How BOSS prices options strategies
How BOSS turns public order books from Deribit, OKX and Bybit into a priced options structure: the data model, how a structure is built, how its cost, payoff, probability of profit and volatility measures are computed, and how the numbers are checked. The Methodology page summarizes the same rules; this document gives the formulas.
Last reviewed:
1. Data pipeline
One ingest process per venue discovers the listed BTC and ETH options (the nearest six expiries, strikes from 0.5 to 1.5 times the index), subscribes to each contract's public ticker over WebSocket and writes every update to Redis with a 30-second expiry, renewed on each tick, then publishes a notification. The web server only reads: it subscribes to those notifications, keeps the quotes in memory and pushes a re-render to the open pages that show the contract, at most twice a second per page.
Each quote carries the venue's own timestamp. A structure is shown as outdated when its newest quote is more than 60 seconds old. Slower series (futures and perpetual books, funding, daily and hourly index closes, delivery prices) are polled on their own schedules.
2. Normalization
Every price is converted to US dollars per one coin of underlying before any computation. Deribit and OKX quote BTC and ETH option premiums in coin; with F the venue's forward for that expiry:
premium_USD = premium_coin × F
which is the conversion the venues use themselves. Bybit quotes in dollars. Implied volatility is stored in percent. Greeks are taken from each venue's model per contract, in dollars (OKX's dollar-denominated set; its coin-denominated one is ignored). Mid price m = (bid + ask) / 2; with one side missing, the venue's mark (Deribit, Bybit) or last trade (OKX) stands in.
3. Building a structure
Strategies are data: each is a list of legs, each leg a right (call, put or underlying), a direction s (+1 bought, −1 sold), a ratio r and a strike offset. Nothing in the code is specific to one strategy.
- The default expiry is the first one at least 7 days away.
- The at-the-money strike
K₀is the listed strike nearest that expiry's forward. - An offset
nbecomesK₀ + n × Δ × w, whereΔis the most common spacing between listed strikes andwthe width chosen on the page (separate widths for a condor's body and wings), then snapped to a listed strike.
4. Execution cost and fees
Each leg's execution price is the ask when bought and the bid when sold: e_i = ask_i if s_i = +1, bid_i if s_i = −1. The estimated taker fee per contract is
f_i = min(rate × F, cap × m_i)
with rate 0.03% on all three venues and cap 12.5% (Deribit) or 7% (OKX, Bybit). The structure's net cost, positive when paid, is
C = Σ s_i r_i e_i + Σ r_i f_i
and its slippage is max(0, Σ s_i r_i e_i − Σ s_i r_i m_i), the cost of crossing the spread. Only the top of each book is used, so these hold for the size quoted at the best bid and ask.
5. Payoff, breakevens and extremes
At expiry each leg is worth its intrinsic value v_i(S): max(S − K_i, 0) for a call, max(K_i − S, 0) for a put, S for the underlying. The profit after costs at expiry price S is
P(S) = Σ s_i r_i (v_i(S) − e_i) − Σ r_i f_i
P is piecewise linear with kinks only at strikes. Breakevens are found exactly by locating sign changes between consecutive kinks and interpolating linearly. The maximum profit and loss are taken over S = 0 and every strike; above the highest strike the slope Σ s_i r_i over calls and underlying legs says whether the profit or the loss is unbounded. These figures are only defined when all legs share one expiry.
6. Probability of profit
The price at expiry is taken as lognormal around the forward, with the expiry's at-the-money implied volatility σ and time to expiry t in years (365-day year):
ln S_T ~ N(ln F − σ²t/2, σ²t)
the risk-neutral distribution implied by the market. With Φ the standard normal distribution and d(x) = (ln(x/F) + σ²t/2) / (σ√t), the probability of profit is the sum, over each interval [a, b] between consecutive breakevens where P > 0, of Φ(d(b)) − Φ(d(a)). A single flat volatility ignores the skew.
7. Volatility and basis
- ATM IV of an expiry: the mean of the call's and the put's IV at the strike nearest the forward.
- Fixed horizon
h(for example 7 days): total variancew(t) = IV(t)² × tis interpolated linearly between listed expiries, flat outside them, andIV(h) = √(w(h) / h). - Realized volatility over 7 days, from daily index closes
C_0 … C_7at 00:00 UTC:r_k = ln(C_k / C_(k−1)),RV = √365 × stdev(r_1 … r_7) × 100, with the sample standard deviation. The same formula on every venue. - Smile: the out-of-the-money option's IV at each strike, within
F × e^(±3σ√t). - Basis:
F / index − 1; annualized as× 365 / DTEonly from 7 days to expiry.
8. Margin in paper trading
Two estimates, from published rules. The standard model charges each sold option a percentage of the index plus its mark, with each venue's parameters; bought options need none. The scenario model follows Deribit's portfolio margin: the whole position is revalued with Black-76 (zero rate) for index moves from −15% to +15% in 2.5% steps and implied volatility shocked up by 45% and down by 30%, each scaled by (30 / DTE)^0.3 under 30 days and ^0.13 from 30 days, with a downward shock floored at 5% of IV. The worst loss on the grid, plus 0.01 × index per net sold contract per strike and expiry, is the maintenance margin; initial margin is 120% of it.
9. Scanner
The scanner builds 46 of the 58 strategies on every tracked venue, asset and expiry, around seven anchor strikes near the money and every width, and keeps the structures every leg of which can be executed and whose maximum profit after costs is positive. Results are recomputed every 30 seconds. The default order is probability of profit within risk tiers: limited loss with reward-to-risk of at least 0.20, other limited-loss structures, unlimited loss. Structures with the same payoff are collapsed into one.
10. Validation and limits
The formulas are tested against fixtures with hand-derivable answers, for example a bull put spread whose breakeven must equal the sold strike minus the net credit, and spot-checked against live data from the venues' public APIs. Known limits: top-of-book prices only; regular-tier fees without discounts; one flat volatility in the probability of profit; a zero interest rate in model curves; margin estimates that don't reproduce account-wide offsets. See the methodology page for the plain-language version and the data status page for live freshness.