Thursday, May 26, 2022

Cryptocurrency Catch-22

A major criticism of Bitcoin is that its blockchain processes only around 230K transactions/day, of which only about 10% are "economically meaningful. That is less than 5 "economically meaningful" transactions between individuals' wallets per minute. 90% are wash trades, and 7.5% transactions between exchanges.

As I described in Fixed Supply, Variable Demand, when the limited supply of Bitcoin and Ethereum transactions meets spikes in demand, the result is huge spikes in fees. This has made cryptocurrency boosters such as Vitalik Buterin unhappy:
Buterin didn’t predict the rise of NFTs, and has watched the phenomenon with a mixture of interest and anxiety. ... their volume has overwhelmed the network, leading to a steep rise in congestion fees, in which, for instance, bidders trying to secure a rare NFT pay hundreds of dollars extra to make sure their transactions are expedited.
The solution is obvious, greatly increase the rate at which the system can process transactions. Ethereum 2 is proposed to implement sharding, allowing parallelism. Avalanche claims 3400 transactions/sec with 1.35sec finality. Problem solved! Not so fast. Follow me below the fold.

First, very low fees sound like a good idea but they are a double-edged sword. The Solana blockchain's competitive advantage was very low fees and the result was:
On April 30, NFT minting bots began flooding the Solana network with 4 million transactions per second, causing the network to lose consensus. The project tweeted that "Engineers are still investigating why the network was unable to recover, and validator operators prepare for a restart." The network was offline for seven hours.

This is hardly the first instability the network has demonstrated, much to the chagrin of its users. Transaction flooding is an issue on Solana in part because of the low transaction fees compared to networks like Bitcoin and Ethereum, which have relatively high gas fees that would make flooding extremely expensive.
Avalanche didn't make that mistake:
In Avalanche, we use transaction fees, making such attacks costly even if the attacker is sending money back to addresses under its control.
So, the fee per transaction has to be high enough to deter flooding, but low enough to attract transactions.

Miners' income comes from both fees and the block reward. The point of the reward is to incentivize enough mining to make the cost of a Sybil attack greater than the profit to be made from a successful attack. Thus it can't be too low. But the block reward works by creating new coins, so it is inflationary. Thus it can't be too high.

But there is a further complication to this Catch-22. Bitcoin regularly reduces the block reward, aiming eventually to eliminate it and depend solely on fees. Raphael Auer showed that, as the rewards in a block become less than the fees in the block, the system becomes unstable. Another way of expressing this is that the average reward per transaction must be more than the average fee per transaction. Bitcoin handily satifies this condition at present, the total cost per transaction is around $100, but the average fee per transaction is around $2. This has the useful effect of disguising the real cost of Bitcoin transactions.

Thus we have the following constraints:
  1. Fees need to be high enough to deter flooding.
  2. Fees need to be low enough to attract transactions.
  3. Rewards need to be high enough to keep the blockchain secure.
  4. Rewards need to be low enough to preserve the value of the currency.
  5. The reward per transaction needs to be more than the fee per transaction.
If we vastly increase the supply of transactions, well above the demand for transactions, the fee per transaction that the market will sustain will decrease significantly, risking violating constraint 1. The average fee per transaction will be less than the reward per transaction, satisfying constraint 5, but the total per-transaction cost (reward+fee) will be high, risking violating constraint 4.

If, on the other hand, the supply of transactions is below the demand, the fee per transaction will spike (as we have seen), risking violating constraints 2 and 5.

Thus we need to ask in whose interest a vast increase in the supply of transactions would be? It pretty clearly isn't in the miners' interest, as we see with the history of attempts to increase the Bitcoin block size. Liimiting the supply of transactions allows the miners to profit handsomely in periods of high demand. At first glance low fees would appear to be in the users' interest, but in practice it isn't. Low fees lead to flooding, and outages like Solana's. In practice, miners control the system, so even if it were in the users' interesst it wouldn't matter. Catch-22.

1 comment:

David. said...

It is estimated that about 20% of the Bitcoins that have been mined are lost, so the actual supply is about 80% of 19M or 15.2M. If we assume that the rate of loss is constant (probably it has decreased) this is about a 1.5%/year reduction in the supply, or deflation.

Each year 52,560 new Bitcoins are mined, an increase of 0.35% in the supply, or inflation. So Bitcoin is safely deflationary and not at risk of violating constraint 4.

At the current "price" under $30K the mining reward is notionally around $1.58B/yr. A better way to look at it is that it is $187K/block. This is significantly more at risk of violating constraint 3 than it used to be.

But, as David Gerard has reported in mid-April:

"Bitcoin miner Riot Blockchain (RIOT) produced 511 BTC in March and holds 6,062 BTC. ... HIVE Blockchain Technologies produced 278.6 BTC and over 2,400 ETH in March. As of 3 April, HIVE held 2,568 BTC and 16,196 ETH. ... Miners just love holding cryptos, see. It’s not that they can’t sell them for fear of crashing the market, because number can only ever go up."

In mid-April Bitcoin was around $40K. Even if they had sold all their rewards since, the Bitcoin stashes of those two miners alone would have lost more than $86M, or about 110 days income. If they were borrowing against their HODL-ings to get cash to pay bills, they would likely have had margin calls.

The miners can't sell the notional $1.58B/year of their rewards because there isn't $1.58B/year coming in to the system from "greater fools", so they have to HODL this currently wasting asset.

This means that the actual reward per block in fiat terms is much less than $187K, significantly raising the risk of violating constraint 3. And in fact the hash rate has dropped about 10% and the difficulty decreased to match.