Micro auctions, and why they make an exchange fair

The problem it solves

Almost every exchange in the world matches orders continuously and in time order. An order arriving a microsecond earlier is filled first. That single design choice creates the entire high-frequency trading industry, because when being first is worth money, it is rational to spend enormous sums on being first.

The costs are well documented. Firms have built microwave tower networks and leased satellite links to shave milliseconds off transmission. In crypto the equivalents are fast relays, specialised networking and priority gas bidding. None of it produces a better price for anyone. It is a pure transfer from slower participants to faster ones, and it consumes real resources to achieve nothing.

• Latency arbitrage — traders with a speed advantage extract value from slower ones without offering better prices.
• Adverse selection — a market maker quoting tightly gets picked off the moment prices move, before they can cancel. They defend by widening spreads, which everyone then pays.
• The arms race — continuous spending on speed that improves execution for nobody.

How a micro auction works

A micro auction replaces continuous matching with a rapid sequence of batch auctions. On Rocket, each block is an independent auction lasting roughly 100 milliseconds. Within a block the sequence is:

• Traders submit limit orders and cancellations during the block.
• At block close, aggregate supply and demand curves are constructed from everything received.
• A single uniform clearing price is chosen — the price that maximises the volume able to trade.
• All eligible orders execute at that same price, with pro-rata allocation at the margin.
• Residual orders either roll forward into the next auction or expire.

The property that changes everything

There is no time priority within a block. Whether your order arrived first or last in the interval is completely irrelevant — every order in the batch is treated as simultaneous. The only dimensions that determine your outcome are the price and the size you were willing to show.

Once arrival time stops mattering, spending money to arrive sooner stops making sense. The arms race has nothing to compete over. Latency no longer pays; price does.

To a human it still feels continuous. A hundred milliseconds is far shorter than reaction time, so placing an order and seeing a fill is indistinguishable from trading on a conventional book. The experience is unchanged while the incentives underneath are inverted.

Two worked examples

Price improvement. Alice is willing to buy at up to 100.10 and Bob is willing to sell at 100.00 or above. On a continuous book, Alice's order executes at 100.10 and Bob later sells at 100.00, with the spread accruing to whoever sat in between. In an auction, both orders land in the same batch and clear at roughly 100.05. Alice pays less than her maximum, Bob receives more than his minimum, and the surplus stays with the two of them.

Adverse selection damped. A market maker has posted an ask at 100.10 when fair value suddenly jumps to 101.00. On a continuous book, the fastest taker lifts that 100.10 offer instantly and the maker sells a full point below fair value. In an auction, the new buying interest joins the same batch and pushes the clearing price to around 100.90. The maker still sells slightly below fair value, but the loss is cushioned rather than total — so they remain willing to quote tightly next time.

Why front-running and MEV disappear

Front-running requires a first transaction to get ahead of. Sandwich attacks require ordering a victim's trade between two of your own. Both depend entirely on the existence of a sequence within the matching process.

In an auction there is no sequence. Every order in the block clears together at one price, so there is no position in a queue to buy, bribe or race for. The extractive strategies that dominate onchain trading do not become harder — they lose the structure they operate on. Value that would have leaked to searchers and reordering bots stays with the traders.

What it produces

• Tighter spreads, because makers no longer pad quotes against being sniped.
• Deeper books, because pro-rata allocation makes displaying genuine size rational rather than dangerous.
• Better fills on large orders, which clear at one price instead of walking a flickering book.
• Fairer prices, because overlapping buy and sell interest clears inside the spread rather than each side paying it.
• Execution at your limit price or better, never worse.

Fairness is more than the matching engine

Auction matching removes the speed advantage, but an exchange has other places to quietly sort its users into classes. Rocket closes those too, and the claim to be the fairest venue rests on the whole set rather than the matching engine alone.

• Flat fees at every size. One maker and taker rate for everyone, rather than a volume ladder that charges a first-time trader several times what the largest desk pays for an identical trade.
• Funding paid between traders. On perpetuals, funding is exchanged peer-to-peer between longs and shorts with no protocol cut, so the mechanism that keeps the contract honest is not also a revenue line.
• Self-custody throughout. You trade from your own wallet; the protocol margins and settles onchain rather than holding your balance in an operator account.
• Verifiable settlement. Because trades settle onchain, anyone can check them on the Rocketberg explorer rather than trusting what the operator publishes.
• One price per block, published. There is no hidden internalisation, no payment for order flow, and no privileged routing tier.

Where it applies on Rocket

Everything matched on Rocket clears through the same auction: the BTC and ETH options chains, and perpetual futures across crypto, tokenised equities and commodities. There is no premium tier with different execution and no separate venue for larger participants.

Options benefit the most. An options market maker quoting a ladder of strikes and expiries is exposed at every one of them, and on a continuous book a fast taker picks off whichever quotes went stale when spot moved. Makers defend by widening spreads and showing less size, and traders pay that padding on every contract. Auction matching cushions the stale quote instead of handing it over, which is why the effect on an options chain is more visible than on a liquid perpetual.

The Rocket Liquidity Provider vault sits on the same mechanism. It provides liquidity, runs market making strategies and absorbs liquidations — activities whose economics depend directly on not being systematically picked off.

This is not an experiment

Auctions are among the most widely trusted mechanisms in finance and beyond. Google allocates billions of ad impressions daily through a continuous sequence of auctions, where advertisers compete on price rather than speed. The U.S. Treasury issues sovereign debt through uniform-price auctions. Major stock exchanges already run call auctions at the open and the close — precisely the moments when liquidity is deepest and accurate price discovery matters most.

The pattern is consistent: wherever price is the variable that matters most, auctions emerge as the superior mechanism. For the overwhelming majority of traders, the difference between ten and a hundred milliseconds is irrelevant, while the price they trade at matters enormously. Running auction windows short enough to feel instant makes price, rather than speed, the axis of competition.

Blockchains already reach distributed consensus on each block. Auction matching extends the same principle to reach consensus on price.