Participants · 11 / 19Preview · not deployed
For market makers
What a seat is worth, with an expected-value walkthrough of a bid.
A seat is a short, exclusive trading franchise on one pool: trade first, trade fee-free, set and collect the fee, and during DAY keep the price honest. It is sold by auction before every session and paid for in full, upfront. This page is for the firms and individuals who might bid: what a seat is worth, how to reason about a bid, and what it costs to get it wrong.
Who should bid#
A seat suits someone who already does three things: prices $NVDA or another underlying continuously, including pre-market; hedges on a reference venue at low cost; and runs on-chain execution that does not fall over. Arbitrage desks, tokenized-equity market makers and dominant LPs fit that description. A dominant LP bidding on its own pool mostly pays itself, which is neutral, and is covered on Edge cases.
It does not suit someone looking for passive income. A seat is a cost first and a revenue only if you work it.
What a seat is worth#
seat value = expected gap capture during the bell window
+ expected fee-free arbitrage during the session
+ expected swap fees collected from other traders
- expected cost of duties (rebalancing work + slashing risk)
- hedging, gas and inventory costs
Which terms matter depends on the session. An OPEN seat is mostly the first line. A DAY seat has no bell window unless there is a halt, earns from the second and third lines, and carries the whole fourth line. NIGHT and WEEKEND seats have no duties and the widest fee caps, and their value is mostly fee income plus fee-free repositioning as news arrives. The rights are detailed on Seat rights, the duties on Seat duties and slashing.
Expected value of an OPEN seat#
Take an OPEN seat on a $4M $NVDA/USDG pool with liquidity concentrated at ±20%. For the first 20 seconds only the regent can swap, so the regent captures the whole gap between the pool price and the opening price. "Gap" here means pool price at 09:30 ET versus the opening price, not close-to-open: if the NIGHT session already moved the pool toward the pre-market price, the remaining gap is smaller.
Gap capture per scenario is computed by the same code the simulator uses. Down gaps of the same size give nearly the same figure.
| Scenario (illustrative) | Probability | Gap | Gap capture | Weighted |
|---|---|---|---|---|
| Quiet | 50% | 0.5% | $129 | $65 |
| Normal | 30% | 1% | $516 | $155 |
| Active | 14% | 2% | $2,055 | $288 |
| Large | 5% | 3% | $4,602 | $230 |
| Earnings | 1% | 6% | $18,141 | $181 |
| Expected gap capture | $919 |
Gap capture grows roughly with the square of the gap, so the tail dominates: the two rarest rows carry a large share of the expected value. Then add the rest, all illustrative:
| Line (illustrative) | Amount |
|---|---|
| Expected gap capture | $919 |
| Swap fees: $400,000 of third-party volume at a 0.25% fee | $1,000 |
| Fee-free arbitrage in the rest of the session | $200 |
| Hedging, gas and inventory costs | -$150 |
| Cost of duties (none: duties apply in DAY only) | $0 |
| Expected value of the seat | $1,969 |
Gap capture is gross. It assumes you can hedge the other leg at the opening price, which is a job in itself.
From expected value to a maximum bid#
The bid is certain and the value is not. A rational bidder demands a margin for variance, capital and model error. With an illustrative margin of 20%:
| Amount | |
|---|---|
| Expected value of the seat | $1,969 |
| Rational maximum bid (illustrative margin) | $1,575 |
| Result on a quiet morning at that bid | -$396 |
| Result on an earnings morning at that bid | $17,616 |
At this bid the regent loses money on the most common morning and earns it back on rare ones. A bidder who cannot carry that pattern should not bid.
The table is what you use the evening before. The auction hard-closes 2 min before the session, by which time pre-market trading has revealed most of the gap. Late bids should be conditioned on it. Everyone else sees the same pre-market price, so on a big-gap morning expect the auction to price the gap, not to miss it.
How the rules shape bidding#
The raise rule. Each bid must beat the standing bid by at least +5%. With a maximum of $1,575, you can only answer a standing bid up to $1,500. Above that the required raise exceeds your limit and you are out, even though the standing bid is below your value. Small creeping raises can leave you on the wrong side of that step; a jump bid close to your maximum avoids it. In the preview model the standing top bidder cannot raise himself.
The anti-snipe rule. A bid in the last 60 s extends the auction by 60 s. Waiting for the final second buys nothing, because rivals get time to answer. Extensions stop at the hard close, 2 min before the session; after a run of extensions the last close is firm. Have your real maximum in before then. In the preview the scheduled close is 10 min before the session, which is a preview assumption.
The reserve. The opening bid must meet the reserve: 50% of the median of the last 10 winning bids for that session type. It is public. If your value is below it, do not bid; the seat goes unsold and the pool runs in fail-safe mode. If you are the only bidder, you win at your opening bid. Full rules are on The seat auction.
The winner's curse#
A seat has roughly the same value to every competent bidder. In such an auction the winner is whoever held the most optimistic estimate, and the most optimistic estimate is usually too high. If you win often and easily, check your model before you celebrate. Shade your bid, and shade it more when you know less than rivals about the underlying, the hedge or the pool's flow.
No bond, no bid#
Bidders must lock a $RGNT bond sized by pool tier before they can bid. A bid from an address without a bond is rejected. The bond is what gets slashed for DAY breaches, at a rate per breach-second, split 50% to LPs, 10% to the poke() caller and 40% burned.
| Tier | Pool TVL | Bond |
|---|---|---|
| 1 | $5M and above | 25,000 RGNT |
| 2 | $1M and above | 10,000 RGNT |
| 3 | Below $1M | 2,500 RGNT |
See The $RGNT token.
Operational checklist#
- Registered executor. Fee-free swaps and the bell window only work through the executor contract registered for the session. Make sure no third party can route through it.
- Bell window automation. 20 seconds at ~100 ms blocks is ample for a machine and nothing for a human. Halt resumptions open a new bell window without notice.
- Oracle monitoring during DAY. Track the Chainlink feed and your deviation against ±max(current fee, 0.30%). You have 60 s of grace. Your competitors are paid to call poke().
- Fee policy per session. Bounds run from 0.05% to 0.50% in DAY, 1.00% in OPEN, 2.00% in NIGHT and 2.00% in WEEKEND. A higher DAY fee widens your band and removes outside help in holding it.
- Halt handling. Duties are suspended, the fee cap is raised, and a new bell window opens on resumption.
- Calendar. Early closes end DAY at 13:00 ET. See Sessions and the calendar.
The downside#
A bid is sunk. It is paid in full and escrowed before the session starts, and it streams to LPs whether you trade or not. There is no refund for a quiet morning, a failed hedge, a bug in your executor or a change of mind. The one exception is sequencer downtime: rent that could not stream while the chain was down is refunded pro rata.
On top of the bid, a DAY regent risks the bond. And everything here runs on contracts that do not exist yet and, once deployed, would carry risk of total loss and require an independent audit before any real funds. You can rehearse bidding with paper balances in the preview. See Risks and disclaimers.