Desk dictionary.

The tendency of a resting quote to be filled most often when it is wrong. Traders who know more than the quote choose when to trade with it; the market maker doesn’t get to choose who trades.

Our take The cost that never appears on the ticket. It turns up later, in the mark.

The last day the data behind a figure runs to. A figure printed today may be as of an earlier day; the as-of date is the one that says how old it is.

Our take An undated figure is an old figure that hasn’t admitted it yet.

The risk that two positions meant to offset each other settle on slightly different things: another source, another time, another rounding rule. If the two readings land on opposite sides of a threshold, both positions can lose.

Our take Two contracts that sound alike are not a hedge until their rules agree, line by line.

A price someone stands ready to buy a contract at, with a size; the best bid is the highest one showing. A market maker’s bid is a standing offer to buy.

Our take A price without a size is a rumor.

The average of (forecast − outcome)² over a set of forecasts, counting the outcome as 1 if the event happened and 0 if it didn’t. Zero is perfect. Saying 50% about everything scores 0.25.

Our take Unflattering by design, which is what makes it useful.

How well stated probabilities match what then happens. Of all the contracts a calibrated forecaster prices at 70¢, about 70% resolve YES.

Our take Easy to claim and slow to show. A calibration curve needs a great many resolved contracts before it says anything you can’t argue with.

The organization that stands between the two sides of a matched trade, holds the collateral behind it and pays out at settlement, so neither side depends on the other to pay.

Our take It is why a fill on a regulated exchange is good, whoever was on the other side.

The requirement that related prices agree with each other. The outcomes in a set should add up to about 100¢, and on a ladder of “above” contracts a higher threshold should never be priced above a lower one.

Our take Some mispricing is visible without any model at all, which is why we check the whole set before any one price in it.

A mark taken from the cautious end of a reasonable range instead of its middle. For a contract you hold, that is the lower bound of an independent price estimate: an estimate of 40¢ to 46¢ is counted at 40¢.

Our take If a mark has to be wrong, it should be wrong in the direction that disappoints.

A dated entry at the foot of a published note saying what was wrong and what changed. The earlier claim is never changed silently, and every correction is also listed in one log, newest first.

Our take A mistake corrected in the open costs less than one found by a reader.

The period a set of data spans: from its first month to its as-of date, or “ongoing” while it is still being added to.

Our take Say where the data starts, so nobody mistakes a short history for a long one.

A short description of one set of data: what one row is, where it comes from, its fields, how it is aggregated and rounded, the period it covers, how often it is refreshed, its as-of date, what it leaves out and where it falls short. A card describes the data and carries no values.

Our take If you can’t say how a number was counted, it isn’t ready to print.

How much size rests in the order book at and near the best prices. A tight quote with nothing behind it moves as soon as anyone trades.

Our take Width is what gets quoted in reports. Depth is what someone actually trades against.

An exchange designated by the CFTC to list futures and other contracts, event contracts among them. It must meet the core principles of the Commodity Exchange Act, which include surveilling its markets, giving fair access and preventing manipulation.

Our take One rulebook, applied the same way to every account on the book.

Here, a binary contract: it pays a fixed amount, $1, if a stated event happens and nothing if it doesn’t. Its price in cents reads as a probability: at 62¢, the market is putting the event at about 62%.

Our take The plainest derivative going: one question, a fixed payout and a deadline.

A control fails closed when, unable to tell whether it is safe to carry on, it stops. A control that fails open keeps going while it works things out.

Our take A control that fails open is a log message.

Our own estimate of the probability that an event happens, written in cents. It sits at the centre of a quote, and it need not be where the contract last traded.

Our take The last trade tells you about one trader. Fair value is a claim about the event.

A cost rule: when part of a position is sold or settles, the contracts bought earliest count as the ones that left. It decides how a result splits between realized and unrealized.

Our take Dull, fixed and written down in advance, which is everything you want from an accounting rule.

An order, or part of one, that has traded. A partial fill trades some of the size and leaves the rest resting on the book. The figure on the Record page, trades executed, counts each fill as one trade, whatever its size.

Our take Getting filled is easy. Being glad you were filled is the job.

To hit the bid is to sell at someone’s bid. To lift the offer is to buy at someone’s offer. Both take liquidity from the book.

Our take Neither is rude. Both are informative.

Orders from traders who know something the quote doesn’t reflect yet: a faster read of a release, a better model, a closer reading of the rules.

Our take It can’t be picked out in advance, only priced in. It runs heaviest in the minutes after news.

The net position a market maker is left holding when trades on the two sides don’t cancel out. It carries the contract’s risk until it is offset or settles.

Our take A by-product of quoting, never the goal. One-sided flow leaves it on the desk regardless.

The risk that a price moves straight to a new level with no trading in between, as when a result is announced. Every event contract ends with a jump, to 0¢ or 100¢.

Our take Volatility measured on quiet days says nothing about the minute that matters.

A control that cancels resting orders and stops new quotes when a limit is breached, or when the system can’t confirm it is behaving as expected. Ours sits outside the strategy code, so a fault in the strategy can’t switch it off.

Our take The default state is no quotes. Everything else has to earn its way back on.

How easily a contract can be traded, in size, without moving its price much.

Our take Mostly a question of whether the quote is still there at the moment someone needs it.

A scoring rule that charges −log(p), where p is the probability a forecast gave to what actually happened. It punishes confident misses far harder than the Brier score does.

Our take Put 0¢ on something that then happens and the penalty is infinite, which is a fair summary of the experience.

A maker’s order rests on the book and waits to be traded against. A taker’s order trades at once against one that is already resting. Venues often charge, or credit, the two differently.

Our take Makers wait to be chosen and takers do the choosing. Most of market making follows from that asymmetry.

The value given to an open position before it settles, usually its current fair value or market price. Marking to market records gains and losses as that value moves.

Our take An estimate with a decimal point. Settlement is the only valuation that needs no opinion.

A trader who keeps a bid and an offer up on a contract, ready to trade either way, and earns the difference when both sides fill.

Our take Paid to be there. Charged for being wrong.

The price halfway between the best bid and the best offer.

Our take A handy reference point at which nobody is actually offering to trade.

Combining positions that depend on the same underlying fact, so limits apply to their joint result in each outcome. A position that pays when the others lose reduces the worst case.

Our take Two modest positions on the same reading are one immodest position.

A price someone stands ready to sell a contract at, with a size; the best offer is the lowest one showing.

Our take Seen from a buyer in a hurry, the offer is the price of not waiting.

The list of resting bids and offers on a contract, by price and size.

Our take The one place in a market where opinions come with size attached.

A group of contracts on mutually exclusive outcomes of one event (Outcome A, Outcome B, Outcome C), exactly one of which pays. Their prices should add up to about 100¢.

Our take When a set adds up to 104¢, fix the arithmetic before forming any view on the event.

For a contract priced at p, the variance of its payoff is p(1 − p). It is largest at 50¢ and shrinks toward 0¢ and 100¢.

Our take The reason two cents of width at 50¢ and two cents at 90¢ are different deals.

A cap on exposure, set per contract, per event or across the whole book, and measured as the worst loss at settlement.

Our take Set before the trade, while everyone is calm.

An order accepted only if it would rest on the book. If it would trade at once against a resting order, the exchange cancels it (or, on some exchanges, reprices it) instead.

Our take The line between making and taking, enforced at the door.

A test an order must pass before it is sent: its size, its price against a band around fair value, and what the event’s worst case would be if it filled.

Our take The cheapest place to catch a mistake is before it has a counterparty.

A way of scoring probability forecasts under which your best expected score comes from reporting what you actually believe. The Brier score and log loss are both proper.

Our take It makes shading a forecast a bad habit with a measurable cost. That is the whole design.

Trading with a firm’s own money, for its own account, with no client funds involved.

Our take Nobody else’s money is exposed, and there is nothing to sell you. Both are on purpose.

Where an order sits among others resting at the same price. Venues commonly fill resting orders at a price in the order they arrived, so earlier orders fill first.

Our take The front of the queue fills first, including when the flow is informed.

A bid and an offer shown together, each with a size: 61¢ bid, 63¢ offered, 100 contracts a side.

Our take A forecast you are obliged to trade on, which is what makes it worth reading.

A term in a market-maker agreement setting how a market maker must quote: for example, for how much of the trading day, at what maximum width and at what minimum size, and on which contracts.

Our take A promise about the busy hours, not only the quiet ones.

A realized result is locked in: the position has been sold or has settled. An unrealized result is the current mark on open positions minus what they cost, and it can still change.

Our take Unrealized is a forecast about your own results, which is why we count it at the conservative end.

A chart of forecast probability against observed frequency, with forecasts grouped into bins. A calibrated forecaster’s points sit on the diagonal.

Our take Points on the diagonal look reassuring. The count in each bin tells you whether to be reassured.

The moment an event’s outcome is determined under the contract’s rules.

Our take The rulebook decides it. The first headline only thinks it does.

The source a contract’s rules name for deciding the outcome, such as an official publication or a named measurement at a stated time: Station A’s reading at 4 p.m., say.

Our take Read it before the price. Most surprises at settlement were written down in advance.

A change to a figure that was already published, because it was counted wrong or the data behind it changed. It is dated, says what changed and why, and is listed with the corrections.

Our take A figure restated in the open is worth more than one quietly replaced.

An exchange’s rules, together with the terms of each contract it lists: what the contract pays on, the source that decides it, when trading stops, its tick size and any position limits.

Our take The fine print is the product. Read it before the price.

Information published at a known time, such as an official statistic. Every contract tied to it can reprice in the minute it comes out.

Our take The calendar is the most useful risk model nobody brags about.

An exchange control that stops an order from filling against a resting order from the same account or owner, by cancelling one side or both instead.

Our take A trade with yourself is never the trade you meant to make.

When a resolved contract pays out: 100¢ to the side that was right and 0¢ to the other.

Our take A position is open until it settles, however obvious the outcome looks.

Placing orders you intend to cancel before they can fill, to mislead others about supply or demand. It is prohibited on regulated US markets and by venue rules.

Our take A quote you won’t honor is a lie with a timestamp.

The gap between the best bid and the best offer on a contract. When the quote is our own, we call it width.

Our take It pays for immediacy, adverse selection, inventory and running costs. It pays nobody for being absent.

An input to a quote (fair value, the order book, the reference data a contract settles on) that has stopped updating within its expected interval.

Our take Width prices the risk you can see. A stale input is risk you can’t, so the quote comes down until it recovers.

A series of contracts on one measurement at different thresholds: Station A’s high above 80°F, above 85°F, above 90°F. On a ladder like that, a higher rung can never be worth more than a lower one.

Our take Ladders are where incoherence hides, one rung at a time.

A research note built on data described by a dataset card. It states its question, the data it uses and the date that data runs to, its method step by step, and its limits, so a reader can follow how each figure was reached.

Our take A study shows its working before its answer.

What a position gains or loses at settlement, in each outcome. A YES bought at c cents ends at +(100 − c) or −c per contract.

Our take The one risk number a calm week can’t talk down.

The smallest step a contract’s price can move in, often 1¢ on event contracts.

Our take Near 1¢ and 99¢, a single tick is a large share of what is left to move.

A note’s version is one plus the number of its corrections: a note with none is version 1. The date it last changed is the latest of the date it was published and the dates of its corrections.

Our take A version tells a citation which text it means.

A trade in which the same party, or parties acting together, sit on both sides, creating volume with no real change in who holds what. It is prohibited on regulated US markets and by venue rules.

Our take Volume traded with yourself adds nothing to a market except a misleading statistic.

The distance between a market maker’s own bid and offer, in cents. A 61¢ bid and a 63¢ offer are two cents wide.

Our take Narrow is easy in a quiet hour. The width that counts is the one still up when the news lands.

The two sides of an event contract. YES pays if the event happens and NO pays if it doesn’t, so the two prices on one contract come to about 100¢.

Our take Buying NO at 37¢ and selling YES at 63¢ end in the same place. The paperwork differs.