Batch Auction

A batch auction is a trading method that collects buy and sell orders over a period of time and executes all of them together, at one uniform clearing price, at the end of that period. It contrasts with continuous trading, in which each incoming order is matched immediately against the best opposing order in the order book. Batch auctions are also called call auctions or uniform-price double auctions.

Stock exchanges have long used batch auctions to open and close the trading day. Economists have proposed running them every fraction of a second to blunt the speed race in high-frequency trading (HFT), and in crypto they are used by DeFi protocols to protect traders from front-running and to run fairer token sales. The batch auction is a cousin of the Dutch auction, which also settles many buyers at one price.

Picture a farmers' market where, instead of haggling stall by stall all day, everyone writes down what they want to buy or sell and at what price, and drops the slips in a box. Every hour the organiser opens the box, finds the one price at which the most apples change hands, and everyone who agreed to that price trades at it.

Nobody gains by rushing to the stall a few seconds earlier, because all slips in the box are treated the same. Being first no longer matters; offering a good price does. That is the core idea of a batch auction.

At the end of each batch, the auction builds a demand curve from all buy orders and a supply curve from all sell orders, then picks the price that maximises the quantity traded:

• $p^{*}$ is the clearing price paid and received by every matched order.
• $D(p)$ is the total quantity buyers want at price $p$ or better (buy orders with a limit at or above $p$).
• $S(p)$ is the total quantity sellers offer at price $p$ or better (sell orders with a limit at or below $p$).

Worked example: buyers submit limit orders for 100 units at $10.20, 200 at $10.10 and 300 at $10.00; sellers offer 150 at $9.90, 100 at $10.00, 50 at $10.10 and 300 at $10.20. At $10.10, demand is 300 and supply is 300, so 300 units trade. At $10.00 demand rises to 600 but supply falls to 250; at $10.20 supply is 600 but demand falls to 100. The clearing price is $10.10, and every matched buyer and seller trades at it, even those whose limits were better. Exchanges add tie-break rules when several prices give the same volume.

In "The High-Frequency Trading Arms Race: Frequent Batch Auctions as a Market Design Response", published in the Quarterly Journal of Economics in 2015, Eric Budish, Peter Cramton and John Shim argued that continuous trading creates mechanical arbitrage opportunities that go to whoever is fastest. Using millisecond exchange data, they found that correlations between closely related instruments break down at very short horizons, so firms invest heavily in speed to capture stale quotes. They proposed replacing continuous trading with uniform-price batch auctions held at short fixed intervals, such as every tenth of a second, so that competition shifts from speed to price. The proposal has been widely debated by regulators and exchanges but has not replaced continuous trading in major U.S. equity markets. Many venues do use opening and closing auctions, and these concentrate a large share of daily trading volume.

On public blockchains, pending transactions are visible before they are confirmed, so bots can reorder them to profit at users' expense, for example through sandwich attacks on automated market makers (AMMs). Batch auctions remove the advantage of ordering:

• CoW Protocol. Users sign trade intents off-chain; third parties called solvers compete to settle each batch, and the protocol's documentation states that the same asset pair clears at a uniform price within a batch, making the order of trades irrelevant. Where two users want opposite trades, the protocol can match them directly (a "coincidence of wants") without touching on-chain liquidity, and the winning solver must deliver at least each user's signed price.
• Gnosis Auction. Launched in 2021, this platform runs batch auctions for token offerings: the seller sets a minimum price and bidders submit a maximum price and quantity, and all winning bidders pay the same clearing price. Its stated aims were fairer price discovery and resistance to front-running bots, compared with first-come sales that reward a sniper bot.

• Fairness. All participants in a batch get the same price, reducing the value of speed and of seeing orders first.
• Liquidity pooling. Orders are concentrated at one moment, which can deepen liquidity and narrow slippage for large trades.
• Delay. Traders must wait for the batch to close, which suits investors less well when prices are moving fast.
• Complexity. Running auctions across many assets at once is computationally harder, which is why DeFi versions rely on competing solvers.

Batch auctions are a question of market design rather than a trading strategy, an alternative to a continuous order book quoted by a market maker. They are studied alongside dark pools and algorithmic trading as ways to manage the costs of fast, fragmented markets.

• CoW Protocol documentation: Batch auctions
• Budish, Cramton and Shim: The High-Frequency Trading Arms Race (NBER working paper)
• Gnosis Auction smart contracts (GitHub)