Quoting engine
Spread and size as functions of short-horizon volatility, queue position and current inventory. Quotes are priced from the microprice rather than the mid, and they are firm: placed to be filled, sized to be defensible.
Proprietary trading · Market making
Proprietary market making in liquid electronic markets.
Penguin Path Advisors streams firm two-sided quotes, carries the inventory on our own balance sheet, and earns the spread for bearing risk that other participants want to shed — systematically, at scale, on our own capital.
No clients. No external money. The desk’s edge is measured in basis points and defended in microseconds.
01 — The business
Market making has a simple income statement.
On every round trip the desk earns the half-spread and pays three costs: adverse selection (the markout against informed flow), hedging slippage, and fees. In expectation, per unit traded:
E[π] = δ̄ − m̄ − h̄ − f̄, profitable iff δ̄ > m̄ + h̄ + f̄
(P&L)
Every system on the desk exists to move exactly one of those terms. Quoting models raise the captured spread δ̄. Toxicity filters and markout attribution cut adverse selection m̄. Cross-hedging engines cut h̄. Venue selection and market-making scheme economics cut f̄. That inequality is the whole business — everything else on this page is machinery for widening it.
02 — The desk
An inventory problem wearing a pricing problem’s clothes.
Anyone can quote a spread. The business is deciding how wide, how deep, and how skewed — per instrument, per venue, every few microseconds — while the inventory you accumulate argues with the prices you show.
2.1
2.2
Inventory & skew
Position shifts the reservation price; the reservation price skews the quotes; hard bounds cap what the model may accumulate. Inventory mean-reverts to target through the session and ends it flat by design.
2.3
Adverse selection
Every fill is marked out at multiple horizons and attributed. When measured toxicity rises — informed flow, regime breaks, stale-quote sniping — spreads widen and size steps down. Quotes widen; they do not vanish.
2.4
Hedging & risk transfer
Inventory that cannot be worked off is hedged in correlated instruments — index against single names, futures against cash — with basis risk measured, not wished away. Exposure is managed at the book level, not per symbol.
price size
ask 101.27 1,900
ask 101.26 3,100 ← ours
mid 101.245 spread 2.0 bp
bid 101.24 3,400 ← ours
bid 101.22 1,800
inv −420 · skew +0.5 bp · σ̂(30s) 4.1 bp
Exhibit 1. Illustrative quote placement. The desk’s quotes sit one tick inside the away market on both sides; a short position of 420 units shifts the quote centre 0.5 bp above the prevailing mid against the current 30-second volatility estimate. The engine prices from the microprice; the ladder is annotated against the mid for readability. Values are illustrative, not a performance representation.
03 — The model
The model is public. The edge is in the calibration.
The starting point is the Avellaneda–Stoikov formulation: given mid price s, the desk quotes not around s but around a reservation price, with a spread that never collapses to zero.
r(s, q, t) = s − q γ σ2(T − t)
(1)
δa + δb = γ σ2(T − t) + (2/γ) ln(1 + γ/κ)
(2)
Here q is signed inventory, γ the risk-aversion parameter, σ a short-horizon volatility estimate, κ the order-book liquidity parameter, and T − t the time remaining in the session. Equation (1) is why inventory skews the quotes; the second term of (2) is why spreads stay open even at zero inventory; the factor (T − t) is why inventory weighs most heavily when there is most session left to carry it — the end-of-day flat is enforced by an explicit control, not by the model.
The equations are on every desk’s whiteboard. What separates desks is everything the paper leaves as an exercise: estimating σ at thirty-second horizons, fitting κ per venue per hour of day, valuing queue position, pricing from the microprice — and knowing the regimes in which the model must not trade at all.
04 — Risk & governance
The controls are not an afterthought. They are the product.
A market maker’s licence to operate is earned by behaving well in fast markets. Every control below is enforced in the path of the order — in software, not in a policy document — with a full, replayable order-lifecycle audit trail. Risk limits are owned by management, set in writing, and cannot be widened by the systems they constrain.
| Hard limits, in-path | Market conduct |
|---|---|
| Position and notional caps, per instrument and per book | Quotes are firm and executable — placed to be filled |
| Order- and message-rate throttles | Spreads widen under stress; they do not vanish |
| Fat-finger price collars on every order | Systems and controls aligned with international algorithmic-trading standards, including MiFID II RTS 6 where we face EU venues |
| Kill switch; cancel-on-disconnect at every venue | Order-lifecycle audit trail, replayable |
| Self-match prevention | Venue market-making scheme obligations honoured |
05 — Contact
Venues, clearing, counterparties.
For venue and liquidity-programme relationships, clearing and prime enquiries, or counterparty due diligence, write to the desk. Engineers and researchers: we are always reading CVs from people who think in nanoseconds and basis points.
contact@penguinpathadvisors.comPenguin Path Advisors Ltd is a Cayman Islands company trading proprietary capital only. We do not manage external money, accept client funds, or provide investment services, portfolio management or investment advice to any third party. Nothing on this site is an offer, a solicitation, or advice.