Skip to content
DeFi Reference

Perpetual futures — funding rates and venue design

1. TL;DR

Section titled “1. TL;DR”

A perpetual future (“perp”) is a derivative contract with no expiry date that stays pinned to the price of an underlying asset through a periodic funding payment between the two sides of the trade, rather than through delivery. Dave White’s four mental models — thermostat, PNL loan, loan swap, and one-day future — all describe the same instrument, and the same funding-based mechanism turns out to price stablecoins, margined futures, and constant-product AMM positions too. Real venues implement the same idea three different ways: a virtual pool that only trades against itself, an oracle-priced pool where liquidity providers are the counterparty (GMX, Jupiter Perps/JLP), or a matching engine — off-chain order book or on-chain auction — that pairs takers with makers (dYdX, Hyperliquid, Drift).

2. Explain it simply

Section titled “2. Explain it simply”

Analogy

Section titled “Analogy”

Imagine two neighbors, Alice and Bob, who want to bet on the price of a house down the street without ever actually buying it. They agree to “pretend” Alice bought the house from Bob at today’s price, and every night they settle the day’s price change in cash — whoever’s side lost money pays the other, and the pretend price resets to whatever the house is really worth that evening. As long as they keep making these nightly payments, the arrangement can run forever, with neither of them ever touching the house. A crypto perpetual future works the same way, except the nightly cash settlement is called a funding payment, the neighbors are called longs and shorts, and the “house price” is the exchange’s own index feed.

ELI5

Section titled “ELI5”

A perpetual is a bet on an asset’s price that never expires. Instead of trading the asset itself, you trade a contract, and once a day (or more often) whichever side is currently “winning” on paper pays a small ongoing fee to the other side. That payment is called funding, and it is sized so betting through the contract stays about as costly as just holding the real asset, which keeps the contract’s price close to the real one. You put down a deposit called margin to cover possible losses, and if losses eat through it, the exchange automatically closes your position — a liquidation.

Step-by-step walkthrough

Section titled “Step-by-step walkthrough”

Scenario: Alice opens a 10x long perpetual on BTC with $1,000 of margin when the index price is $100,000; the exchange charges a funding payment of (mark − index) per contract per day, paid once daily at midnight.

  1. Position opened (state: margin $1,000, notional $10,000, i.e. 0.1 BTC-equivalent, mark = index = $100,000). Alice deposits $1,000 and opens a long position worth $10,000 of BTC exposure. No BTC changes hands; the position is purely a claim on price differences.
  2. Price rises before funding (state: index $103,000, mark $103,050). Alice’s unrealized profit is $10,000 × (103,000/100,000 − 1) = $300, so her margin (mark-to-market) is now $1,300. The perp’s mark price is trading $50 above the index because demand for longs has pushed it up slightly.
  3. Funding is paid (state: midnight, mark $103,050, index $103,000). Longs pay shorts (mark − index) × contracts = $50 × 0.1 = $5, deducted from Alice’s margin, now $1,295. This small tax on longs nudges some of them to close, pushing mark back toward index.
  4. Price falls sharply (state: index drops to $87,000, an 13.6% decline). Alice’s position loses roughly $10,000 × (87,000/100,000 − 1) = −$1,300, wiping out her $1,295 margin. The exchange liquidates her position at or before her margin hits zero, closing it out and, if the loss overshoots what her margin covered, drawing the shortfall from the exchange’s insurance fund.

Common misconceptions

Section titled “Common misconceptions”
  • Myth: A perpetual is redeemable for the underlying asset. Reality: “You can’t redeem the perp for bitcoin or anything else” — it is a cash-settled bet kept in line with spot only by the funding payment (White, Paradigm, 2021-03).
  • Myth: Funding is a perp-specific invention. Reality: “Everything Is A Perp” shows the same logic prices crypto-collateralized stablecoins (a 0-perp) and margined futures like dYdX’s (a 1-perp) too (Robinson, Clark & Leone, Paradigm, 2024-03).
  • Myth: All perp exchanges price trades the same way. Reality: GMX and Jupiter Perps quote straight from an oracle against a pool with no order book, while dYdX and Hyperliquid run matching engines and Drift runs both plus a backstop AMM.
  • Myth: Liquidation happens exactly when margin hits zero. Reality: Exchanges liquidate earlier, at a maintenance threshold, so the insurance fund — not a counterparty — absorbs any overshoot (White, Paradigm, 2021-03).

If you only remember one thing

Section titled “If you only remember one thing”

A perpetual future has no expiry and no delivery — it stays priced like its underlier purely because whoever is on the “too expensive” side of the trade keeps paying the other side until that stops being true.

3. How it works

Section titled “3. How it works”

Mark, index, and the funding payment

Section titled “Mark, index, and the funding payment”

Every perp has a mark price — its own traded price on the venue, used to compute PnL and trigger liquidations — and an index price, usually a volume-weighted feed from external spot exchanges. In the simplified model White uses to explain FTX-style perps, the periodic funding fee paid by longs to shorts is

Funding=(Mark−Index)×contracts\text{Funding} = (\text{Mark} - \text{Index}) \times \text{contracts}

per funding period; when mark trades below index, the sign flips and shorts pay longs. This makes the perp a feedback control system: the more the mark price drifts above the index, the more longs pay, which should make going long less attractive and pull the mark back down — though White is careful to note this is a “marginal” incentive, not a guarantee, since funding’s effect on price depends on unpredictable trader behavior. Real venues typically implement funding as a rate rather than a fixed dollar figure — Drift, for example, derives its funding basis from “the last funding rate (normalized by its oracle TWAP, annualized, and adjusted by the protocol’s funding-rate offset), then decayed linearly to zero over the course of one funding period” (Drift docs, 2026-08) — but the underlying idea, taxing the side that pushed mark away from index, is the same.

A perpetual is the limit of an expiring future

Section titled “A perpetual is the limit of an expiring future”

White’s “perp as one-day future” argument makes the funding constant concrete. Picture a long perp holder who, each hour, sells 1/24th of her position and uses that hour’s funding payment plus the proceeds to buy 1/24th of a real, physically-settled unit of the underlying. After enough hours her time-weighted exposure converges to that of a futures contract expiring in exactly one day — because a 100%-per-day funding coefficient is precisely the rate that makes 1/24th of the gap close every hour. More generally, a funding coefficient λ\lambda (the fraction of the mark–index gap paid per unit time) makes a perpetual price identically, ignoring interest-rate term structure, to an expiring future with maturity

T=1λT = \frac{1}{\lambda}

Halving the funding coefficient doubles the effective expiry. This is the sense in which “perpetual future” is not a metaphor: it is literally the T→T\to continuous limit of an infinite chain of expiring futures, each rolled into the next by the funding payment instead of by a trader manually re-entering a new contract.

Everything is a perp: margined futures as a 1-perp

Section titled “Everything is a perp: margined futures as a 1-perp”

“Everything Is A Perp” generalizes the funding idea to a power perpetual with power pp: an instrument whose funding is Mark−Indexp\text{Mark} - \text{Index}^p, backed by a vault with collateral ratio

Collateral ratio=(collateral qty)×(collateral price)(perp qty)×(index price)p\text{Collateral ratio} = \frac{(\text{collateral qty})\times(\text{collateral price})}{(\text{perp qty})\times(\text{index price})^{p}}

Setting p=0p=0 with ETH collateral and a USD numeraire reproduces an over-collateralized crypto stablecoin (funding =Mark−1= \text{Mark}-1, pulling the stablecoin toward 1).Setting1). Setting p=1$ with USD collateral reproduces a fully-collateralized synthetic long ETH position, and — critically — the paper shows that “overcollateralized perpetuals backed by more than 100% collateral can be converted into undercollateralized perp futures like those traded on dYdX” simply by repeatedly selling the minted token back into more collateral, the same mechanism a lending-market looping strategy uses. This is why the authors call ordinary margined perpetual futures 1-perps: mechanically, a maximally-leveraged, undercollateralized power-perp vault with p=1p=1. The p=2p=2 and other non-linear cases (“power perpetuals” proper) are covered in /derivatives/power-perps-and-everlasting-options/.

Three venue architectures

Section titled “Three venue architectures”
  1. Oracle-priced pool (trader-vs-LP). Jupiter Perps has no order book at all: “traders borrow assets from the Jupiter Liquidity Pool (JLP) to open leveraged positions,” positions are “priced using onchain oracles, which means trades execute at the displayed price without orderbook slippage,” and a price-impact fee substitutes for slippage to protect the pool (Jupiter docs, 2026-08). GMX uses the same shape — GM/GLV pools are the counterparty to every trade, quoting “the oracle index price rather than relying on an order book or external market makers” (GMX docs, 2026-08).
  2. Order book with off-chain/keeper matching. dYdX (originally an Ethereum StarkEx layer-2, now its own Cosmos application chain) and Hyperliquid run a conventional central limit order book, matching resting maker orders against incoming takers; Hyperliquid’s HyperCore “includes fully onchain perpetual futures and spot order books” with “one-block finality” and throughput of “200k orders / second” (Hyperliquid docs, 2026-06).
  3. Hybrid auction + book + AMM. Drift layers three liquidity sources for every taker order: a JIT (just-in-time) auction, where the requested price walks linearly from the taker’s best price toward their limit over roughly 5–10 slots (about 2–4 seconds at Drift’s 400 ms baseline slot time) while market makers race to fill it early at a better price; the resting decentralized order book (DLOB); and Drift’s own AMM as backstop liquidity, which competes on price at each level rather than filling only as a last resort (Drift docs, 2026-08).

The earliest design in this space, the virtual AMM (vAMM) used by protocols like early Drift v1 and Perpetual Protocol, simulates a constant-product curve with no real reserves purely to derive a mark price — it takes no counterparty risk itself, but needs an external vault to fund PnL and liquidations, which is why most vAMM-based venues have since migrated to oracle-pool or order-book designs.

4. Worked numeric example

Section titled “4. Worked numeric example”

Continuing Alice’s position from §2: 10x long BTC perp, $1,000 margin, $10,000 notional (0.1 BTC-equivalent), opened at index = mark = $100,000, funding paid once daily as (mark − index) per contract.

Day 1 (mark drifts above index). Index closes at $103,000; the perp’s own mark, pulled up slightly by long demand, closes at $103,050. Funding owed by longs: (103,050−103,000)×0.1=$5(103{,}050-103{,}000)\times 0.1 = \text{\textdollar}5. Alice’s margin: starting $1,000, plus unrealized PnL of $10,000 × (103,000−100,000)/100,000 = $300, minus \text{\textdollar}5 funding = \text{\textdollar}1,295.

Day 2 (price reverses hard). Index falls to $87,000 (a 13.6% single-day move — extreme but not unheard of in crypto). Alice’s PnL versus her $100,000 entry: $10,000 × (87,000−100,000)/100,000 = −$1,300. Her $1,295 of margin cannot absorb this; if the exchange’s maintenance-margin threshold is, say, 1% of notional ($100), the liquidation engine should have closed her out well before the loss reached $1,295 — but assume a fast market gapped through the trigger. The shortfall of $1,300 − $1,295 = $5 is covered by the venue’s insurance fund rather than by Bob (her counterparty), exactly the design point White makes: “the system is safe to the extent that liquidations happen quickly enough and the insurance fund is sufficiently capitalized.”

What Alice paid for 10x leverage. Over two days she paid $5 in funding and lost her full $1,000 principal plus the insurance fund absorbed $5 more — a reminder that at 10x leverage, a 13% adverse move is enough to erase the position entirely, funding costs are a small fraction of total risk, and the exchange’s own solvency (not Alice’s counterparty) is what stands behind any gap-through-liquidation loss.

5. Where it’s used

Section titled “5. Where it’s used”

Ethereum

Section titled “Ethereum”
  • dYdX — pioneered the order-book perpetual on an Ethereum StarkEx layer-2 (dYdX v3) before migrating fully off Ethereum to its own Cosmos SDK chain for v4; still the reference design for a fully order-book-matched on-chain perp exchange. docs.dydx.xyz
  • GMX — decentralized spot and perpetual exchange on Arbitrum, Avalanche, and MegaETH; up to 100x leverage, no order book, oracle-priced trades against GM/GLV pools, with liquidity providers earning “63% of the fees generated from trading, liquidations, borrowing fees, and swaps” (GMX docs, 2026-08). gmx-docs.io
  • Hyperliquid — not an Ethereum protocol but included here as the benchmark CLOB venue that both Ethereum and Solana perp DEXs are measured against: its own L1 (HyperBFT consensus) running a fully on-chain order book (HyperCore) plus an EVM environment (HyperEVM) for permissionless building on top of that liquidity (Hyperliquid docs, 2026-06).

Solana

Section titled “Solana”
  • Drift Protocol (rebranded “Velocity Protocol” in its own docs as of the 2026-08 fetch) — perpetuals-only order book combining JIT Dutch auctions, a decentralized limit order book, and an AMM backstop, with fills allocated in price order across whichever liquidity source is best at each level. docs.drift.trade
  • Jupiter Perps / JLP — trader-vs-LP oracle-priced perpetuals with up to 250x leverage on SOL, ETH and wBTC; JLP holders “act as the counterparty to all trades,” earning 75% of trading, swap, and minting/burning fees while bearing trader PnL as pool risk (Jupiter docs, 2026-08). docs.jup.ag

6. Risks, attacks, and incidents

Section titled “6. Risks, attacks, and incidents”
  • Mango Markets, 11 October 2022 — over $110 million. Avraham Eisenberg used roughly $10 million to rapidly buy MNGO on three exchanges feeding Mango’s oracle, jumping its price “over 13-fold during a 30-minute span,” then used that inflated collateral value to borrow out “over $110 million in digital assets” (CFTC v. Eisenberg press release, 2023-01). He later returned roughly $67 million while keeping about $47 million; both the CFTC and SEC charged market manipulation, though a 2025 federal ruling found his conduct did not meet the specific fraud statute charged — an unresolved “gaming the rules vs. manipulation” tension regulators are still litigating.
  • Hyperliquid JELLY incident, 26 March 2025 — up to $12 million. An attacker opened a roughly $4.1 million short on the thinly-traded JELLYJELLY token, then pumped its price over 400% across exchanges to force their own position underwater; Hyperliquid’s liquidity pool (HLP) inherited the short when the order book couldn’t absorb the liquidation, briefly showing up to $12 million of unrealized loss against HLP’s ~$230 million balance. Validators delisted JELLY within about two minutes, settling at the attacker’s $0.0095 entry price rather than the ~$0.50 market price; the Hyper Foundation reimbursed most affected users (Halborn, 2025-03, secondary).
  • GMX V1, 9 July 2025 — $42 million. A reentrancy bug in the executeDecreaseOrder function let an attacker’s contract regain control mid-transaction, updating the list of short positions without updating the global average short price used to calculate GLP’s value, artificially inflating perceived pool value and letting the attacker redeem GLP for far more than it was owed. The attacker returned the funds in exchange for a negotiated $5 million bounty (Halborn / QuillAudits, 2025-07, secondary).
  • Drift Protocol, 1 April 2026 — approximately $285 million. Attackers spent months building social trust before exploiting Solana’s durable-nonce feature to get Security Council multisig members to pre-sign dormant transactions that, once triggered, transferred admin control; the attackers then whitelisted a worthless self-created token as collateral and used it to withdraw real assets. TRM Labs attributed the attack with “medium-high confidence” to North Korea-linked actors previously tied to the 2024 Radiant Capital hack (TRM Labs, 2026-04, secondary). Unlike the other incidents here, this was a governance/access-control failure rather than an oracle or funding-rate exploit — a reminder that a perp exchange’s admin-key and multisig security matters as much as its pricing mechanism.
  • Structural risks common to the design. Cascading liquidations in a fast market can outrun an insurance fund, as in Alice’s worked example; thinly-traded assets are especially vulnerable to funding-rate and oracle manipulation (Mango, JELLY); oracle-pool venues (GMX, JLP) concentrate trader-vs-LP risk in a single pool that must be large enough to cover worst-case drawdowns; and order-book venues depend on makers actually showing up during volatile moves, which is exactly when JELLY-style liquidity gaps appear.

7. Open problems

Section titled “7. Open problems”
  • What counts as manipulation versus legitimate trading? The Mango case shows current law and protocol design have not settled this: Eisenberg argued (and a judge partly agreed) that using a protocol exactly as coded is not fraud, even when the economic effect is identical to theft.
  • Decentralization versus safety trade-offs during a crisis. Hyperliquid’s two-minute validator-coordinated delisting after the JELLY incident protected users but relied on a level of centralized, discretionary intervention that critics argue undercuts the “fully on-chain, no counterparty” pitch of perp DEXs (OneKey, 2025-03, secondary).
  • Regulatory clarity for DeFi perps in the US. Paradigm’s 2025-05 comment letter to the CFTC argues that “DeFi perps remain effectively off-limits in the U.S., trapped in regulatory gray zones,” and asks for a dedicated advisory committee; as of this writing no such framework exists, and perps — which Paradigm says made up “93% of all crypto derivatives volume” by 2025 — trade in a legal gray zone for US-based users (Grieve & Slaughter, Paradigm, 2025-05).
  • Concentration risk in pooled-liquidity designs. As oracle-pool venues like GMX and JLP grow, a growing share of trader PnL is absorbed by a single shared vault; how much drawdown such a pool can safely underwrite, and how it should be capitalized against tail scenarios, remains an active risk-engineering question without an industry-standard answer.

8. Ethereum vs Solana

Section titled “8. Ethereum vs Solana”
AspectEthereumSolana
Dominant venue architectureOrder book (dYdX, historically on an Ethereum L2) and oracle-priced pool (GMX)Hybrid JIT-auction + order book + AMM (Drift); pure oracle-priced pool (Jupiter Perps/JLP)
Counterparty to trader PnLGM/GLV liquidity providers (GMX) or the opposing order-book trader (dYdX)JLP pool (Jupiter Perps) or DLOB/JIT makers plus AMM backstop (Drift)
Settlement cadence~12 s blocks; keeper- and oracle-driven execution400 ms slots (trending toward 200 ms); JIT auctions measured in slots, not seconds
Benchmark for bothHyperliquid’s own L1 (HyperBFT, 200k orders/sec, one-block finality, as of 2026-06) sets the throughput and UX bar neither Ethereum nor Solana perp DEXs yet match on-chainSame
Highest-profile incidentsMango Markets (Solana, Oct 2022) and GMX (Arbitrum, Jul 2025) show oracle and reentrancy risk cut across chainsDrift’s Apr 2026 governance exploit shows admin-key risk is independent of the trading mechanism

Both ecosystems converge on the same three architectures — vAMM, oracle-pool, and order book — because the underlying funding-payment mathematics is identical regardless of chain; what differs is which architecture each chain’s block time and fee market make viable. Solana’s sub-second slots make a per-order Dutch auction (Drift’s JIT) cheap enough to run on every single trade, something that would be prohibitively slow and costly on Ethereum’s ~12-second blocks, which is one reason Ethereum-native venues lean more heavily on oracle-pool designs or move order-book matching off-chain entirely (dYdX v4, Hyperliquid).

9. Reference doc

Section titled “9. Reference doc”

The reference

Section titled “The reference”

The Cartoon Guide to Perps — Dave White (Paradigm), 8 March 2021. paradigm.xyz/writing/the-cartoon-guide-to-perps

Summary of the reference

Section titled “Summary of the reference”

White opens by confessing that despite perps trading “tens or sometimes hundreds of billions of dollars per day,” his own understanding of the mechanism (“funding rate,” right?) felt thin, so the post builds up four independent, mathematically equivalent mental models. Perp anatomy first defines mark price (the perp’s own traded price, used for PnL and liquidations), index price (the underlier’s external price), and the funding fee paid between longs and shorts whenever mark and index diverge, plus margin and liquidation. Perp as thermostat frames funding as a feedback-control system nudging mark back toward index, while flagging the honest caveat that markets are hard to predict and the feedback is only “marginal.” Perp as PNL loan walks through Alice and Bob informally agreeing to settle a pretend bitcoin trade in cash with interest on unrealized profit, re-deriving the funding formula from first principles. Perp as loan swap re-derives the same formula by having Alice and Bob swap a dollar loan for a bitcoin loan and net the interest. Perp as one-day future is the most technical section: it shows that continuously rolling a fraction of a perp position into the physical underlying, funded by that period’s funding payment, converges to the same time-weighted exposure as a future expiring in exactly one day when funding is 100%-per-day — generalizing to an equivalent expiry of 1/λ1/\lambda for any funding coefficient λ\lambda. A closing section on margin and liquidations works a numeric example of a 100x position wiped out by a 1–2% price move and introduces the insurance fund as the backstop when a liquidation doesn’t happen in time.

Key quotes

Section titled “Key quotes”

“For the cost of $1, a 100x bitcoin perp… behaves like $100 of bitcoin… especially considering you can’t redeem the perp for bitcoin or anything else.” (Introduction)

“When the index price and the mark price diverge, the system transfers a funding fee between those who are long the perp… and those who are short the perp.” (Perp anatomy)

“Anon’s time-weighted exposure to bitcoin through the perp during this process ends up being the same as it would be if she were holding a futures contract expiring in exactly 24 hours.” (Perp as one-day future)

“Generally speaking, the system is safe to the extent that liquidations happen quickly enough and the insurance fund is sufficiently capitalized.” (Margin and liquidations)

How to read the original

Section titled “How to read the original”

Background needed: comfort with long/short terminology and the idea of margin trading; no options or stochastic-calculus background is required for the first four sections. Skip the deep mechanics of the “one-day future” section on a first pass if the geometric intuition (“she keeps transmuting her position into real bitcoin”) is enough — the formal proof is unpacked further in the companion paper covered below. The hardest paragraph is the “one-day future” walkthrough itself: the trick is realizing that the same funding payment being received simultaneously (a) compensates the discount from selling at a stale mark price and (b) exactly finances the marginal bitcoin purchase, so the two effects are not double-counted — reread the specific dollar amounts ($10k profit, $400/hour) slowly rather than trying to generalize the argument abstractly on a first pass.

What changed since

Section titled “What changed since”
  • Paradigm’s own follow-up, “Everything Is A Perp” (2024-03), generalizes this framework, showing margined futures are a “1-perp,” stablecoins a “0-perp,” and constant-product AMM positions a “0.5-perp” — see the second reference below.
  • Perp exchange architecture has diversified well beyond the single-venue, single-mark-price model White assumed: oracle-priced pools (GMX, Jupiter Perps) removed the order book entirely, and hybrid designs (Drift) run three liquidity sources simultaneously (2022–2026).
  • The regulatory question White doesn’t touch has become the central open issue for US-based perp trading, per Paradigm’s 2025-05 CFTC comment letter (see §7).

The reference

Section titled “The reference”

Everything Is A Perp — Dan Robinson, Joe Clark, Andrew Leone (Paradigm), 6 March 2024. paradigm.xyz/writing/everything-is-a-perp

Summary of the reference

Section titled “Summary of the reference”

The post makes three claims that unify seemingly unrelated DeFi primitives under one funding-payment framework. It first defines a perpetual generically as a contract that “tracks and gives exposure to an index, and exchanges regular payments that are larger the further the traded price (mark) is away from the target price (index),” and a power perpetual as a perpetual on indexp^p. Claim 1 shows a crypto-collateralized stablecoin is a p=0p=0 power perp: since price0=1\text{price}^0=1, funding is simply Mark−1\text{Mark}-1, which is exactly the incentive structure that pulls a stablecoin toward its dollar peg. Claim 2 shows margined futures (like dYdX’s) are p=1p=1 power perps: an over-collateralized synthetic asset can be turned into an under-collateralized, leveraged perp future by repeatedly selling the minted token back in as more collateral, a mechanically explicit version of a looping strategy. Claim 3 argues constant-product AMMs (Uniswap) are close to a p=0.5p=0.5 power perp, because a full-range LP position’s value scales with price\sqrt{\text{price}}; a portfolio short the 0.5-perp against the LP position isolates the pool’s trading fees from its price risk, implying that in equilibrium, expected Uniswap fee income should equal σ2/8\sigma^2/8 for a zero-rate approximation, where σ2\sigma^2 is the variance of the pool’s relative price returns. The paper closes by noting the natural extension to squared and higher-power perpetuals (Squeeth) and to Taylor-series portfolios of whole-number-power perps as a way to approximate arbitrary payoff functions.

Key quotes

Section titled “Key quotes”

“A perpetual is a contract that tracks and gives exposure to an index, and exchanges regular payments that are larger the further the traded price (mark) is away from the target price (index).” (Definition)

“This means that overcollateralized perpetuals backed by more than 100% collateral can be converted into undercollateralized perp futures like those traded on dYdX.” (Claim 2)

“This gives us a nice result that the equilibrium Uniswap fee should be the funding rate for a 0.5 perpetual.” (Claim 3)

How to read the original

Section titled “How to read the original”

Background needed: the Cartoon Guide’s funding-fee model above, plus a basic sense of collateralized-debt-position mechanics (deposit collateral, mint a token). Skip the Taylor-series “what’s missing” section on a first pass — it is a forward-looking research direction, not part of the core argument. The hardest part is Claim 3’s replication argument: it helps to separately track two portfolios — “long the LP position” and “short the 0.5-perp” — and notice that the only difference in their payoffs is that one collects trading fees and the other pays perpetual funding, which is what pins the two cash flows to be equal in a no-arbitrage equilibrium.

What changed since

Section titled “What changed since”
  • The paper’s power-perp framing builds directly on “Power Perpetuals” (2021-08) and “Everlasting Options” (2021-05); see /derivatives/power-perps-and-everlasting-options/ for the full derivation of non-linear power perps.
  • Since 2024, oracle-pool perp venues (GMX, Jupiter Perps/JLP) have made the “LP-as-counterparty” relationship this paper formalizes into the dominant retail perp design on both Ethereum L2s and Solana (§5).

Secondary references

Section titled “Secondary references”
  1. Jupiter Perps & JLP documentation — read if you want the concrete mechanics of a live oracle-pool perp venue matching this paper’s Claim-2/Claim-3 style of reasoning.
  2. Drift Protocol JIT auction docs — read if you want to see an on-chain Dutch-auction alternative to a static funding-payment order book.
  3. Paradigm, “DeFi Perps Deserve a Path Forward in the US” (2025-05) — read if you want the regulatory angle these two mechanism papers don’t cover.

10. Sources

Section titled “10. Sources”