This writing has been deprecated by full analysis available on my Substack.

https://dataalways.substack.com/p/partial-withdrawals-after-the-shanghai

Jan 30, 2023 edit:

Looking at the actual data, the numbers presented here are an underestimate. The original estimate assumed a linear increase in adoption of 0x01 credentials, but in reality Lido has been the largest setter of this option. Lido alone accounts for 66% of 0x01 validators, which means the rewards profile was front-loaded to Spring 2021 when Celsius and 3AC were deploying recursive stETH trades.

The rewards are underestimated: 95,000 ETH vs ~ 250,000 ETH, but since the excess weighting comes almost exclusively from liquid staking derivatives, the resulting sell pressure should be comparable to the original estimates.

As is tradition in the Ethereum community, people have been throwing out basically random numbers about the sell pressure from partial withdrawals.

The big factor that many people are missing in their analysis is the difference between validators who set 0x01 credentials, and those who did not. The Withdrawals FAQ dives into the difference, but the basic story is that with the evolution of the consensus specs, genesis validators and other early depositors didn’t set their credentials for automatic withdrawals and will need to wait until after withdrawals are enabled to update them (my understanding is that these changes are rate limited).

Per Christine Kim, as per Ben Edgington, as of January 9, 2023, there were 186,722 validators with 0x01 credentials (of 488,876) that will receive automatic partial withdrawals.

New validators started setting 0x01 withdrawal credentials in the fall of 2021 (the earliest I found with a brief search was validator in October 2021). At this point in time 0x01 credential setting was very rare, unlike today where it is very common (although still not required).

If we assume that effectively zero were set before that date, and that since then we’ve seen a linear increase in the rate of 0x01 credential setting, then the centroid of 0x01 deposits is at 70% of the time passed, or about 140 days ago.

Look at a sample of validators who deposited their ether to the beacon chain 140 days ago, they have on average 32.5 ETH, and therefore should receive automatic withdrawals of 0.5 ETH. Total automatically skimmed rewards should therefore amount to approximately 94,000 ETH (as per January 2023). This number will increase up until the Shanghai upgrade; current base rewards yield about 4% per year. With Shanghai estimated at 3 months away, validators should accrue 1% of their effective balance (32 ETH), ie: 0.32 ETH before withdrawals are enabled. Therefore, we should expect the 94,000 ETH number to grow to approximately 150,000 ETH.

However, since October 24, 2021 the breakdown of staking deposits has been as follows:

• 51% to liquid staking derivatives

• 20% from whales

• 16% to staking pools

• 8% to centralized exchanges

• 5% from others

A more accurate breakdown of automatically skimmed withdrawals should therefore fall between:

• 73,500 ETH if we only remove liquid staking derivatives.

• 37,500 ETH if we only consider whales and *others--*making the assumption that centralized exchanges and pools will likely not have their shit together.

Hal Press assumes 25% of skimmed ether will be sold, which gives us a range between 9,500 ETH and 19,000 ETH of sell pressure from this event. At a price of $1,600/ETH this amounts to: $15 million to $30 million dollars.

I think this estimate is a little low, but either way we're back to full withdrawals likely being the dominant factor in terms of sell pressure. The churn limit will probably be 8 at the Shanghai upgrade, which corresponds to 1,800 validator exits per day or 57,600 ETH.

These withdrawals will also be staggered throughout the day and we’re unlikely to see a coordinated dumped from people who are withdrawing. That said, the FUD shall undoubtedly continue.