bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 2쪽 중 1쪽
Derive Protocol
Sean Dawson Dominic Romanowski Anton Cheng Vladislav Abramov
with
Joshua Kim Jake Fitzgerald Nick Forster Timothy Gorham Ksett
July 2023
Abstract
In this paper, we introduce the Derive Protocol, a generalized risk engine running on Derive L2 - an Ethereum
rollup built with the OP Stack. A fundamental redesign of previous versions, the Derive Protocol offers a capital
efficient, modular and rapidly upgradable platform for traders, all whilst ensuring that users maintain self-custody
of their funds.
1 Introduction
Since launching in 2021, Lyra V1 [1] and its subsequent
updates Avalon and Newport have been the leading
decentralized options exchange with over $1.5 billion of
notional volume traded.
Though an incredible feat, comparing these numbers
to those in the traditional world is humbling. Volumes at
such exchanges often range into the tens, if not hundreds,
of billions per day. Derive’s volumes must therefore grow
by several orders of magnitude in order to compete with
those in the centralized space.
In this paper, we detail the Derive Protocol, otherwise
known as V2. This is a fundamental redesign of Lyra,
one which we believe will be the bedrock of a fair,
democratized and decentralized financial system that can
compete with traditional entities.
The Derive protocol can be summarized with the
following key features:
• Generalized Risk Engine: Any financial instrument
(option, dated futures, binaries, etc) can be traded,
margined, settled and liquidated on the Derive
Protocol. At launch, support will be provided for
1
Derive
Figure 1: High level schematic of the Derive Protocol.
European options (with capital efficient spreads),
perpetual futures and spot assets.
• Capital Efficiency: Traders will have access to
sophisticated margin methodologies at launch,
unlocking substantial improvements in capital
efficiency compared to previous versions.
• Modularity: Derive’s smart contracts define various
key features like account structure, risk management
and a wide array of supported assets. These building
blocks can be replaced or built upon more easily
than previous versions of Lyra.
2 The Protocol
The Derive Protocol is a collection of smart contracts
on Derive L2, an Ethereum rollup on the OP Stack
governed by the DAO. There are three main components
to the Protocol: subaccounts, managers and assets.
The Protocol is modular by design, allowing for swift
iteration, meaning many of these layers will be further
enhanced over time. For a visual representation of the
protocol, see Figure 1.
3 Subaccounts
Subaccounts are the base unit of the Derive Protocol. All
users will be able to permissionlessly mint a subaccount
entity which houses all of their supported assets. In
turn, all of a subaccount’s assets will be governed by a
manager, explained below.
Subaccounts are ERC-721 compliant NFTs which store
a list of {asset, subId, balance} where:
• the asset is a smart contract which encodes rules
that define the basic properties of an instrument
(such as the fact that an option requires a strike
price). Assets are extensible and can be programmed
for any purpose. E.g. dated futures and binary
option assets can be added at a later date.
• subId is a unique label that specifies the parameters
necessary to fully define the instrument. E.g. the
subId of an option will contain information about its
strike, expiry and whether it is a call or put.
• balance is the net balance of the instrument. This
can be either positive (a credit) or negative (a debit)
The subaccounts contract has two main functionalities:
1. Ensure that only authorized parties (the user,
2
Derive
managers) are able to change the subaccount’s
balance.
2. Inform all relevant parties about any trade involving
the subaccount.
This was made possible through the use ofhooks;
external function calls to relevant contracts that occur
during and after any asset transfer. They are:
• asset hook : this invokes the asset to validate
the transfer and determine the final balance. For
example, an asset contract may apply interest
accrued to a transfer event, as well as monitor the
time since the last such payment.
• manager hook : this calls the manager to validate
the final state of the subaccount. The managers
can enforce different rules by reverting transfers
that make the final state invalid. The manager
of a subaccount might read all necessary data
from oracles or asset contracts to calculate margin
requirements.
This architecture allows different managers and assets
the flexibility to create unique validation rules for various
subaccount actions.
4 Assets
An asset contract is responsible for determining the
results of a transfer, trade, deposit, withdrawal or
settlement. To support deposits and withdrawals, as well
as interest accrual, an asset has the privilege to update
its own balances in any subaccount.
Assets and managers are linked; certain assets will
only be supported by specific managers and vice versa.
This also means that assets can be fungible across
managers. This relationship is maintained via the
aforementioned hooks.
At launch, there will be four types of assets available:
• cashAsset: Each manager supports onecashAsset
(say, USDC). This is often referred to as thequote
asset.
• option: European options on a given underlying
asset.
• perp: Perpetual futures on a given underlying asset.
Unlike other exchanges, perpetuals can be settled
by the user at any point in time. This is referred to
as continuous settlement. Further, funding paid or
received by the user is also continuously applied.
• wrappedERC20: Wrapped ERC20 assets such as
wBTC. Balances for wrappedERC20s can only be
credits.
Assets can also assist the managers at settlement or
during subaccount state validation. For example, the
CashAsset contract stores all relevant data about a
subaccount’s accrued interest. The manager may want
to consider this interest when assessing the risk of a
portfolio.
4.1 CashAsset
Each risk manager supports a single quote asset. As with
options and perpetuals, users can hold credit or debt
balances of the quote asset. Unlike these other assets,
users with debit balances will have to pay interest on
this sum. This interest will then be distributed to all
other users with credit balances. Thus, the quote asset
facilitates a native lending market over its supported
managers.
At launch, USDC will be the only supported quote
asset.
For example, if ETH has a spot price of $2000, then
a user holding one ETH in their account can borrow,
say, $1600 of USDC using their ETH as collateral. In
doing so, they enter a debit balance and pay 5% interest.
All other users with credit USDC balances will receive a
share of this interest payment.
The interest rate is set via a simple piece-wise curve
like in Aave (see [3]). Quote utilization will be defined
as the ratio of total debit balances over total credit
balances. When utilization is low, the interest rate paid
by borrowers will be low. As utilization increases, the so
too will the interest rate. Past a certain threshold, the
rate of change in interest accelerates greatly.
5 Managers
Managers are smart contracts that specify certain rules
for subscribed subaccounts. At launch, there will be
two managers available, which ensure that subaccounts
maintain appropriate margin when transacting. These
are the standard risk manager (SRM) and the portfolio
margin risk manager (PMRM). When creating a
subaccount, the user must subscribe to a manager. We
stress that managers are modular. One major benefit
of this design choice is that risk managers can be easily
designed for any set of assets or purpose, allowing for
diverse trading and effective risk management.
Managers are given the privilege to alter a
subaccount’s balances since they are responsible for
both trade settlement and liquidations. Since managers
are smart contracts and are therefore immutable, this
privilege cannot be used maliciously.
In the Derive Protocol, risk managers fulfil the
following functions:
3
Derive
1. Setting Margin Rules: Managers set the margin
requirements of all supported subaccounts. To avoid
being subject to liquidation, users must satisfy
their maintenance marginmandate. To open a new
position, the user must fulfil their (stricter)initial
margin requirements. These rules are transparent
and verifiable by anyone.
2. Liquidations: Managers define the conditions for and
enforce liquidations.
3. Settlement: Managers are responsible for settling all
trades.
Both the SRM and PMRM will have (adjustable) open
interest caps on supported base assets, options and perps.
5.1 Standard Risk Manager
The standard risk manager gives traders an intuitive,
feature rich entry point for trading and margining
options, perpetual futures and their base assets on
the Derive protocol. The SRM supports the following
instruments:
• Quote asset (USDC at launch)
• Various base (ERC20) assets (such as wETH,
wBTC)
• European options and perpetual futures on these
base assets
Key features of the SRM include
1. Cross-margin: traders use all quote and base
balances as collateral for all open positions in a
subaccount. Isolated margin is obtained by opening
a separate subaccount.
2. Multi-asset collateral: traders will be able to use
various base assets to collateralize their perpetuals
and options. For instance, a user will be able to
collateralize a short BTC perpetual with ETH.
3. Capital Efficient Spreads: Normally, the margin
required for short options is the mark-to-market
value of the option and a percentage of the spot.
Long options provide no margin offset while base
assets are valued at a discount to their market value.
To facilitate capital efficient spreads for traders,
the SRM is able to cap the margin requirement
of spreads at their maximum loss. For example,
consider Alice’s portfolio consisting of a long $200
call spread, i.e. 1 long ETH $1800 call expiring in
2 weeks (normally requiring $0 of margin) as well
as 1 short ETH $2000 call with the same expiry
(typically requiring $320 of margin). The SRM will
require precisely $0 of margin - a huge improvement
in capital efficiency over previous versions. For a
short spread, the maximum margin required will be
capped at the maximum loss.
4. Leveraged Perpetuals: Perpetuals can be traded
using the SRM with high capital efficiency.
5. Borrowing: If enabled, users can borrow the quote
asset against their wrapped ERC20 base assets.
If relevant data feeds havelow confidence, then the
SRM will require extra initial margin for all base assets,
perpetuals and short options. In addition, if the quote
asset depegs, then further margin will be required for all
perpetuals and short options.
5.2 Portfolio Margin Risk Manager
The PMRM brings portfolio margin, an industry
standard for allowing high levels of capital efficiency, to
Derive.
The PMRM will support the quote asset, the (single)
denominated base asset, as well as corresponding options
and perpetuals. Consequently, if a user wishes to use
portfolio margin to trade ETH options and BTC options,
then two PMRM subaccounts will be required.
To compute the portfolio margin, the manager finds
maxLoss, the market loss the portfolio would suffer
under various scenarios, where each scenario is specified
by a given spot and implied volatility shock. Extra
contingency marginis added to account for risk factors
not captured inmaxLoss, i.e.
Margin = mfactor × (contingencies + maxLoss)
where mfactor is typically 1.0 when computing
maintenance margin and a larger value for initial margin.
The manager uses the following contingencies:
• Base Asset Contingency: A small percentage of the
spot price per base asset held.
• Perpetual Asset Contingency: A small percentage of
the spot price for each perpetual contract held.
• Option Contingency: For each strike, the net number
of short options is computed. A small percentage of
the spot is added to the margin requirement for each
net short option.
• Oracle Contingency: A large percentage of the spot
price is added to the initial margin requirements per
base, option and/or perpetual if the relevant data
feeds have a low confidence score.
4
Derive
There is also a fifth contingency; thebasis contingency
which computes the loss the portfolio will suffer if the
forward basis moves against the user for all expiries.
If greater thanmaxLoss, then the basis contingency
replaces maxLoss in formula for margin.
Finally, if the quote asset depegs, thenmfactor will
increase but only when computing the initial margin.
More detail on the methodologies for the SRM and
PMRM will be provided in upcoming documentation.
5.3 Marking and Settlement
Both managers mark options using theBlack76 pricing
model, see [2] for more detail.
To ensure traders can smoothly unwind their hedges,
options on both managers settle to a time weighted
average price (TWAP) of the underlying price in the 30
minutes leading up to expiration.
In this period, options will be marked to a linear
combination of the forward and TWAP prices with
weights determined by proximity to expiration. Since
part of the settlement price will be fixed, this means
options will experience so-calleddelta decay. I.e. a
normally 100 delta call 15 minutes from settlement will
only have 50 delta of underlying exposure.
5.4 Risk Reducing Trades
For both managers, a subaccount cannot open new
positions or increase the size of existing positions when
the final state has insufficient initial margin. That said,
if a transaction results in a reduction in the subaccount’s
maintenance margin requirements (i.e. the transaction
reduces risk), then the update is always permitted.
5.5 Risk Assessors
Due to the large number ofBlack76 function calls,
portfolio margin computations are very gas intensive. To
avoid limiting PMRM account sizes, entities approved
by governance calledverified integratorswill be able to
bypass all but one risk (namely, the mark-to-market,
i.e. solvency) check on chain. While it is possible for
verified integrators to approve transactions that result in
liquidatable (but not insolvent!) positions, such accounts
can be immediately liquidated by any willing participant,
removing this risk from the system. Further, the verified
integrator role can be revoked if the party in question is
found to be acting improperly.
It is important to stress that, due to the solvency
check, verified integrators can never allow insolvent
positions to be opened.
6 Partial Liquidations
Users who fall beneath their margin requirements will be
subject to a partial liquidation.
Liquidations will be a transparent process and open to
all participants, i.e. any user can act as a liquidator and
the user being liquidated will be able to monitor their
subaccount throughout this process.
The liquidation mechanism is designed to expeditiously
remove risk from the system while at the same time
ensure the most beneficial user experience for traders
being liquidated.
6.1 Liquidation Auctions
The way liquidations will work for both managers is as
follows.
1. A user is flagged as being under their maintenance
margin requirements. Any user can flag this
subaccount by calling the functionstartAuction().
2. The entire portfolio is put up for auction. The initial
offer is a small percentage discount to its mark-to-
market value.
3. This discount increases over time. At any point, a
user (liquidator) can agree to take on any percentage
of the portfolio at the current discounted mark-to-
market value.
4. To ensure liquidated users get a fair deal, the
maximum percentage a liquidator can take is a
function of the current discount, portfolio mark-to-
market value and current margin requirements. I.e.
a portfolio just shy of its margin requirements will
only need, say, 10% liquidated while an account near
insolvency will require nearly 100% to be liquidated.
5. The discount increases until a sufficient amount
of the portfolio is liquidated (by potentially many
liquidators) or the discount reaches 100% (where a
liquidator can take on up to all of the portfolio for
free).
6. An insolvent auction then begins, where bidding
starts at $0 and decreases up to a large (negative)
value based on the subaccount’s initial mark-to-
market value. Liquidators can always take up to
100% of the portfolio and get paid by the Security
Module (see below) based on the current (negative)
offer.
6.2 The Security Module
The Security Module (SM) is a smart contract that will
hold reserves of the quote asset (i.e. USDC) to pay out
funds to liquidators in the event of an insolvency.
5
Derive
Should the SM run out of funds to do so, the insolvent
debt will have to be shared amongst all users.
This will be done via atemporary withdrawal fee. The
insolvent debt of the system, as a fraction of sum of the
insolvent debt and total deposited (i.e. wrapped) quote
asset, is applied as a fee to users wishing to withdraw
their quote asset.
For example, say there is $10,000 of bad debt and
$100,000 of USDC in the system. A fee of
10, 000
100, 000 + 10, 000 = 9.09%
is applied to all withdrawals. So, if Alice wants to
withdraw $5000, then she pays a $454.5 withdrawal fee.
The temporary withdrawal fee provides no incentive
for being the first to withdraw funds and also prevents
liquidation cascades, thereby increasing the system’s
integrity.
7 Protocol Fees
Every risk bearing entity in the architecture can receive
fees for doing so. The protocol has three main sources of
fee generation:
1. Trading fees: All transactions that increase the open
interest (and therefore risk) of the system will be
charged a fee. This fee will be a function of the spot
price of the underlying asset and size of the trade.
2. Liquidation fees: Users who are liquidated will be
charged a small fee that scales with the risk and
market value of their subaccount.
3. Interest rate spreads: A percentage of all interest
paid by quote asset borrowers is taken as a fee.
All fees will accrue to the Protocol owned Security
Module which improves the robustness and resiliency of
the protocol.
Entities approved by governance can bypass the
trading fee in exchange for, say, a different fee scheme
pledge. These approvals can be rescinded if the entity
does not meet its requirements.
8 Parameters and Governance
There are many parameters that govern the protocol; this
is especially true for the managers. These parameters
also have to be updated whenever there is a market shift
to ensure both capital efficiency and security for traders.
All parameters in the protocol are controlled by a
smart contract. In Derive’s case, the smart contract will
be owned by the short term executor. For details on the
short term executor, see [4].
9 Data Feeds
The managers, in order to accurately assess and control
risk, require oracles to provide the following data:
1. Implied Volatility (IV): IV for supported tenors will
be provided as raw stochastic volatility inspired
(SVI) surfaces. These are specified by 5 parameters:
(a, b, ρ, m, e). See [5] for more detail. Consider a
strike K with log-moneyness k, where k = log(K/F )
and F is the current forward price with time to
expiration τ. The implied volatilityσ of strike K
is given by
σ = 1√τ qa + b[ρ(k − m) +p(k − m)2 + e2].
2. Forward: The forward price for supported tenors is
also supplied. As mentioned, close to expiration, this
becomes a linear combination of the TWAP of the
underlying and forward prices.
3. Risk Free Rates: This is an interest rate curve
defined at each supported tenor.
4. Spot: This is the spot (index) price of the underlying
assets sourced from multiple outlets.
5. Perpetual Price: This is the trading and impact
prices of the perpetual future.
Each piece of data supplied also has an associated
confidence score. Low confidence feeds result in extra
initial margin being required for both the SRM and
PMRM.
More detail on the methodology for deriving the
implied volatility and forward bases will be announced
shortly.
References
[1] Lyra Whitepaper. 2021, S. Dawson, N.
Forster, D프로젝트로 돌아가기