Backpack Learn

Appearance

Glossary — Market Microstructure & Speed Bumps ​

If any term in the docs is unfamiliar, check here first.

Market Structure Basics ​

  • Market microstructure — The rules, mechanics, and plumbing of how an exchange actually works: how orders get matched, who gets priority, what fees are charged. Think of it as the exchange's "operating system."
  • Order book — The list of all resting buy orders (bids) and sell orders (asks) on an exchange, organized by price. The order book is the market.
  • Maker — A trader who adds liquidity by placing a limit order that rests on the book, waiting to be filled. They "make" the market.
  • Taker — A trader who removes liquidity by sending an order that immediately executes against a resting order. They "take" from the market.
  • Aggressive order — An order that crosses the spread and trades immediately (a taker order). A buy order at or above the best ask, or a sell order at or below the best bid.
  • Passive order — An order that doesn't trade immediately and instead rests on the book (a maker order). A buy below the best ask, or a sell above the best bid.
  • Spread (bid-ask spread) — The gap between the best bid (highest buy) and the best ask (lowest sell). A $100.00 bid and $100.05 ask = 5 cent spread. Tighter spread = cheaper to trade.
  • Depth (book depth) — How much size (quantity) is available at or near the best prices. "Deep book" = lots of volume near the top, meaning large orders won't move the price much.
  • Midpoint — The price exactly halfway between the best bid and best ask. Used as a reference for measuring execution quality. If bid = $100 and ask = $101, midpoint = $100.50.
  • Price discovery — The process by which the market determines the "correct" price of an asset through the interaction of buyers and sellers. New information (news, trades on other venues) gets incorporated into the price through order flow.
  • Fill rate — The percentage of your limit orders that actually get executed. Higher fill rate = your resting orders trade more often instead of sitting untouched.
  • Price-time priority — The standard rule for who gets matched first: best price wins, and at the same price, whoever arrived first wins. Almost all exchanges use this.
  • Basis point (bp / bps) — 1/100th of 1%. So 1bp = 0.01%. A 2.5bp fee on a $100,000 trade = $25. Used because percentages get awkwardly small in trading.

Speed & Latency ​

  • HFT (High-frequency trading) — Trading strategies that depend on extremely fast execution, often holding positions for milliseconds to seconds. Firms invest in specialized hardware and co-located servers to be microseconds faster than competitors.
  • Co-location (colo) — Placing your trading servers physically inside the exchange's data center, right next to the matching engine. This minimizes network latency to single-digit microseconds. Exchanges typically charge for this.
  • FPGA (Field-Programmable Gate Array) — Specialized hardware chips that can be programmed to process trading logic in hardware rather than software. FPGAs can react to market data and send orders in nanoseconds, much faster than any CPU-based system.
  • Latency arbitrage — Profiting from being faster than other participants. Specifically: seeing a price change on one venue and racing to trade on another venue before the quotes there update. The "arbitrage" isn't finding a mispricing — it's exploiting the time lag between venues. No skill, no insight, just speed.
  • Stale-quote sniping — The act of trading against an order that the poster would have cancelled if they'd had time. The quote is "stale" because it reflects an old price that has already moved on other venues. The "sniper" picks it off before the poster can react.
  • Information asymmetry — When one party in a trade knows something the other doesn't. In latency arb, the fast trader knows the price has moved on another venue, but the market maker hasn't seen it yet. The fast trader exploits this gap.

Market Maker Concepts ​

  • Inventory risk — The risk a market maker takes by holding assets. If an MM buys 10 BTC at $100,000 and the price drops to $99,000, they're down $10,000. MMs constantly try to stay close to flat (equal long and short exposure).
  • Adverse selection — When the trades you get filled on are systematically bad for you. For market makers, this means their buys tend to happen right before the price drops, and their sells happen right before the price rises — because the counterparty (the sniper) knew which way the price was going. It's not bad luck — it's a structural problem caused by faster traders picking off stale quotes.
  • Toxicity of flow — How "toxic" (harmful) the incoming order flow is to market makers. Toxic flow comes from informed traders who know something the MM doesn't (like a price move on another venue). Non-toxic flow comes from retail traders, hedgers, and others trading for reasons unrelated to short-term price prediction. MMs want non-toxic flow and avoid toxic flow.
  • Quoted spread — The spread you see displayed on the order book at any moment. The gap between the current best bid and best ask.
  • Effective spread — What traders actually pay, measured as 2x the distance between their fill price and the midpoint at the time of the trade. This captures the real cost of trading, including any price impact.
  • Realized spread — What the market maker actually earns after accounting for price movement. If an MM sells at $100.05 and the price stays at $100.00, the realized spread is $0.05. But if the price moves to $100.10 right after (adverse selection), the realized spread is -$0.05. This is the real measure of MM profitability.
  • Phantom liquidity (ghost liquidity) — Orders visible on the book that disappear before you can trade against them. Happens when MMs use ultra-fast cancel strategies — they show size to attract flow but pull it the instant anyone tries to hit it. The liquidity looks real but isn't.
  • Quote flickering — Market makers rapidly posting and cancelling orders, so their quotes only exist for milliseconds at a time. This reduces their exposure to sniping but makes the book unreliable for other traders.
  • Quote stuffing — Flooding an exchange's matching engine with massive volumes of order submissions and cancellations (thousands per second) not to trade, but to overwhelm competitors' systems, slow down their data processing, and game rebate calculations. Illegal in TradFi under market manipulation rules; largely unenforced in crypto.

Speed Bump Specific ​

  • Symmetric speed bump — A delay that applies equally to all message types: new orders, cancels, and modifications. IEX uses this. Everyone is delayed by the same amount, no exceptions.
  • Asymmetric speed bump — A delay that applies only to aggressive (taking) orders, while passive actions (cancels, modifications, new resting orders) pass through instantly. This gives market makers an explicit head start to update their quotes.
  • TSX Alpha — A Canadian stock exchange (part of the Toronto Stock Exchange group) that implemented an asymmetric speed bump. They delay only marketable (aggressive) orders by a few milliseconds while processing cancellations immediately. It was one of the first real-world implementations of an asymmetric design and generated significant debate about whether favoring one side of the market is fair.
  • Aequitas NEO — Another Canadian exchange that implemented an asymmetric speed bump, similar in philosophy to TSX Alpha. Canada has been a testing ground for these mechanisms.
  • Stickier books — When an order book has a speed bump (especially a longer one like 100ms), quotes tend to stay on the book longer instead of flickering on and off. The book is "sticky" — orders stick around. This means the displayed depth is more reliable and traders can trust that the liquidity they see is actually available to trade against.
  • Slower price discovery — The trade-off of a speed bump: if incoming orders are delayed, the order book takes longer to reflect new information. With a 100ms bump, the book could be up to 100ms behind the "true" price during fast moves. For retail traders this is irrelevant. For institutions trading large size during volatile moments, it means the book price might not match reality.

NBBO & Equity Market Structure ​

  • NBBO (National Best Bid and Offer) — The best available buy price (highest bid) and sell price (lowest ask) across all US stock exchanges at any given moment. Calculated by the SIP from quotes on NYSE, NASDAQ, IEX, CBOE, and 12+ other exchanges. By law (Reg NMS Rule 611), brokers cannot execute trades at prices worse than the NBBO.
  • Reg NMS (Regulation National Market System) — SEC regulation adopted in 2005 that links all US equity exchanges into a single national market. Rule 611 (Order Protection Rule) prohibits "trade-throughs" — executing at a price worse than the NBBO. Rule 612 prohibits sub-penny quoting on lit exchanges for stocks priced ≥$1.
  • SIP (Securities Information Processor) — The system that aggregates quotes from all US stock exchanges and calculates the NBBO. Two SIPs exist: CTA/CQS (for NYSE-listed) and UTP (for NASDAQ-listed). Adds ~500μs of processing latency. Firms can pay for faster direct exchange feeds to calculate their own NBBO.
  • Solver — An entity that competes to fulfill a user's intent at the best price. In DeFi, solvers route across DEXes and bridges. In TradFi, the solver is a broker with direct market access to stock exchanges. Solvers don't need to hold liquidity — they need to access it.
  • RFQ (Request for Quote) — A trading model where the trader asks dealers/solvers for a price rather than trading against a public order book. The trader expresses intent, multiple solvers compete to quote the best price, and the trader accepts the best quote. Dominant model for institutional OTC, options, and block trades. See RFQ Guide.
  • Price improvement — Getting a better price than the NBBO. If NBBO ask is $227.44 and you buy at $227.435, you received $0.005/share of price improvement. Typically comes from dark pool midpoint matching or wholesaler execution.
  • Protected quote — Under Reg NMS, a quote that must be honored and cannot be traded through. Only quotes from registered national securities exchanges with automated execution qualify. Dark pools don't publish protected quotes.
  • ATS (Alternative Trading System) — A trading venue that matches buyers and sellers but is not registered as a national securities exchange. Dark pools are the most common type. ATSes don't publish quotes to the SIP but must execute at or better than NBBO.
  • DMA (Direct Market Access) — The ability to send orders directly to an exchange without intermediary routing. TradFi brokers with DMA can access any US equity exchange, dark pool, or wholesaler to seek the best execution for their clients.
  • Wholesaler (retail market maker) — Firms like Citadel Securities and Virtu Financial that buy retail order flow from brokers (Robinhood, Schwab) and fill orders from their own inventory, typically at NBBO or better. They profit from the spread while providing price improvement to retail traders.
  • Best execution — A broker's legal obligation to seek the most favorable terms reasonably available for a customer's transaction. This includes price, speed, likelihood of execution, and settlement. Enforced by FINRA and the SEC.

Crypto Specific ​

  • MEV (Maximal Extractable Value) — Value that block producers (validators/sequencers) can extract by reordering, inserting, or censoring transactions within a block. It's the on-chain equivalent of front-running. Example: a validator sees your large DEX buy in the mempool, inserts their own buy before yours (pushing the price up), then lets your trade go through at the worse price.
  • Mempool — The queue of unconfirmed transactions waiting to be included in the next blockchain block. On public blockchains, everyone can see the mempool, which means traders can see your order before it executes and front-run it.
  • Sequencer — In Layer 2 / app-chain architectures, the entity that decides the order of transactions before they're finalized. Whoever controls the sequencer controls transaction ordering, which means they could extract MEV (or implement rules to prevent it).
  • Batch auction — Instead of matching orders one-by-one as they arrive (continuous matching), collect all orders over a time window, then match them all at once at a single clearing price. Within a batch, it doesn't matter who submitted first — everyone gets the same price. This eliminates latency advantage entirely within each batch.
  • Clearing price — In a batch auction, the single price at which all orders in the batch are matched. It's calculated to maximize the volume traded — the price where supply meets demand for that batch.
  • Commit-reveal scheme — A two-step process: first, submit a cryptographic hash of your order (the "commit" — proves you placed it without revealing what it is), then after a delay, reveal the actual order details. Prevents front-running because no one can see your order between commit and reveal.
  • Intent-based trading — Instead of placing specific limit orders, traders submit "intents" (e.g., "swap 1 ETH for at least 2,500 USDC"). Specialized solvers compete to fill the intent at the best price. Removes the order book entirely, so there's no latency race.
  • Co-location in crypto — Same concept as traditional finance: placing your servers physically close to the exchange's matching engine. Unlike TradFi, crypto co-location is mostly unregulated — exchanges can offer whatever access they want to whoever they want, which can create unfair advantages.