Trade Options on Hyperliquid
Source: http://x.com/i/article/2050552332560023552
TLDR
- HIP-4 price markets (BTC above X by Y) can act as dated options but with their payoff structure differs materially from traditional options.
- The greeks (delta, gamma, vega, theta)
don't port over well to individual HIP-4 options.
- Chaining together a lot of binary options, however, can approximate the same exposure as traditional options. The greeks also converge back to normal behavior.
Hyperliquid launched HIP-4 outcome markets on Saturday with "BTC above 78213 on May 3 at 2:00 AM?" as their first experimental market.
For now, Hyperliquid seems to be only launching 15m and 1d BTC markets according to their docs and the live nav-tabs on the frontend.
Individually, these outcome markets can simulate options-like behavior but don't map options microstructure 1:1. However, chaining a bunch of these individual binary markets together can provide traditional options exposure quite well.
This becomes quite interesting as HIP-4 expands out from 15m and 1d markets towards longer-dated markets. Being able to trade options along with spot and perps markets can enable strategies that did not have a home onchain until now.
Below, we will discuss why individual HIP-4 markets don't have the same payoff structure and the same greeks. Then, we will discuss how to construct vanilla options-like payoff and greeks through an options strip. Lastly, we will discuss why portfolio margin will be crucial for HIP-4 in order to enable strategies that work across spot, perps, and outcome markets.
There's a few math formulas sprinkled throughout for a deeper understanding but feel free to skip over them since they are not crucial. Every subheading is broken up into Math and Intuition.
Note: Please see full disclaimer attached in the first reply to this article.
Binary HIP-4 Options ≠ Vanilla Options
For the purposes of an example, let's consider HYPE is trading at $40 for the rest of this article. Now consider the "HYPE > $45 by June 30" market:
- YES pays $1 if HYPE > $45, $0 otherwise.
- NO pays $1 if HYPE < $45, $0 otherwise
- Let's say the market trades at $0.30. This is the market's risk-neutral probability that HYPE settles above $45.
- The terminal payoff is a Bernoulli random variable: $1 with probability p, $0 with probability 1-p. The long posts p per contract, the short posts 1-p, and total escrow per contract is $1.
The most important caveat however is that the terminal payoff is a step function, not a ramp.
If HYPE finishes at $45.01 the market pays $1. If the HYPE finishes at $100, the market still only pays $1.
A vanilla $45 call keeps paying more as HYPE rallies. A single binary caps payout at $1 regardless of how far the rally extends.
The core attractor of a call option is right-tail convexity.
You pay a small premium to participate in theoretically infinite upside with a capped downside. Let's say you buy YES >$50 at $0.25 your maximum return is $0.75/$0.25 = 3x. The same vanilla call can ride a 10x+ if the underlying keeps increasing.
Options Greeks Don't Port Over to HIP-4
We can roughly think of a vanilla option's value as:
dV≈Δ dS+12Γ (dS)2+Θ dt+ν dσdV \approx \Delta\,dS + \tfrac{1}{2}\Gamma\,(dS)^2 + \Theta\,dt + \nu\,d\sigma
As a brief refresher:
- V → option price
- S → underlying's price
- Δ → the option's sensitivity to the underlying. If Δ = 0.1, the option price increases by +0.1 for every $1 move in the underlying
- Γ → the option's sensitivity to the delta (i.e. Γ = dΔ}/dS}) . Γ is how options gain convex upside. If the delta changes fast, Gamma rips making the option more expensive
- Θ → the option's sensitivity to time. For a vanilla long call, Θ < 0 and the position bleeds premium each day toward expiry.
- vega → the option's sensitivity to implied volatility. If vega = 0.10, a 1 percentage-point rise raises the option price by $0.10.
Other helpful terms:
- K → strike price
- d₁ → CDF of d₁ = vanilla option's delta
- d₂ → CDF of d₂ = risk-neutral probability of finishing ITM
Δ
For a vanilla call,
Δ=N(d1) \Delta = N(d_1)
where N(d₁) is the CDF for d₁
Delta moves monotonically from 0 to 1 like a sigmoid function:
- Deep OTM it barely moves (Δ → 0)
- Deep ITM it tracks spot nearly dollar for dollar (Δ → 1)
With a binary, delta is a hump.
It is low when the event is obviously unlikely, high near the strike, and low again when the event is almost certain.
The Math: Consider a cash-or-nothing call under Black-Scholes,
p(S)=e−rτN(d2)p(S) = e^{-r\tau}N(d_2)
Differentiating with respect to spot:
Δbinary=∂p∂S=e−rτφ(d2)Sστ\Delta_{\text{binary}} = \frac{\partial p}{\partial S} = \frac{e^{-r\tau}\varphi(d_2)}{S\sigma\sqrt{\tau}}
since:
∂∂SN(d2(S))=φ(d2)⋅∂d2∂S=φ(d2)⋅1Sστ\frac{\partial}{\partial S}N(d_2(S)) = \varphi(d_2)\cdot\frac{\partial d_2}{\partial S} = \varphi(d_2)\cdot\frac{1}{S\sigma\sqrt{\tau}}
where φ(d₂) is the pdf of d₂
Or simply:
Δbinary=e−rτφ(d2)Sστ\boxed{ \Delta_{\text{binary}} = \frac{e^{-r\tau}\varphi(d_2)}{S\sigma\sqrt{\tau}} }
As such, binary delta is largest near the strike and decays on both sides.
Intuition:
- Deep OTM: HYPE is far below K, so a small move barely changes the probability of finishing above K. Delta is near zero.
- Near the strike:
The event is close to a coin flip, so every dollar of spot movement materially shifts the probability. Delta is at its maximum.
- Deep ITM: The event is already likely and probability is pinned near 1. Further spot moves barely change the binary's price. Delta decays back toward zero.
For instance, if HYPE is trading at $35 with strike at $45, a $1 move to $36 barely changes the odds. If HYPE is at $44.50, the same $1 move can flip the market's view materially. And if HYPE is at $60, that $1 move barely matters because the event is already priced in.
Γ
Gamma is super powerful in vanilla options since it grants convexity to an option.
Γcall=φ(d1)Sστ\Gamma_{call} = \frac{\varphi(d_1)}{S\sigma\sqrt{\tau}}
In vanilla options, gamma is always positive.
Big moves help disproportionately and small moves only cost theta.
In binary options:
Γbinary=∂Δ∂S=−e−rτφ(d2) d1S2σ2τ\boxed{ \Gamma_{\text{binary}} = \frac{\partial\Delta}{\partial S} = -\frac{e^{-r\tau}\varphi(d_2)\,d_1}{S^2\sigma^2\tau} }
φ′(d₂) = −d₂·φ(d₂), and d₂ + σ√τ = d₁
In binary options, gamma can work against you since it can flip negative.
The sign depends on d₁ and flips at/slightly before the strike.
When Gamma flips negative the position essentially becomes short convexity.
Intuition:
- Below the strike, d₁ < 0 and gamma is positive --> long convexity.
- Above the strike, d₁ > 0 and gamma is negative --> short convexity.
- The crossover happens at d₁ = 0, i.e., S = K·e^(−(r+σ²/2)τ) → slightly below the strike for short tenors and small rates
ν
For a vanilla call, vega is positive:
νcall=φ(d1)ST\nu_{\text{call}} = \varphi(d_1)S\sqrt{T}
Higher vol helps a vanilla call because it increases the chance of larger right-tail payoffs. The payoff is uncapped on the upside, so more dispersion is valuable.
For a binary:
νbinary=−e−rTφ(d2)d1σ\boxed{ \nu_{\text{binary}} = - \frac{e^{-rT}\varphi(d_2)d_1} {\sigma} }
The sign again depends on d₁.
Intuition:
- If the binary is OTM, more vol helps because it increases the chance of crossing the strike.
- If the binary is ITM, more vol can hurt because it increases the chance of falling back below the strike.
Θ
For a vanilla call, theta is always negative:
Θcall=−φ(d1) S σ2T−rKe−rTN(d2)\Theta_{\text{call}} = -\frac{\varphi(d_1)\,S\,\sigma}{2\sqrt{T}} - rKe^{-rT}N(d_2)
Time decay erodes the extrinsic value of the option. The longer you hold a vanilla long call, the more premium you bleed.
Intuition:
- For a binary, theta can sign-flip around the strike:
- OTM: Theta is negative for the long. As time passes, the probability of crossing the strike tends toward 0, and the binary's price falls toward $0.
- ITM: Theta is positive for the long. As time passes, the probability tends toward 1, and the binary's price rises toward $1.
This is best understood with the following example:
- Suppose HYPE is at $55, the binary is "HYPE > $45", and the YES price is $0.85 with one week left.
- If one day passes and HYPE is still $55, there is less time for HYPE to fall back below $45.
- The event is now more locked in, so the YES price might move from $0.85 to $0.90.
- That price increase came from time passing, not spot moving. That is positive theta for the long binary.
The reverse happens OTM. If HYPE is at $35 and the binary is "HYPE > $45", one day passing gives HYPE less time to climb above the strike. The YES price falls, so theta is negative.
This is impossible for a vanilla long call. The binary's capped payoff means time can work for or against you depending on which side of the strike you sit on.
Rebuilding Vanilla Options using Binary Options
Ignoring premium for a second, a vanilla call pays:
(ST−K)+(S_T-K)^+
That means:
Pay me one dollar for every dollar the underlying finishes above the strike.
We can rewrite that as:
(ST−K)+=∫K∞1ST>k dk(S_T-K)^+ = \int_K^\infty \mathbb{1}_{S_T>k}\,dk
Discretizing, we can write this as:
(ST−K)+≈∑i1ST>Ki(S_T-K)^+ \approx \sum_i \mathbb{1}_{S_T>K_i}
where Kᵢ is the i-th strike in the ladder.
However, practically speaking, options will not exist for every tick (e.g., $45.01, $45.02, ...), they will be spaced out (e.g., $45.0, $45.5, $46.0, ...)
Let ΔK be the distance between strikes, so so Kᵢ = K + i·ΔK. The discrete approximation becomes:
(ST−K)+≈ΔK∑i1ST>K+iΔK(S_T-K)^+ \approx \Delta K \sum_i \mathbb{1}_{S_T>K+i\Delta K}
where:
- ΔK is the distance between strikes.
- i is the strike index.
- K+iΔK is the i-th strike in the ladder.
The strip matches the vanilla up to the cap ($65) and above the cap, the strip clips at $20 while the vanilla keeps growing.
The tighter the strips, the closer it mimics a vanilla capped call:
Thus, HIP-4 option strips only allow for capped calls. We can formally write the payoff of a $45→$65 strip as:
∫45651ST>k dk=(ST−45)+−(ST−65)+\int_{45}^{65}\mathbb{1}_{S_T>k}\,dk = (S_T-45)^+ - (S_T-65)^+
In other words, HIP-4 $45 to 65 call, is equivalent to a 45-65 capped call/call spread where we are long the $45 call and short the $65 call. Below the cap the strips behaves like a vanilla long call and above the cap, it stops participating entirely – but it tracks the payoff quite closely and the greeks also port over well as we will see below:
Pricing a HIP-4 Option Strip
Looking at vanilla options:
C(K,T)=∫K∞e−rTQ(ST>k) dkC(K,T) = \int_K^\infty e^{-rT}\mathbb{Q}(S_T>k)\,dk
For all-or-nothing options, the no-arb price at strike k is:
p(k)=e−rTQ(ST>k)p(k)=e^{-rT}\mathbb{Q}(S_T>k)
where ℚ is the risk-neutral measure: the pricing probability distribution after risk premia and discounting.
Thus, price of vanilla option:
C(K,T)=∫K∞p(k) dkC(K,T) = \int_K^\infty p(k)\,dk
Extrapolating to a HIP-4 strip:
Cstrip≈ΔK∑ip(K+iΔK)\boxed{ C_{\text{strip}} \approx \Delta K\sum_i p(K+i\Delta K) }
For the finite $45→$65 strip, we would use ∫₄₅⁶⁵ p(k) dk or ΔK · Σ_{Kᵢ<65} p(Kᵢ)
Intuition for pricing a HIP-4 strip:
- Each outcome market gives one probability price.
- A vanilla call is the sum of those probability prices across strikes.
- No implied-vol inversion is needed if the probability ladder already exists.
Rebuilding the Greeks using Strips
As a reminder, a single binary has anomalies in the Greeks:
- Delta is humped.
- Gamma flips sign.
- Vega can flip sign
However, the greeks converge back to vanilla options when we "strip" together a bunch of binaries:
Δ
A strip is a sum of underlying binaries. As such, the Greeks add linearly as well.
Δstrip(S)=ΔK∑i∂p(Ki;S)∂S →ΔK→0 ∫KKmax∂p(k;S)∂S dk\Delta_{\text{strip}}(S) = \Delta K \sum_i \frac{\partial p(K_i; S)}{\partial S} \;\xrightarrow{\Delta K \to 0}\; \int_K^{K_{\max}} \frac{\partial p(k; S)}{\partial S}\,dk
⟹ ∂∂S∫KKmaxp(k;S) dk=∂∂S[C(K,S)−C(Kmax,S)]=N(d1(K,S))−N(d1(Kmax,S))\implies \frac{\partial}{\partial S}\int_K^{K_{\max}} p(k; S)\,dk = \frac{\partial}{\partial S}\bigl[ C(K, S) - C(K_{\max}, S) \bigr] = N(d_1(K, S)) - N(d_1(K_{\max}, S))
We know N(d₁) = Δ :
Δstrip(S)=Δcall(K,S)−Δcall(Kmax,S)\boxed{\Delta_{\text{strip}}(S) = \Delta_{\text{call}}(K,S) - \Delta_{\text{call}}(K_{\max}, S)}
Intuition:
- The strip's aggregate delta is the vanilla $45 call's delta minus the vanilla $65 call's delta.
- As K_max → ∞, N(d₁(K_max, S)) → 0, and the strip delta converges to the uncapped vanilla call delta.
Γ
Γstrip(S)=ΔK∑i∂2p(Ki;S)∂S2 →ΔK→0 Γcall(K,S)−Γcall(Kmax,S)\Gamma_{\text{strip}}(S) = \Delta K \sum_i \frac{\partial^2 p(K_i; S)}{\partial S^2} \;\xrightarrow{\Delta K \to 0}\; \Gamma_{\text{call}}(K, S) - \Gamma_{\text{call}}(K_{\max}, S)
Each individual binary's gamma is +ve below its strike and -ve above it.
However, when the strikes are dense the -ve tail of one binary's gamma overlaps with the +ve tail of another and they cancel out.
Intuition:
- Thus, we get +ve gamma near the lower edge of the strip and -ve gamma near the upper cap.
- Again, as K_max → ∞, and ΔK → 0, Γ stays +ve since our window becomes so large that only the lower edge matters.
ν
Vega rebuilds the same way:
νstrip(S)=ΔK∑i∂p(Ki;S)∂σ →ΔK→0 νcall(K,S)−νcall(Kmax,S)\nu_{\text{strip}}(S) = \Delta K \sum_i \frac{\partial p(K_i;S)}{\partial \sigma} \;\xrightarrow{\Delta K \to 0}\; \nu_{\text{call}}(K,S)-\nu_{\text{call}}(K_{\max},S)
For the continuous capped strip:
νstrip(S)=ST[φ(d1(K,S))−φ(d1(Kmax,S))]\boxed{\nu_{\text{strip}}(S)=S\sqrt{T}\left[\varphi(d_1(K,S))-\varphi(d_1(K_{\max},S))\right]}
Intuition:
- If the cap ($65) is far from spot, the second term is tiny because vega is concentrated near ATM and far-OTM strikes carry almost no vega. The strip behaves like a normal long vanilla call.
- If spot moves near or past the cap, the second term grows and partially cancels the first. The strip's net vega shrinks and eventually the strip stops being cleanly "long vol."
Discrete vs Continuous
Vanilla options are continuous terminal payoffs.
If HYPE settles $0.1 higher, a deep ITM call pays $0.1 more.
However, binary strip options are discrete and have stepwise payoffs.
If HYPE settles $0.10 higher but not enough to cross into the next market, the expiry payout is unchanged.
This creates a subtlety:
- Before expiry, the strip's mark-to-market price still moves continuously, because the underlying probabilities move continuously.
- At expiry, the payoff is a staircase, because each leg resolves discretely. Thus, payout is chunky because each binary either pays or does not.
Capital Constraints
Longs pay ΔK · Σᵢ pᵢ, shorts pay ΔK · Σᵢ (1−pᵢ). For OTM strips Σᵢ pᵢ ≪ Σᵢ (1−pᵢ). Thus, shorts have higher capital requirements. In traditional options markets, these are offset by portfolio margin.
However, without portfolio margin, this asymmetry compounds because each binary market is treated as independent and does not net across offsetting positions.
Suppose one shorts two strips at the same time:
- A $45→$65 capped call strip: blows up if HYPE rallies above $65.
- A $20→$35 capped put strip: blows up if HYPE crashes below $20.
HYPE cannot do both at once, yet HIP-4 charges collateral against each tail separately. Without portfolio margin, the capital requirements for short-vol positions and market makers shoots up astronomically. That said, since HIP-4 will mainly have markets against which you cannot portfolio margin (sports, politics, etc), so it will be interesting to see if any markets are portfolio margined if at all.
As such, constructing options using HIP-4 strips is somewhat of a half-baked solution and does not simulate the full power and flexibility of options. If options are to be truly implemented they might benefit from a standalone HIP-upgrade.