A market on live player counts.

Concurrent players, CCU, is gaming's scoreboard. Launch-day peaks make mainstream news, chart races between rival releases run for weeks on social media, and "dead game" is a taunt with an integer attached. The metric is public, high-frequency, emotionally load-bearing for a very large audience, and, until now, not tradeable anywhere.

The feed itself has properties that shape everything in this document. Steam's endpoints serve a cached snapshot on a roughly ten-minute cycle, and the ranked charts endpoint self-reports each snapshot's generation time. Different network edges roll over at different moments, which means a sophisticated poller can observe a new value seconds to minutes before a casual reader. Every third-party chart site is downstream of the same single Valve counter, so no independent second source exists. And because idle logged-in clients count as players, the metric is cheaply manipulable for free-to-play titles and expensively manipulable for paid ones.

Respawn does not pretend these properties away. The market structure is built from them.

Games cannot issue stock, so Respawn trades them in sessions on the index price: the game's surprise, how far it beats or misses its own tape, the median of its count at the same weekday and time over the trailing four weeks, measured in the game's own volatility unit. A game one unit above its tape prices about ten percent higher than one exactly on it; a game that climbs exactly as the clock says it will does not move. A session locks before the first measurement and closes at the game's own peak hour; a long pays the change in the index price from the opening mark to the closing mark, a short the reverse, and longs and shorts are matched against each other at lock with a seeder pool on the lighter side of thin books. At the close mark the position settles back to its holder's vault, capped at 50 percent, and its holder re-enters the next session if they want. A session runs to its close: there is no early close on an on-chain session, because a verified mark is the only price the system will ever act on.

Settlement function

The seeder pool backs thin books. At lock, longs and shorts are matched against each other pro rata; the seeder takes the lighter side of an uneven book, bounded by a published cap per session and a published share of its own balance and never beyond the actual imbalance, and any excess on the heavy side is returned. It is an account inside the vault with no key and no view: it never picks a side, it is on whichever side has less by the same rule every session, and it is funded from the fees the exchange earns, recycled into liquidity without leaving the vault. Its position is an ordinary position, settled on the same marks, paying the same fee, voided with everyone else when a feed degrades, and losers settle first. A degraded window voids the session by rule: no value moves, the fee is waived, and every position returns to its vault unchanged. The exact formulas are published and enforced in the contract.

The pool model is built and locked. Under it a position opens instantly at the verified index plus a spread and closes at the index less it; a shared pool takes the net of longs and shorts on every game as counterparty, never a side of its own choosing, inside two caps derived from its own usable equity and an absolute per-game ceiling sized below the cost of buying fake players; the spread widens as a game's cap fills and hourly funding flows from the heavy side to the light side, so an imbalance is priced rather than refused; and if the pool ever owes more than its equity, gains are paid pro rata at a published ratio and losses are unaffected. Perpetual trading turns on as soon as community-token creator fees are enough to seed the liquidity pool; until then every game trades in sessions. The Depth season is the latest it happens. Sessions retire at that point, every open position carries across at its last mark, and depositor access to the pool, under the same shares, a withdrawal delay, and an exposure guard, follows.

Why the surprise, not the count

Every game on Steam rises and falls with the same daily and weekly wave. A position priced on the raw count is decided by the clock and shared by every market: on one day of live observations across 82 listed games, the afternoon window rose on 94 percent of games at once and the evening window fell on 88 percent, while the full-day median move was 3 percent. Priced on the surprise, every window split within a point of fifty-fifty with real dispersion, and per-game volatility spanned an order of magnitude, which is why payouts are measured in each game's own unit. The tape and the volatility band for every game are signed into every mark and are replayable from committed observations.

Pairs

A pair is long one game against another. At launch it is two positions in two books, one long and one short, locked and marked in the same sessions, so the shared wave cancels and only the disagreement between the two games pays. The single-ticket pair, settling on the ratio of the two surprises in a combined volatility unit, follows after launch.

Settlement inputs come exclusively from Steam's public endpoints, with the ranked charts endpoint as the canonical source because it alone self-reports snapshot generation time. The relay works as follows:

1. Independent reporter instances fetch the canonical snapshot.
2. Each reporter signs a typed message binding the snapshot hash, source timestamp, fetch timestamp, game, and value.
3. The settlement contract accepts an observation only at reporter quorum with valid timestamps, and commits it on-chain immediately.
4. Marks are medians of accepted observations across published windows, keyed to the source's own timestamps, never to wall-clock fetch time.
5. Every snapshot is committed on-chain whole at the source's own timestamp: every listed game's count as calldata under a root a quorum of reporters signed, one transaction per snapshot however many games are listed. When a session's window ends, at lock and again at the game's peak hour, each reporter computes the thirty-minute median from its own signed observations under the published rules and signs the window mark with a hash of the exact observations it used; the session contract accepts a mark only when a quorum of distinct reporters signed the identical mark and evidence. A reporter that saw a different feed does not match and does not count. The accepted mark opens the session, or closes it and settles every position to its vault at that mark. The index feed the pool model uses, a quorum-signed mark every five minutes priced on chain and bounded to a published step per hour, is built and deployed alongside; it prices nothing until perpetual trading turns on.

The registry enforces the rules in code: a quorum of distinct reporters (five of seven at launch), a source timestamp no older than twenty minutes at fetch, reporter clocks within five minutes of the chain, strictly increasing source timestamps per game, at least three samples per window, and a stale threshold of fifteen minutes measured from the last observation before the window, so a feed that died just before a window opened degrades it. The same rules run in every reporter and in the contract, so a mark is only ever what the rules produce.

Medians over windows are chosen deliberately: a single snapshot can be mistimed or manipulated, but shifting a median requires sustaining a false value for the majority of samples, in public, on charts the entire internet watches.

Degraded feeds stop the clock, they never improvise. If the source timestamp goes stale beyond the published threshold, the window is degraded and the session voids by rule: no value moves, the fee is waived, and every position returns to its vault unchanged, under objective rules enforced in code. There are no discretionary reversals, no post-hoc clarifications, and no governance vote over outcomes, the three failure modes that have damaged comparable venues.

Anyone can replay any mark: published windows, on-chain snapshots, the evidence hash on-chain, public chart history. Disagreement would be provable.

The central threat in a player-count market is not block-level extraction; it is minutes-scale information asymmetry against a cached public feed. Respawn's answer is structural rather than procedural:

Respawn's revenue is a toll on trading, never a book against users: a fee of 5 percent of the value moved at each session close, paid by the side that gained, plus an entry take of about 10 percent on competition pools. Nothing is charged when nothing moves, on a void, or on a deposit. The venue's income is identical whichever side of any market wins, which removes every incentive that makes discretionary venues dangerous. At launch the seeder pool's share of every fee, 100 percent, is swept to its vault balance and becomes liquidity at the next lock; the share is a published parameter, dialled down through the timelock as organic depth arrives. When perpetual trading turns on, the toll becomes 10 basis points of the position at open and again at close with the pool's share credited to pool equity, plus a published spread and hourly funding. Deposits and withdrawals carry no Respawn fee.

Comparable products built on asset churn have a uniform record of collapse once churn slows; a small toll on the value each session moves, and later a small toll at each open and close with a spread that widens only with imbalance, are the surviving model. The fee engine above is the entire business.

The community token stands apart from the exchange. The community token launches with the platform. It has no utility on Respawn, no rights, and no claims: it is not equity, carries no share of fees or revenue, has no governance power, and plays no role in trading. Its market value is independent of the exchange and may go to zero. Launch mechanics are subject to counsel review, and the only official announcements come from respawn.family; any contract address not announced there is fake. The exchange's own economics remain exactly as described above: a small toll at each open and close, requiring nothing from users except that markets keep being interesting. Creator fees earned from community token trading flow to the exchange and seed the liquidity pool; that flow runs from the token to the exchange, never the other way, and gives holders no claim on the pool or on Respawn.

Traders sign in with X. Authentication creates an embedded self-custodial wallet on Robinhood Chain through Privy; keys are generated and used inside secure infrastructure such that neither Respawn nor Privy ever holds them whole, and export is always available. Respawn stores public identifiers only: the stable X subject, username, avatar, and wallet address. Before balances pass a policy threshold, a recovery method is required so no X account event can strand funds. Traders who prefer their own wallet connect it instead.

Trading is live ETH from the first session. A trader deposits ETH from their own wallet into the Respawn session vault on Robinhood Chain, a position reserves ETH from that vault when it opens, and the position settles back to the same vault at the session close. Nothing on Respawn is simulated and no balance is issued by the venue: every position is the trader's own ETH, and withdrawal to the trader's own wallet is always available after the published delay.

Positions are public by design. Every position joins the open record attached to its X identity when it opens. The open book is the social product; the lock before the first measurement is what makes publicity safe, because nothing published at lock can be traded against.

Respawn deploys on Robinhood Chain, an EVM environment, chain id 4663, with development on testnet and launch on mainnet only after independent contract review. The contract system is modular:

Administrative controls sit behind a multisignature wallet with a time delay, and their scope is published. The venue cannot move user funds, cannot freeze wallets, and cannot reverse a close; guardians can pause, and every unpause and rule change waits out the timelock.

1. Source dependency. All player-count data originates from Valve, whose API terms permit revocation at any time without notice. Feed loss voids the session under published rules: no value moves and positions return to their vault unchanged. Respawn is not affiliated with Valve Corporation or Steam.
2. Market risk. Sessions settle against real movement; participants can lose up to the published cap in a session, and the whole of a position over repeated sessions. Entries lock before the first measurement and a session runs to its close: an on-chain session carries no early close, so ETH reserved for a position is committed until it settles.
3. Seeder and pool solvency. At launch the seeder pool takes only the lighter side of thin books up to published caps; unmatched exposure on the heavy side is returned at lock. When perpetual trading turns on, the pool is the counterparty for net exposure, capped as a share of its own equity; if it ever owes more than it holds, gains are paid pro rata at a published ratio until it recovers, and losses are unaffected. Depositors to the pool bear its losses and wait out a withdrawal delay and an exposure guard.
4. Smart contract risk. Audits reduce risk; they do not eliminate it.
5. Regulatory risk. Real-money markets on measurable outcomes carry jurisdiction-dependent regulatory surface. Respawn treats legal review as a launch gate.
6. The token is not the exchange. community token, launching with the platform, has no utility, rights, or claims on Respawn; its value is speculative, independent of the exchange, and may go to zero.
7. No ownership. A position is a two-sided contract on a public metric, settled session by session. It is not equity, shares, or any interest in any game or company, and confers no rights over any game.

The charts are already a spectator sport. Respawn sells the seats.