quote: It's not that deep: 50% = 2^(-1)
From https://eth2book.info/capella/part3/config/preset/
(Since the debate is a bit lively, I am obliged to say that this is just a shit-post - please don't shoot me.) https://x.com/banteg/status/2088649215417184419 | eip-8363 works backwards from a chosen policy endpoint:
"consensus issuance should be fully cancelled when staked eth reaches 50%."
but the 50% endpoint is not derived from a security optimum, an estimated staking supply curve, or a model of validator composition. it is defended as the "last natural stopping place", essentially a shelling point meme.
it claims favorable effects on validator composition and participant behavior from changing issuance, without actually modeling how the different cohorts respond.
there has been much more epistemically honest and rigorous analysis during the previous 2024 issuance debate. there is an awkward irony that eip-8363 authors now invoke elowsson's work as a part of its intellectual foundation, and thanking him for key insight, while presenting as settled several economic effects his faq very carefully classified as unresolved.
the truth is we don't know the staking supply curve, as we don't know at what yields different groups completely withdraw. that means we simply cannot claim to be able to derive the optimal staking ratio from first principles.
elowsson modeled multiple possible staking supply curves and asked how issuance policies behave across them. he explicitly considered mev, proposer reward variance, economies of scale, the costs solo stakers bear, the minimum yield at which participants are willing to stake (which he calls reservation yield in his research). among the questions he examines is whether reducing issuance could change validator composition in either direction.
and after all that rigorous modeling, his conclusion on whether lower issuance could worsen the solo staker share is: we don't know yet.
the contrast with eip-8363 is striking. the eip takes mathematical properties of its issuance curve and extrapolates them into claims about real-world market behavior: that large operators will become self-limiting, lower issuance will resist capture, and limiting the staking ratio will protect decentralization.
great slogans, but there is still no credible model showing how cex, etf, lst, institutional and solo stake respond to those yields. there is no evidence that delegated stake exits before solo stake. nor is there evidence that the supposed self-limiting property translates into real provider behavior. a rational provider has little reason to refuse a deposit that earns him a fee. aggregate demand for staked exposure is determined by depositors, not by any one provider, and refusing the deposit can simply send it to a competitor while the staking ratio rises anyway.
the missing research questions from 2024 didn't disappear and they haven't suddenly got resolved. yet the level of discussion has cratered so much, the uncertainties from then got promoted into assumptions.