterministic when an assessor decides to accept the assessment, and is derived hasing and then XOR-ing all the salts assessors submit in the commit-reveal phase.
The winning hash is the one closest to the target for that period and the winner gets X tokens.
The current closest hash is stored in the contract during an epoch, so that an individual only has to submit their hash if they know it beats it. Eventually this can be taken care of automatically for users by a script that listens for completed assessments.
You can view the first implementation of the contracts here.
The main line of attack is control of either the target hash or one's bid hash.
The former is easily influenced by miners but because it is set before any bids are allowed, it isn't meaningful to exert control over it.
For the latter an individual knows their address, the address of the assessment, and the amount of tokens they are risking, before they accept the assessment. However they cannot compute the salt and so cannot control the hash.
However because this is a commit and reveal scheme, the last assessor to commit can exert some control by decieding whether or not to reveal. In this case though if they choose not to reveal, they lose their stake and are no longer assessors for that assessment and hence cannot submit any hashes based on it, nor will their salt influence the final one.
One shortcoming of this approach is that an individual needs to have tokens in order to earn tokens. We don't see a way out of this without relying on external, arbitrary mechanisms.
Ideally the network should aim to have a reasonable amount of inflation, taking into account tokens burned during the assessment process. This is to incentivize and allow for spending over holding, and to drive activity in the network.
To achieve this the parameters of the systetm will likely need to be tuned as it runs. Initially the token reward will not matter while the epoch length will be key, but as the system grows this will reverse.
The ultimate goal of all of this is to grow the community and create scalable infrastructure for people to interact with and around. It's not likely we'll get it right on the first try, but we think this is a good starting point.