Token distribution in perspective: Livepeer’s merkle mine not as successful as portrayed

Token generation & distribution has always been a top concern for anyone seeking to establish truly decentralized networks. We’ve seen many methods attempted, but it remains quite challenging to distribute tokens fairly to real users (not speculators) while avoiding concentration of wealth.

So as a token designer, I was very excited when Kyle Samani of Multicoin published A New Model for Token Distribution, praising the innovative work by Livepeer in developing merkle mining, a supposedly significantly superior model. It claimed to get tokens into the hands of real users while producing a significantly lower Gini coefficient. Analysis revealed these claims are inaccurate both for Livepeer specifically and for merkle mining generally. This article will deep dive on the effects of this method of distribution and propose some industry guidelines for future network launches. If you’re unfamiliar with merkle mining, I would recommend reading about how it actually works here.

Gini coefficient in perspective

Multicoin’s article opens up with the following graph, showing the at-times extraordinary wealth concentration among a few select coins.

However, I think it’s also important to show how many addresses there are in total so we can see what percentage these 500 comprise. (Disclaimer: this is an imperfect measure as wallets =/= people, but we make do with what we have.)

Livepeer has so many wallet addresses (even more than Litecoin, which was started in 2011) because:

1. Livepeer airdropped Livepeer Tokens (LPT) on all addresses with >.1 ETH that sent a valid tx over a 3 month period
2. Then, it pooled the remaining LPT and enabled merkle miners to in essence claim a reward for distributing tokens to additional addresses that met the >.1 ETH requirement, but did not send a valid tx in the 3 month period

The resulting distribution among 2.6m wallets appears at first glance to be a highly decentralized distribution. However, a deeper dive into actual holdings and token mechanics dispels that notion.

Resulting token distribution

The top 50 addresses own the lion’s share of tokens relative to the next 450 addresses. This still appears to put Livepeer’s Gini coefficient lower than ETH’s and miles ahead of the comparison ERC-20 tokens. However, this neglects two facts:

1. LPT does not trade on any custodial exchanges (meaning ownership is more centralized) and,
2. the remaining 52% of the token supply has been effectively burned (taken out of circulation) because most of these addresses will ignore their LPT, whether it came from the initial airdrop or subsequent merkle mining

“In the Livepeer network, the founding team keeps 12.35% of the Livepeer token (LPT) supply, which is on the lower side of allocations we have seen throughout late 2016 and 2017. Early investors, in two rounds, have purchased 19% of the token supply in order to fund the early stages of the project. These early investors and the team are on an 18-month and 36-month vesting schedules, respectively.” — “Livepeer cryptoeconomics as a case study of active participation in decentralized networks” by Jake Brukhman, CEO at Coinfund

Out of the 48% owned by the top 50 addresses, about 31.35% is owned by the team and early investors. That leaves 16.65% of the “active” token supply in currently unclaimed territory. However, I would venture that it is highly likely that a sizable portion of that 16.65% is also held by insiders and early investors because:

1. Being already invested, they had every incentive to double down and obtain more LPT at discounted prices (paying only for gas costs) (Case in point — Disclosure: Multicoin Capital was an early LPT investor, participated in the merkle mine, and owns LPT tokens.)
2. The project was not well known and the merkle mine was not widely marketed, so the pool of people who knew about the project and had the financing (to pay for gas costs) and technical skills to participate was very small

Livepeer’s inflation’s impact on token distribution

Compounding the already centralized token ownership is Livepeer’s inflation. In short — if you are not performing work on the network (transcoding) or delegating your tokens, you will not keep up with inflation and your token’s effective purchasing power will diminish.

The Livepeer network currently has a participation rate of 19.62% (% of tokens staked), resulting in an inflation rate of .0752% per round (~daily), which comes out to an approximate effective annual rate of 31.57%, assuming the round rate remains constant. The rate of inflation actually increases until participation hits 50% so annual inflation could be even higher than that. This quickly and significantly dilutes any tokens not being staked.

If we accept that:

1. most of the airdropped / merkle mined tokens (not the miner rewards) are essentially out of circulation,
2. and that the top 50 wallets will be the ones most likely to stake,

Then in one year, it is highly likely that the token distribution will continue to centralize even further, creating a cartel of a few wallets and even fewer transcoder spots. Contrary to Multicoin’s findings, this is by far one of the most centralized distributions I’ve seen to date.

Evaluating merkle mining separately from Livepeer

Although I do not believe Livepeer’s distribution was successful, merkle mining needs its own review. Merkle mining claims to be more fair in that anyone can participate (self-selected distribution) and it puts a real cost on the miner in the form of ETH gas fees, theoretically forcing them to sell a portion of their mined LPT to cover mining costs (forced distribution). This is akin to bitcoin miners selling a portion of their block rewards to pay for electricity. Unfortunately, neither of these claims hold up upon review.

Self-selected distribution

“Payment tokens should look for the widest distribution possible while work tokens should look for concentrated distribution to those wishing to purchase or provide a service to a network for payment.” — Gregory Rocco, Lead Strategist at Alpine

LPT is undoubtedly a work token, so does merkle mining by its nature target the right audience for token distribution? No. There is surely overlap between people willing to merkle mine LPT and those willing to become dedicated transcoders, but how many of the 800 merkle miners fall into that category? And for other networks that require different types of work, how likely is it that merkle mining will target that user base effectively?

Merkle mining is potentially even more prone to centralization because there is no way to prevent a whale from mining an outsized amount of tokens, as Kyle pointed out (his chart below).

Forced distribution

“Of the major cryptocurrencies, Bitcoin has the lowest Gini coefficient. This is at least partially due to age, because early holders slowly sell over time. However, a more important consideration driving concentration of wealth is that Bitcoin is mined using proof-of-work (PoW). In PoW, miners incur fiat-denominated costs, and so they must sell at least some of their mining rewards to keep the lights on.” — “A New Model for Token Distribution” by Kyle Samani

The bitcoin analogy is apt because when the network first launched, no one was selling bitcoin to keep the lights on — there were no buyers! The electricity costs had to be eaten by the miner for continued bitcoin accumulation. It took several years for a sufficiently thick order book to develop at which point the mining economies we see today became possible.

The same is true of merkle mined tokens. There is simply no market demand for these tokens at time of distribution. There is no liquidity. Livepeer is still not listed on exchanges and has thin trading on DEXes like Radar Relay. In short, merkle miners were investors. They were willing to eat the gas costs to accumulate an asset they believed would earn them a future return.

Distributing tokens to real users is hard

The question every project should be thinking about is how to get tokens into the hands of real users. My team and I are working on a methodology now, but in the meantime, please see the following guidelines for how to best go about it.

Separate project funding from token distribution

Those willing to finance a risky investment and those willing to actually use tokens are usually not the same individuals. Can you imagine if someone wanted to invest in Uber so they could take many rides for very cheap a couple years later (since they got in early)? Projects need to seriously consider alternative funding mechanisms to ensure long-term network viability.

Distribute tokens very close to mainnet launch

With work tokens, it’s not about “getting in early.” The network should be up and running, and people that want to perform work should be able to calculate how much money they’re putting in and how much profit they’ll be making from it. Same goes for users. The incentive to have the token needs to be from the utility they will derive by using it, not the price appreciation they hope to get by holding it.

Have all sides of the marketplace participating from day 1

With n-sided marketplaces, it is critical to have all sides be present from the very beginning to mitigate the chicken-and-egg problem. Jumpstarting regular marketplaces is already very challenging, but if you added the friction of requiring your own token, it becomes even harder.

Target your distribution towards the most likely users and workers

Numerai already had users on their platform so they just gave everyone some tokens, proportional to how active they were on the Numerai platform, which could now be staked to combat sybil attacks and increase worker quality. Centralized as hell, but the point stands: identify your user base and get your tokens in their hands.

Every project should be proactively identifying their most likely users and targeting them for distribution. Not every project will have an existing user base, but there are websites, communities, mailing lists, meetups, etc. devoted to almost every demographic and interest. It is highly likely that projects will be able to identify and target specific user demographics and that these efforts will lead to greater network adoption.

Conclusion

While the merkle mine was a noble attempt at improved token distribution, there is still much work (pun intended) to be done to identify a truly optimal methodology for long-term network health. However, I am very optimistic we will figure it out because it’s in the interest of every single project to do so.