Market Making
Quoting a two-sided price — a bid you will buy at and an ask you will sell at — and standing ready to trade either side on demand. It is the strategy family the rest of this wiki’s stack presupposes: a bot that sends a market order is buying from a market maker.
The obligation comes first
The MIT guide defines the job as availability, not prediction: “the job of a Market Maker is to always be available to buy or sell a particular financial instrument at the prices they quote.” Exchanges partner with market makers so that securities stay liquid — tradeable without the price lurching on volume — and in return grant privileged access to order-flow information. Roughly 100,000 primary equities exist and most of them have no natural counterparty waiting on the other side at any given moment; the market maker is the counterparty of last resort, and is generally required to honour the prices it shows.
The spread x@y (bid @ ask) is the compensation for that obligation, and exchanges often cap its width
(ask − bid < k). Every fill hands the maker inventory it did not want — buy from a seller and you now
own something that can fall — so the spread is priced against inventory risk, and post-trade hedging with
options or correlated instruments mitigates what the spread doesn’t cover.
Three inputs to a quote
- Theoretical value — what you think it is worth. Certainty tightens the market and uncertainty widens it: a die roll is 3.5 and quotable tight; the number of ping-pong balls that fit in the Empire State Building is not.
- Last traded price — where the market actually is. When it diverges from your theo, you are weighing your model against everyone else’s, and each observed trade updates the estimate.
- Current position — your net exposure. Market makers want to be flat. Flat quotes symmetrically
around theo; long an asset worth $0.50 quotes something like
$0.43@$0.53, giving up edge on the sell side to shed inventory.
The width is a confidence interval by another name, and the guide teaches it that way: wider under more
uncertainty. But an interval is only a market if someone will trade it. 0@1 billion contains the true
value and is worthless, because “the tightest and fastest markets usually get the majority of trades.”
Correctness and competitiveness pull against each other, and that tension is the whole job.
Adverse selection is the adversary
The counterparty trades against you precisely when they know more. Show 350@370 on a quantity that is
truly 400 and an informed counterparty lifts your ask and takes 30 from you. Two consequences the guide
draws out:
- Every fill is information. If they bought, they think you were cheap — so move the market up, both to incorporate their view and because otherwise they will simply buy again at the same price.
- Trades on both sides bracket the truth. Once someone has both bought and sold to you, their fair value is boxed in between the two midpoints, and subsequent quotes should move less. The guide’s worked game generates 9 units of riskless PNL from two trades this way — profit taken purely from having read the counterparty, not from being right about the underlying.
Skew follows from the same reasoning: if you expect the counterparty to value the thing far above you, you can quote a bid above your own theo, because the trade you actually expect is them buying at your ask.
Where it sits in this wiki
Market making is named in the strategy taxonomy on algorithmic-trading alongside arbitrage, mean reversion, scalping and momentum, and is one of the five tasks trademaster supports. It is the corner of the field where market microstructure (spread, inventory, order flow, adverse selection) is the subject rather than a friction, which is why the standing “beyond crypto / microstructure is unsourced” gap in synthesis narrows here.
The evidence in this corpus is pedagogical, not operational: the guide teaches market making as an interview exercise on quantities like baseball wins and national populations. No source here yet describes a production market-making system — inventory limits, quoting engines, latency, or the exchange programs that grant maker status.
Related
quant-bible-mit-sloan · quant-interview · quant-trading-firm · algorithmic-trading · event-driven-trading · trademaster · synthesis