Thread by @tarunchitra

For those of you who were annoyed at having to read the Staking vs. Lending paper on DocSend — it is finally on @arxiv!

I'm excited to present it at SBC 2020!

In the spirit of @vgr:
1 like == 1 🥵🔥🌶 take about DeFi and security

arxiv.org/abs/2001.00919

@arxiv @vgr 1. On-Chain lending, by allowing for programmatic loans to instantly pass underwriting and settlement, tend to compete with their underlying security chain, as capital flies to the best combination of yields and security

@surja795

2. Staking derivatives (such as proposals from @sunnya97) and on-chain lending are very similar in spirit, but have very different pricing / volatility mechanics. This means that they could combine in a _cancellative_ fashion to preserve PoS network security

@camlCaseTech

3. DeFi on Layer 2 will be less likely to cause bank runs, as channel close/open fees should go up proportional to the total loan size versus main-chain DeFi requiring a single txn (or meta-txn) for arbitrarily large loan size

@AlokVasudev

4. Composability in DeFi leads to unclear exposures (e.g. SNX/MKR manipulation on 12/25/19) make it hard for participants to price their portfolios. One can synthetically perform oracle manipulation via other contracts (e.g. modify UNI price -> affect SNX mark)

@teo_leibowitz

5. PoS systems, unlike PoW systems, force users to view their assets as a portfolio with constantly rebalancing opportunity costs. PoW with liquid hash power derivatives will have this property, but so far this hasn't been super liquid (c.f. @tzhen)

@UthsavC

6. Strictly speaking, DeFi claims on PoS vs. PoW are not equivalent as bearer liens. This is because without finality / BFT, the liens have a strictly greater expected time of coexisiting on two forks (c.f. @dmoroz's work on @tezos forks and longest-chain PoS)

@tzhen

7. In a world with no slashing, there is still a raison d'être for professionalized validator pools that is the same at with mining: income smoothing. Since we have, to first order, Var[Returns] ~ (stake %)^2 / EpochLength, delegates prefer to delegate to bigger pools

@matts783

8. DeFi contracts and Child-Pays-For-Parent are incentive incompatible as this front-running across blocks is way easier (asymptotically cheaper). You can place transactions with no intention to fill and cancel w/ child chains (akin to @CFTC's def. of spoofing)

@JeremyRubin

9. Hybrid PoW/PoS systems (e.g. Decred) are unlikely to be able to get around the capital flight issues of staking until some critical threshold that is at least >2/3

@Leorzhang

10. An impossibility theorem to prove (maybe @nmushegian's grant money can go towards this 🙃):

Cross-Chain Swaps (cc @_prestwich) require collateral qties on both chain >= (1+ϵ) * min(colA, colB), where colX is collat. X's value in a numéraire to inherit security

@niraj

11. Extending this to a PTAS-like conjecture (for @_prestwich, really!):

CC Swaps can inherit δ * MEV if the collateral value >= (1+sqrt(δ) * min(expected block time)) * min(colA, colB)

@jubos

12. Cross-Shard DeFi front-running is going to be fun 😈 Validators can route cross-shard txns in a way to make front-running hard to decipher. This will also be true in L2

@badcryptobitch

@badcryptobitch 13. Routing attacks on Layer 2 will basically reinvent a lot of HFT-style arbs in which ephemeral nodes join the network, open, and close channel to spoof demand. These strategies are hard to prevent in hub-and-spoke topologies, easier in random graphs (or Kautz graphs)

@hasufl

@badcryptobitch @hasufl 14. Impossibility theorem #2 (@nmushegian might find this one more interesting)

Let 𝛾(duration) be the yield curve for an uncollat. algorithmic stablecoin (e.g. @basisprotocol, @CeloHQ w/o the crypto reserve).

Claim: With probability one, Var[𝛾(d)] = ∞

@mstrakastrak

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak 15. Uniswap Theorem: There exists a function space, contained within a Sobolev space, that defines the set of *all* admissible (admit no-arb equilibrium) automated market makers (e.g. @UniswapExchange / @BalancerLabs). Oh wait, @GuilleAngeris might be doing this 😈

@valardragon

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak @UniswapExchange @BalancerLabs @GuilleAngeris @valardragon 16. Second price auctions (or kth price uniform auctions) will not improve front-running in DeFi. Part of this is c.f. @phildaian's Devcon V talk and part of this is because sharded mempools will have more complicated keys than (gas, timestamp)

@zmanian

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak @UniswapExchange @BalancerLabs @GuilleAngeris @valardragon @phildaian @zmanian 17. Trad. Markets: Futures == leverage, Options == hedging. Recently proposals from @opyn, @synthetix_io, @AugurProject, @Flux_Market, @MonteCarloDEX, etc. have made me think that in DeFi

Futures == Oracle Leverage 🔮
Options == Unsecured Lending 🏦

@AlexisGauba

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak @UniswapExchange @BalancerLabs @GuilleAngeris @valardragon @phildaian @zmanian @OPYN @synthetix_io @AugurProject @Flux_Market @MonteCarloDEX @AlexisGauba 18. Something that @alexhevans and @_charlienoyes will fight me about: Put-Call parity doesn't exist in decentralized options because the vol surface has...
a) E[δ^k] = ∞ for k > 1
b) E[𝛾^k] = Ω(1/k)
because of the gas price auction

*ducks*

@snarkyzk

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak @UniswapExchange @BalancerLabs @GuilleAngeris @valardragon @phildaian @zmanian @OPYN @synthetix_io @AugurProject @Flux_Market @MonteCarloDEX @AlexisGauba oops, wrong Opyn: @opyn_

@badcryptobitch @hasufl @nmushegian @basisprotocol @CeloHQ @mstrakastrak @UniswapExchange @BalancerLabs @GuilleAngeris @valardragon @phildaian @zmanian @OPYN @synthetix_io @AugurProject @Flux_Market @MonteCarloDEX @AlexisGauba @alexhevans @_charlienoyes @snarkyzk 19. There exists an alternative to a mixer that combines a @compoundfinance style bonding curve lending contract with a mixer and/or ZK proof generation contract to directly incentivize people to increase the anonymity set in private txns via on-chain lending.

@zooko

20. Optimizing allocations of a staking portfolio (qty staked, qty lent, qty in passive trading, qty in active trading) is only well approximated by the Hull-White model with zero trading fees.
∴ only exchanges that validate can ctsly rebalance staking assets

@BinanceLabs

21. @UniswapExchange at scale is going to need fees driven by volatility and perhaps higher order Greeks — but it is so cool to see the bowels replicating portfolio theory running in practice on millions of 💸💰💵

@jmo_mx

22. Futures markets on gas costs and other forms of variance reduction will provide better incentives for privacy than ceremonies and mixers over the next decade.

@AnnaRRose

23. Bribing endorsers (akin to @dmoroz's paper arxiv.org/abs/1912.02954) works against HotStuff, PaLa, and Tendermint in a statistical sense — you need to do it within 2 rounds + cause a liveness fault.

@buchmanster (and really, @ittaia + @LibraDev too, but I'll get y'all soon!)

24. There’s a threshold at which staking derivatives help with liquidity — afterwards, you get things like the Uniswap/SNX Christmas arb. Computing this is NP-hard!

@tomhschmidt

25. The introduction of fixed rate on-chain lending products (variable rate + swaps) will cannibalize PoS security almost as much as variable rate on-chain lending products.

@0xlildex

26. Interspersing contract level auction mechanics (e.g. Multicollateral DAI collateral auctions) with mempool level auction mechanics will lead to unstable equilibria if they have two different auction clearing types (e.g. first price, combinatorial)

@ArthurB

27. DeFi on Libra will be interesting because you can use contract events to automate front running the reserve rebalancing mechanism without needing an oracle! You’ll need an oracle for other chains

@b1ackd0g @LibraDev @Libra_

28. There will be an on-chain lending vehicle that:
a) facilitates coin voting bribery (e.g. coins are borrowed just for a vote)
b) prices it’s yield in terms of the cost of corruption / collusion (e.g. PNL sharing)
c) uses a convex bonding curve

@AlexandraBerke

29. Gassy Girsanov Conjecture 🤢:
In the limit of infinite size mempools and a gas price distribution with unbounded support, there exists a risk-neutral measure (with rate equal to inflation) for the gas price stochastic process

@alexhevans

30. Capital flight from poorly designed automated market makers will be facilitated by interest rate swaps and Eurodollar-esque products. If only you could do an on-chain prorata order book 🥱

@Alexintosh

31. Native on-chain DEXs will be good for price discovery, but just as we saw rival centralized futures exchanges mucking around with each other’s oracles, I expect malfeasance to abound

@aeyakovenko

https://twitter.com/aeyakovenko/status/1214730567826202625?s=21

https://twitter.com/aeyakovenko/status/1214730567826202625