Five books. One tape. Zero improvisation.
SkyQuant Capital Management is building a systematic SPY and QQQ options desk. Five independent Python processes, 9,251 replayed trades against real option prints, one contract per signal, and a cash floor that is known before the first order goes out.
A narrow business, run on purpose.
Two underlyings, one option type, five rule sets. Nothing here is discretionary, and nothing is levered beyond the debit paid. The whole desk fits in one brokerage account and one repository.
SPY and QQQ, long premium only
A call on a long signal, a put on a short signal, at or near the money, one to three days to expiry, held for minutes. Maximum loss on every trade is the debit, known before entry. No spreads to leg into, no assignment risk to manage, no margin spiral.
On the closed bar, never inside it
Each book reads a completed five-minute bar, applies one deterministic rule, and sends a marketable limit through the spread. The bracket — stop, target, time exit — is submitted in the same order batch. The process cannot change its mind halfway through a trade.
No averaging down, no re-entry
One position per book. A stopped trade is done for the day in that book. No martingale, no widening a stop, no second attempt at the same level. The rules that produced the replay are the rules that trade.
SkyQ is the method, not a product.
Everything the desk does — signal evaluation, fee-aware sizing, order construction, the risk governor, the nightly ledger — is one framework, applied five times. A book is just a rule set plus a code that says what it hunts.
One order path
Every book emits the same object: side, quantity, limit price, stop, target, expiry, time stop. One executor validates it, prices it against the live chain, and sends it. There is no second implementation to drift out of sync, which is the usual way a five-strategy desk quietly becomes five different desks.
One fee model
Tradier Pro: $0.00 equity-option commission, about $0.21 all-in per round trip after clearing, ORF and exchange fees, plus $10 a month. Every number on this site is net of that bill. At the 722.8 trades a month the desk actually runs, cost is a rounding error rather than the edge.
One fill assumption
Exits are modelled as resting limit orders that fill when the underlying touches the level. That is the entire edge, and it is stated on every page rather than buried in a footnote. The first live quarter exists to measure the gap between that assumption and reality.
One risk governor
A per-book contract cap, a floor that is recalculated nightly, and a kill switch that flattens and stops a book when its drawdown crosses its own limit. The governor is not advisory. It is the only component allowed to override a signal.
Successful systematic firms name the method once and then let the books carry short codes — Man AHL, AQR Helix, Two Sigma Spectrum. SkyQ is that name here: one framework, five codes. WICK trades rejection wicks, LEVEL trades reclaims of a rolling level, RANGE trades channel reversion and SIGMA trades VWAP band reversion on both indices.
Each code says exactly what it hunts.
Modelled net income, one contract per signal, after Tradier Pro's full bill. Cash floor is the maximum concurrent premium plus that book's worst drawdown — the amount that has to be sitting in the account for the book to survive its own history.
| Book | Underlying | Hunts | Trades | Win | PF | Net/mo | Floor |
|---|---|---|---|---|---|---|---|
| SkyQ WICK | SPY | Rejection wick continuation | 651 | 83.3% | 2.05 | $850 | $1,863 |
| SkyQ LEVEL | QQQ | Reclaim of a rolling low | 1,782 | 76.4% | 1.49 | $2,493 | $5,904 |
| SkyQ RANGE | SPY | Channel mean reversion | 2,998 | 84.5% | 1.37 | $1,154 | $2,467 |
| SkyQ SIGMA-S | SPY | VWAP band reversion | 2,018 | 86.4% | 1.65 | $1,454 | $1,746 |
| SkyQ SIGMA-Q | QQQ | VWAP band reversion | 1,802 | 86.0% | 1.55 | $1,475 | $2,121 |
| Portfolio | 5 books | — | 9,251 | 83.6% | — | $7,426 | $14,101 |
Replay 2025-10-06 → 2026-10-31, one contract per signal, net of $0.21 per round trip and the $10 monthly platform fee. Hypothetical. Full book detail →
The structure changes at $1.5M, not before.
Today this is a single-member LLC trading its own capital. That is the correct structure for a desk with no live record: cheap, fast, and impossible to oversell. It becomes a limited partnership only when there is enough committed capital to justify counsel, an audit, a fund administrator and a compliance calendar.
The threshold is $1,500,000 in firm, signed commitments — not pledges, not interest, not a group chat. A commitment counts only when a subscription is signed by a verified accredited investor and the money is in escrow. Below that line the fund does not exist, and no outside capital is accepted.
Single-member LLC
Own capital only. No clients, no custody, no advisory relationship. Formation, trademark, DBA and domain completed.
Live on Tradier Pro
Five books, one contract per signal, every fill reconciled against the replay assumption nightly.
Delaware LP, LLC as general partner
PPM and LPA drafted by fund counsel, administrator appointed, audit engaged, subscriptions accepted through escrow.
The limits, before the pitch.
ProvenBuilt and tested
9,251 trades replayed against OPRA contract prints. Deterministic rules. 24–56 offline unit tests per book. Full commission, regulatory and slippage model. No language model anywhere in the order path.
ModelledNot yet live
The replay covers 12.8 months and one volatility regime. Exits assume resting limit fills. No live fill, no paper fill and no broker statement exists yet. Every dollar figure is hypothetical.
Not claimedNo crisis record
No 2008 options track record. The historical options archive starts in October 2025. One book's underlying signal has been tested through 2007–2012, and only on shares. No audited returns, no client capital.
The entire edge depends on resting limit exits. Exit the same 9,251 signals at the printed close of the touch minute instead, and all five books lose money. That is not a footnote — it is the engineering risk the first live quarter is designed to measure, and it is why the launch plan has measurement gates rather than revenue targets.