Market Making Strategy: Getting Paid to Provide Liquidity

The business model in plain terms

A market maker posts a bid and an ask simultaneously. If someone sells to the bid at 99.98 and someone else buys from the ask at 100.02, the maker has bought and sold the same instrument at a 4 cent difference, netting the spread. Repeat thousands of times per day and the accumulated spread is the revenue line.

The service being sold is immediacy. Buyers and sellers rarely arrive at the same instant, so someone must hold inventory in between. Market makers accept that inventory and the risk attached to it, and the spread is their compensation. In this sense the strategy is not speculation at all; it is a warehousing business with price risk.

Adverse selection, the central problem

Suppose you quote 99.98 bid and 100.02 ask. A large seller arrives and hits your bid repeatedly. You now own inventory at 99.98, and the reason you own it is that someone wanted to sell urgently. If they were selling because of information, the price will continue lower and your inventory loses more than the spread you earned.

• Widen quotes when uncertainty rises. Around news releases, earnings, and economic data, the proportion of informed flow rises sharply. Professional systems widen dramatically or withdraw entirely.
• Skew quotes based on inventory. If you are long, lower both quotes so you are more likely to sell than to buy. The skew is a price you pay to rebalance.
• Cancel fast. The ability to pull quotes in microseconds when the order book shifts is the primary technological advantage of professional firms and the reason retail cannot compete in contested venues.
• Size quotes by confidence. Small size in uncertain conditions, larger size in calm ones.
• Track flow toxicity. Measure how often your fills are followed by adverse moves. Rising toxicity means you are being picked off and should widen or stop.

A simplified quoting model

reservation_price = mid - (inventory * risk_aversion * volatility^2 * time_left)

half_spread = (risk_aversion * volatility^2 * time_left) / 2
              + (1 / risk_aversion) * ln(1 + risk_aversion / order_arrival_rate)

bid = reservation_price - half_spread
ask = reservation_price + half_spread

Interpretation:
  - Long inventory pushes the reservation price down, so both quotes
    fall and you are more likely to sell than to buy.
  - Higher volatility widens the spread and increases the inventory skew.
  - Faster order arrival allows tighter quotes, because you can
    rebalance more quickly.

The formula matters less than the two forces it encodes: inventory must be mean-reverting toward zero, and the spread must widen when the risk of holding inventory rises. Any working market-making system, however simple, must express both.

What a realistic retail version looks like

You cannot outrun automated systems in major venues. The accessible versions relocate the competition to places where speed matters less.

Risk controls that are not optional

1. Maximum inventory, in both directions, beyond which you quote one side only.
2. A hard loss limit per session that halts quoting entirely, enforced by code rather than judgement.
3. A volatility circuit breaker: stop quoting when realised volatility exceeds a threshold or when the spread in the book widens abnormally.
4. An event calendar filter: no quoting in the minutes around scheduled releases.
5. A stale-data killswitch: if the market data feed lags or disconnects, cancel all orders immediately. Quoting on stale prices is how automated systems lose large amounts quickly.
6. Position reconciliation: compare your recorded inventory with the venue’s record continuously. A desynchronised position is an unmonitored directional bet.