What your money does between the trade and the resolution

The two moments anyone thinks about are the entry and the outcome. The longest part of a position is neither. It is the interval between them, when the event has not happened, the money is not yours to spend, and nothing on the screen moves except the price.

That interval is where an event contract differs most from the instruments people arrive comparing it to. It is not a deposit against a larger exposure, it cannot be levered, and on most of these venues it cannot be borrowed against. What it is, for the whole of its life, is cash you handed over — held by a clearing house, earning something for somebody, and unavailable to you until a question about the world is answered.

This page is about that interval only. It stops where how money gets back out starts, which is the moment the contract has resolved and you want the proceeds in a bank account.

How it works

A dollar of position is a dollar of cash

The rule is written into the clearing rulebooks rather than into any product page, and the regulated US venues state it in nearly the same words.

ForecastEx's rulebook, version dated 17 September 2026, puts it in the trading chapter as a flat sentence: Forecast Contracts are non-marginable and have no loan value. Its Rule 608, titled Full Collateralization Required, says each member must deposit and maintain funds to fully collateralise any contract held, and that a bid cannot even be entered unless the collateral for the position it might create is already on deposit. Its definitions make the term a borrowed one rather than a house one: a fully collateralised position is the thing defined in CFTC Regulation 39.2.

Kalshi says the same twice, once at the exchange and once at the clearing house. Exchange Rule 6.1(b) states that all trader positions are fully cash collateralised and that no member can take positions leading to an exposure exceeding the funds deposited in the relevant member account. At Kalshi Klear the whole of Chapter 7 is headed Margin, and its first rule — Rule 7.1, Full Collateralization of Contracts Required — says all trades must at all times be fully collateralised and that the company will not clear a position that is not. The chapter then gives the consequence, which is the clearest statement of the design anywhere in these documents: the clearing house maintains no financial resource package for a clearing member default, because it clears only fully collateralised positions. There is no default fund because there is nothing to default on. All of that is in the rulebook amendments filed with the CFTC on 2 May 2025.

The same rule has a second half that is easy to miss. Klear Rule 7.1(B) requires an FCM to reserve a customer's funds and make them unavailable for uses other than collateralizing an order from the moment it is submitted until the order resolves through cancellation, expiry, rejection or execution. The lock therefore starts at the order, not at the fill. A resting limit order that nobody has taken is already holding your money.

Polymarket US builds the same idea into its vocabulary instead of into a margin chapter. Its rulebook, dated 14 September 2026, defines Settlement Amount and Payout Condition in terms of a Fully Collateralized Position, so the collateralised position is the unit the rest of the rulebook is written about.

What follows from this is the thing worth internalising: your maximum loss and your capital commitment are the same number. On a levered instrument they are not, and the habit of thinking about a stake as a fraction of the money it controls does not transfer. Eighty-eight cents at risk means eighty-eight cents gone, not eighty-eight cents of margin against a dollar.

Both sides' cash is held, and the pot is exactly the payout

The clearing house is not holding one side's money. It is holding all of it.

ForecastEx documents this with an arithmetic example in the rule itself. Bids are priced between $0.01 and $0.99, and when the bids for the Yes and No positions sum to $1.00 the exchange pairs them. Each party then posts collateral equal to the price it entered at — the rulebook's own illustration is a $0.20 Yes bid paired with a $0.80 No bid, where the Yes participant posts $0.20 per contract and the No participant posts $0.80. One dollar per contract goes in; one dollar per contract is what settles.

Two consequences follow. The first is that the payout is not topped up from anywhere: whatever a contract pays out was posted by the person on the other side of it before the event happened. The second is the one that matters for this page — at any moment, the total capital immobilised by a market is its open interest multiplied by a dollar, and every cent of it belongs to a trader who cannot spend it.

Leaving early is a claim on somebody else's order

"I can always sell before resolution" is not a property of the contract. It is a prediction about a stranger.

These venues match orders; with the partial exception of an automated market maker, none of them buys your position back. An exit at a price you would accept requires a resting bid at that price, and on a long-dated question, a small market or a contract that has drifted to one extreme, there may be no bid worth taking, or none at all. The mechanisms, and which venue uses which, are the subject of where the liquidity comes from — read it before sizing anything you expect to trade out of rather than hold.

Two details from that page bear directly on the waiting. A market maker program is a commercial arrangement rather than an obligation owed to you: ForecastEx Rule 608(b) allows a market maker alternative collateral terms under an agreement with the exchange, which is a statement about the exchange's incentives to have quotes there, not a promise that quotes will be there on your contract. And a wide book is not a neutral inconvenience when you are the one leaving — the spread is charged once on the whole position, at the moment you have already decided to go.

The honest planning assumption for a long-dated position is that you will hold it to resolution, and that any earlier exit is a bonus. If that assumption changes the size you would take, it is the assumption doing the work, not the forecast.

Four different moments get called "the end"

A market that has stopped trading has not paid. The regulated venues separate the moments in their definitions, which is the most reliable place to see how many there are.

The Polymarket US rulebook defines four separately: Last Trading Date, the last date on which trading is permitted, as specified in the contract terms; Last Trading Time, the last time on that date; Expiration Time, the designated moment the fully collateralised position expires or terminates; and Settlement Date, the date on which amounts owed must be paid. Nothing in the definitions requires any two of them to coincide, and the contract terms are where you find out whether they do.

On some contracts they do coincide, and it is worth seeing one. The ForecastEx unemployment contract's terms and conditions set Last Trading Time as same as Resolution Time, Expiration Time as same as Resolution time, and settlement no later than the day following expiration, unless the outcome is under an Event Review. Three moments collapse into one and the fourth is a day behind — because the contract is pinned to a scheduled statistical release at 7:30 AM CT. That is the tightest case in this catalogue, and it is tight for a reason that does not generalise to a question about an election or a court.

The gap opens where the answer has to be established rather than read off a schedule. Robinhood's settlement article says plainly that settlement does not occur immediately after an event concludes, because the outcome needs to be officially verified first; it gives one to two business days for sports outcomes and says complex events such as elections can sometimes take much longer. It then adds a further step — after settlement your money may not be withdrawable the same day, and typically you wait one business day more.

On chain the last moment is a transaction rather than a date. Limitless's documentation resolves most markets from a price feed at the deadline and gives 24 to 72 hours for the ones its team resolves, and then draws a distinction that is easy to be caught by: market resolution and on-chain redemption readiness are related but not identical, and a market can appear resolved in the API before the payout is fully settled on chain. A dashboard showing a resolved market is not a wallet showing the money.

And every one of these timetables has an escape clause for the case you will care about most. Kalshi Rule 7.2 lets the exchange adjust the expiration date and the timing of expiration where a circumstance would prevent the expiration value being determined accurately, with the last trading date adjusted to follow it; ForecastEx delays the resolution time until a delayed source agency publishes, posting one notice when it delays and a second when the contract resolves. A dispute is its own subject — who decides how a prediction market resolves covers the review processes, the fallbacks and what happens when the question has no clean answer. What belongs here is only the effect on the calendar: the branch where the capital is locked longest is the same branch where you are least sure of being paid.

What it costs

The arithmetic, with its units

The number a reader carries around is the gross one, and it is the least informative of the three worth computing. All of the following is arithmetic on stated inputs — it is a worked example, not a claim about any listed market.

Inputs. A Yes contract bought at $0.88, settling at $1.00 if the outcome is Yes and $0.00 if it is No. A taker fee charged as a coefficient of 0.07 on price times one minus price, which is the shape two of the US venues use; the conversions between the four incompatible units these fees are published in are in what a trade actually costs, and the point here is only that the fee is a number you add to the outlay. A holding period of 243 days — about eight months, roughly what a contract bought in the spring on an event in the autumn will cost you. No fee on the way out, because the position is held to resolution.

Step one, the gross return. $1.00 − $0.88 = $0.12 of gain on $0.88 of outlay, or 13.64%. This is the number that gets quoted, and it is a return per dollar committed, not per year.

Step two, net of the entry fee. 0.07 × 0.88 × 0.12 = $0.0074 per contract, so the real outlay is $0.8874 and the gain is $0.1126. That is 12.69% per dollar committed. The fee has taken about a fourteenth of the profit, which on a contract this far from the midpoint is the smallest of the numbers on this page.

Step three, per year. 1.1269 raised to the power of 365 ÷ 243, minus one, is 19.7% a year. Annualising a short hold flatters it, and this is the version most likely to be quoted back at you. Run the same 12.69% over different holds and the flattery reverses: 791 days is 5.67% a year, and 1,095 days — a contract bought three years before it resolves, which is an ordinary shape for a question about a future election — is 4.06% a year, which is inside the range a savings account or a short-dated Treasury has paid in recent years, for capital you could have spent at any time.

Step four, the one that changes the decision. Every figure above is conditional on being right. If the price is the market's probability, then at $0.88 the position pays 12.69% about 88% of the time and loses everything about 12% of the time. Weight them: (0.88 × 12.69%) + (0.12 × −100%) = −0.83% over the eight months. Not a little less than cash — negative, by roughly the fee. That is not a flaw in the venue; it is what a fairly priced contract does. The contract pays you for being better than the price, not for waiting. The waiting is the part cash does better.

So the honest comparison for a long-dated position is not 13.64% against a savings rate. It is: how much more likely than 88% do I think this is, and is that edge bigger than the fee plus what the 88 cents would have earned elsewhere over eight months? On $1,000 at 4%, eight months is about $26.46. That is the hurdle the forecast has to clear before it has done anything.

Who earns the interest on the collateral

Your collateral is not idle for the party holding it. It sits in a bank or an investment permitted by CFTC rules, and it earns the short-term rate. The only question is whose that is — and it is a rulebook question with three different answers, which is why it cannot be assumed from one venue to another.

Contractually yours, subject to an intermediary. ForecastEx Rule 612 says member funds are held in collateral accounts segregated under Part 22 of the CFTC regulations, that the clearing house may invest them subject to the limits in CFTC Regulations 39.15(e), 22.3(d) and 1.25, and then commits: the clearing house will pass all interest that it earns on monies in Collateral Accounts less any fees, belonging to Members or their Customers, to Members in the form of a monthly coupon payment. Its FAQ describes the same thing as passing 100% of the earnings back to members each month, proportionate to the closing value of their contracts, and notes it is earned whether the contract settles Yes or No. Read the next sentence of the rule before counting on it, though: an FCM member may pass on some or all of the coupon payment to their Customers. The obligation runs to the member, which for an individual is the broker, and how much of it reaches the person who posted the collateral is that broker's decision, not the exchange's. Rule 612(d) closes the loop — the clearing house retains all profits from investment of member funds not paid to members.

Discretionary, above a threshold the venue sets. Kalshi's exchange Rule 8.1, in the chapter headed Investment of Participant Funds, says participant funds may be invested by the clearing house, and that Klear may pay interest to Members' accounts at a floating rate to be determined by Clearing House on funds in Participant's accounts in excess of an amount to be determined by Clearing House. Three discretions in one sentence: whether, at what rate, and above what balance. Rule 8.1(c) states the default explicitly — the clearing house retains all profit from investment of participant funds not paid to members — and the Klear rulebook says the same again from the collateral side, that any interest earned on participant collateral may be retained by the settlement bank or the company. Whatever a venue advertises on top of that is a product decision it can change; the rulebook is what it has committed to, and what the rulebook has committed to is nothing. Check the current terms on the venue itself before treating a rate as part of the return.

Not provided for at all. The Polymarket US rulebook defines the fully collateralised position, the settlement bank and the settlement date, and contains no provision for paying interest on participant funds. That is an absence in one document rather than a claim about the product, and it is the right way to read it: where a rulebook creates no entitlement, there is none to rely on, and anything paid is a promotion.

And on chain, the question is a different one. A position on a non-custodial venue is collateral locked in a contract rather than a balance at a clearing house, so there is no operator earning a rate on it and none to pass on. The cost of the wait is the same — a stablecoin sitting in a conditional-token contract is a stablecoin not deployed anywhere else — but there is nobody to ask about it.

And where that leaves the sum

Add the four together for a long-dated position and the ranking is usually the reverse of the one people expect. The trading fee is the smallest term. The spread is larger, and is charged again if you leave early. The opportunity cost of the collateral over the months held is larger still. And the largest term of all is the probability-weighted one from step four, which is negative unless your forecast genuinely beats the price.

What you can do about it

Work out the annualised number before you enter, not the gross one. Three inputs: the price you pay including the fee, the payout, and the number of days to resolution. Divide, raise to the power of 365 over that number of days, subtract one. If the result is close to what cash pays, the trade is being asked to be right for very little, and the months are not free.

Then weight it by the price. The market's price is its probability. Multiply your net return by it, subtract the rest of the time, and you have the expected result of taking the trade at the posted price. If that number is negative — and at a fair price after fees it will be — the entire case for the position is your disagreement with the price. Write down what the disagreement is. If you cannot, the position is a way of holding cash for eight months, at a cost.

Find the resolution date before the entry price. On a regulated venue it is in the contract's terms and conditions, under the names the rulebook defines: Last Trading Date and Time, Expiration Date and Time, Settlement Date. ForecastEx publishes one terms document per product code; Polymarket US publishes contract terms read against the rulebook's definitions. Ask specifically whether the contract is subject to early resolution — the ForecastEx unemployment terms say in their instructions that that one is not, which means it runs the full distance even if the answer becomes obvious first.

Check the exit before you need it, on the contract you are actually buying. Look at the resting bid on your side of the book, at the size behind it, and at the spread, and price a full exit at those levels rather than at the mid. Do this on the long-dated contract specifically: depth on this month's question tells you very little about a question resolving next year. Where the liquidity comes from sets out what to look at on each kind of venue.

Ask the venue the interest question in its own vocabulary. Not "do you pay interest" — the useful questions are: is it paid on the cash balance, on the collateral behind open positions, or on both; what is the minimum balance; who is obliged to pay it, the exchange, the clearing house or my broker; and is the obligation in the rulebook or is it a programme that can be withdrawn. On ForecastEx there is a fifth: since the coupon is paid to the member and a member may pass on some or all of it, ask your futures commission merchant what share of it reaches you. A firm that cannot answer that in writing is answering it.

Prefer the shorter contract when the edge is the same. Two contracts on questions you are equally confident about, resolving four months apart, are not the same trade. The shorter one returns the collateral sooner, is exposed to fewer of the escape clauses above, and can be redeployed. The longer one has to pay for the extra months out of the same twelve cents.

Size for the branch where you cannot get out. The planning case is not the modal one where the market stays liquid and the source publishes on time. It is the one where there is no bid, the source is late, a review is opened, and the money is committed until the contract ends. That branch decides the size; the forecast decides the direction.

Re-read the rulebook, not the marketing page, when the rate matters. Everything above about who earns the interest comes from filed rulebooks and terms documents, read on 21 September 2026 — the Kalshi and Klear rules as filed with the CFTC on 2 May 2025, the ForecastEx rulebook dated 17 September 2026 and the Polymarket US rulebook dated 14 September 2026. Rulebooks are amended by filing and programmes are withdrawn by announcement, and neither arrives in your inbox. The cards in prediction market venues carry each venue's economics and the date this catalogue last read them.