One order path, used five times.
SkyQ is the shared machinery behind every book: how a signal becomes an order, what a fill is assumed to cost, and what happens when a book stops behaving. The books differ in their rules. They do not differ in how they trade.
Six steps, no branches.
A signal that cannot be executed by this path is not a signal. That constraint is what keeps five strategies from becoming five different businesses.
Evaluate on the completed bar
Each book recomputes its own features from raw OHLCV. No indicator is trusted from a cache, and no decision is taken on a bar that has not closed.
Refuse the trade if the tape is wrong
Stale data, a spread wider than the model, a bar older than the timeout, or a book at its contract cap: any one of them drops the signal and logs why.
Reprice the contract at the signal level
The chain snapshot is not trusted for delta at signal time. The contract is repriced from the underlying print, so the limit reflects the market as it is, not as it was when the chain was polled.
A marketable limit through the spread
Never a market order. The limit is priced a defined number of ticks through the touch, so a fast tape cannot take an unbounded fill while a passive one still has a chance to improve.
Stop, target and time exit in one batch
The children are submitted with the entry as an OCA group. The position is never naked, and no exit depends on the process waking up at the right moment to place it.
Nothing rests overnight
A hard flatten before the close cancels any unfilled entry and closes anything still open. Overnight gap risk is not part of the model, so it is not part of the book.
Why one path matters
Five strategies with five executors is five places for a fill assumption to drift, five stop-loss implementations to audit and five ledgers to reconcile. Here there is one executor and one ledger; a book is a rule set that produces a signal object, and nothing else.
What the books do not control
- They cannot size beyond the contract cap.
- They cannot place a second entry after a stop.
- They cannot widen a stop or extend a time exit.
- They cannot trade outside the regular session.
Five books in one account means a broker or network failure hits all of them at once. That is an accepted trade-off for a desk of this size, and it is why the governor — not the books — owns the kill switch.
The edge is a passive fill. Say it plainly.
Exits are modelled as resting limit orders that fill when the underlying touches the level. Re-run the same 9,251 signals with the opposite assumption — exit at the printed close of the touch minute — and every book loses. This is the most important number in the plan.
| Book | Net, resting limits | Net, close of touch | Swing |
|---|---|---|---|
| SkyQ WICK | +$10,184 | −$4,477 | −$14,661 |
| SkyQ LEVEL | +$29,993 | −$26,358 | −$56,351 |
| SkyQ RANGE | +$11,542 | −$16,274 | −$27,816 |
| SkyQ SIGMA-S | +$16,433 | −$11,944 | −$28,377 |
| SkyQ SIGMA-Q | +$16,931 | −$11,664 | −$28,595 |
| Portfolio | $95,038 | −$70,717 | −$165,755 |
Same signals, same contracts, same fees. The only difference is whether an exit gets the passive fill or pays the printed price. Hypothetical in both columns.
Touch equals fill
A resting limit at a level is filled when the underlying trades through that level. No queue position, no partial fill, no improvement.
The orders really do rest
These are GTC limit exits sitting in the book before the level is reached, not orders typed after the fact. That is a materially different claim from a backtest that fills at the close.
Queue position is real
On a fast touch, the resting order may be behind size that clears first. The first live quarter measures this per book, in cents per contract, and the answer decides whether the desk scales.
Tradier Pro at $0.21 a round trip.
Fee structure is not a footnote at 722.8 trades a month — it is roughly 40% of the gross edge at retail pricing, and about 2% at Tradier Pro. The whole replay was re-priced on this schedule.
| Line item | Tradier Pro | Per round trip | 12.8 months (9,251 RT) |
|---|---|---|---|
| Equity & ETF option commission | $0.00 per contract | $0.00 | $0 |
| Clearing | $0.0775 per contract, each side | $0.155 | $1,434 |
| ORF | $0.02 per contract, each side | $0.04 | $370 |
| Exchange / regulatory fees | pass-through | ≈$0.015 | $139 |
| Platform | $10 per month | — | $128 |
| All-in | Tradier Pro | $0.21 | $2,071 |
TAF is $0.00 through 31 December 2026 and is not modelled after that. Index options are $0.35 per contract on this schedule; the desk trades ETF options only. Compare: the same 9,251 round trips at $1.30 would cost $12,026.
The only component allowed to override a signal.
A strategy cannot be trusted to stop itself; the thing that stops it has to sit outside the strategy. These are the invariants the governor enforces on every cycle.
Why long premium only
A debit is a defined loss. There is no assignment to manage, no short-gamma tail, no margin call at the worst moment, and the cash floor can be computed before the trade rather than discovered after it. The desk gives up the higher win rate of premium selling in exchange for a loss that cannot exceed what was paid.
Long premium bleeds theta and pays the spread twice. Every book has to clear that drag before it earns a dollar, which is why the trade counts are high and the holds are short: minutes, not days.
What the framework does not do.
No forecast of direction
Each book is a conditional response to a completed bar. It has no view on where the index closes, and no model that claims one.
No live parameter fitting
Parameters were fixed before the replay and are not re-optimised on live data. Re-fitting in production is how a backtest becomes a story.
No live record yet
12.8 months, one volatility regime, no broker statement. The first live quarter is a measurement exercise, and it can fail.