How Much Screening Buys a Second? Speed, Selection, and the Maker's Spread

In the invisible markets where digital assets are traded, liquidity providers act as the steady hands that keep prices moving smoothly. They stand ready to buy and sell at any moment, offering a two-sided price to anyone who wants to trade. In exchange for this service, they charge a small fee built into the price difference between buying and selling, known as the spread. This spread is their insurance; it must be wide enough to cover the losses they suffer when they trade with someone who knows more than they do. If a trader knows the price is about to jump, they will trade against the provider before the provider can update their price, pocketing the difference. The provider then relies on the fees collected from less-informed traders to pay back those losses. This balance is the core of how these markets function, but it becomes fragile when the speed of information changes.

The challenge intensifies in the world of automated market makers, which are computer programs that set prices based on the amount of money sitting in a pool. Unlike human traders or traditional exchanges that can update prices instantly, these programs often only refresh their prices when a new trade happens. On the Ethereum network, for instance, a new block of transactions is created roughly every twelve seconds. If the price of an asset changes on a faster, centralized exchange during those twelve seconds, the automated pool is left quoting a stale price. A trader watching the faster exchange can see the new price, wait for the automated pool to still be quoting the old one, and then execute a trade that guarantees a profit at the pool's expense. The pool is essentially paying for a price the rest of the world has already left behind.

Researchers Aleksei Adadurov and his colleagues set out to measure exactly how much this delay costs and to find a way to quantify the trade-off between speed and caution. They asked a fundamental question: if a market maker cannot update their price as fast as the competition, how much better must they be at spotting and rejecting trades from informed opponents to survive? The team built a model to calculate the precise rate at which a slower provider must screen out informed traders to break even with a faster, unscreened provider. Their work reveals that the shape of price movements matters more than just their average size. They found that price changes are not smooth and predictable like a bell curve; instead, they are often silent for long stretches and then jump suddenly. This jagged reality means that simply being faster is not the only solution, and that the cost of being slow is higher than traditional models suggest.

To test their theory, the researchers examined four major pools trading a digital dollar against Ethereum, comparing their performance against a fast, centralized exchange over three months in 2026. They discovered that the price of Ethereum does not move every second. In fact, 73.4 percent of one-second subslots see no change in the mid-price. When the price does move, it often jumps significantly, creating a distribution that is far more extreme than standard financial models predict. Because the price stays still so often, the risk of a sudden, large jump accumulates quickly over time. When the researchers compared a twelve-second update cycle to a one-second cycle, they found that cutting the wait time from twelve seconds down to one eliminated 84.5 percent of the expected price shock. This means that a provider stuck with a twelve-second delay faces a massive disadvantage.

The study calculated that to match the performance of a provider who updates every second, a provider who updates every twelve seconds would need to screen out at least 84.5 percent of the informed traders trying to exploit their stale prices in the low-toxicity limit. This is a very high bar. Traditional models, which assume price changes follow a smooth, bell-shaped curve, would suggest that a screening rate of only 71.1 percent is sufficient. The researchers showed that this standard model is dangerously optimistic because it misses the reality of how prices actually move. The "fat tails" of the real data—those sudden, large jumps—mean that the cost of being slow is much steeper than previously thought. If a provider cannot screen out that many informed traders, they must find a way to update their prices faster, or they will lose money.

The researchers also looked at what happens when the market is trending, moving steadily in one direction. In these moments, the stale pool is picked off more aggressively on the side of the trend. Trades that go with the trend earn a significantly better effective price for the trader than trades that go against it, creating a spread difference of 11.1 basis points in favor of the trend. This confirms that the delay in updating prices is not just a theoretical risk but a real, measurable cost that shifts wealth from the pool to the informed trader. The study concludes that the ability to screen flow is a powerful defense, but it is not a perfect substitute for speed. When the market moves in a way that is hard to predict and prone to sudden jumps, the only way to fully neutralize the risk of staleness is to reduce the time the price is exposed.

Ultimately, this research provides a clear metric for the value of speed in digital markets. It shows that a single second of reduced delay buys back a specific amount of protection against informed traders. For a market maker, the choice is no longer just about how fast they can update, but about how well they can identify and reject the trades that will hurt them most. The findings suggest that in a world where prices can jump unexpectedly, the cost of waiting is high, and the only way to match a faster competitor is to be exceptionally good at filtering out the traders who know the price is about to change.