Quant Memo
Core

Inventory Limits And Position Caps

The hard rules a market maker sets on how much of one instrument it will hold, and why every quoting decision ultimately bends around them.

Prerequisites: Inventory Management for Market Makers, Skewing Quotes To Manage Inventory

A market maker earns the bid-ask spread by constantly buying from sellers and selling to buyers, but every trade it does leaves it holding — or short — some amount of the instrument. If a market maker only ever bought, it would eventually own a warehouse of stock it never wanted, exposed to every price move against it. Inventory limits are the guardrail that stops "buy low, sell high, repeat" from quietly turning into "accumulate a giant directional bet."

The idea in one line

A position cap is a maximum number of shares (or notional dollars) a market maker will hold, long or short, in a given name. Hit the cap and the desk stops quoting on that side, or starts pricing itself out of further trades in that direction.

How the limit gets set

The cap isn't arbitrary. Desks size it against three things: how much capital is backing the strategy, how volatile the instrument is, and how fast the position can be unwound if the market turns. A cap of 50,000 shares means something very different for a stock that trades $2 million a day than for one trading $200 million a day — the first could take hours to exit without moving the price, the second could be flattened in seconds. A common rule of thumb caps the position at some multiple of the average daily volume the desk can realistically trade out of within an acceptable time window without excessive market impact.

Worked example: quoting into a limit

A desk sets a cap of ±20,000\pm 20{,}000 shares in a stock. It starts the day flat. A wave of sell orders lifts its offers repeatedly; after a string of buys the desk is long 15,000 shares — 75% of the cap. Two things happen at once: the desk skews its quotes, lowering both bid and offer slightly so it's less attractive to sell to and more attractive to buy from (a buyer is more likely to lift an offer that's already cheap), nudging the position back toward flat. If the position keeps growing and hits 20,000, the desk pulls its bid entirely — it will still offer stock (since selling reduces the long position it's trying to shed) but will no longer buy more, no matter how attractive the price looks, until inventory comes back under the cap.

long inventory, shares quote price mid cap: 20,000 offer (barely moves) bid (drops away) bid pulled
As the desk's long position climbs toward the 20,000-share cap, the bid drops away (less willing to buy more) while the offer barely moves — until the bid is pulled entirely at the cap.

As the position climbs from flat toward the cap, the bid and offer skew apart — the closer to the limit, the more one-sided the quotes become, until one side stops posting altogether.

What this means in practice

Position caps are set per instrument, per desk, and often aggregated across an entire trading floor so that correlated books (say, all the semiconductor names one trader covers) don't collectively blow through a firm-wide risk limit even if no single name does. Breaching a cap isn't just a P&L question — many firms treat it as a hard risk-control event that triggers automatic order cancellation, and repeated breaches usually mean the desk's sizing was wrong, not that the limit should simply be raised.

An inventory limit converts an open-ended risk (accumulating an unbounded position) into a bounded one, by making the market maker's own quoting behavior — skewing prices, then refusing one side of the market — the mechanism that keeps the position inside a pre-set band.

Related concepts

Practice in interviews

Further reading

  • Avellaneda & Stoikov, High-frequency trading in a limit order book
  • Cartea, Jaimungal & Penalva, Algorithmic and High-Frequency Trading, ch. 10
ShareTwitterLinkedIn