Modern Day Latency Arbitrage: Predicting Price Changes
Over the past few weeks, IEX® upgraded the IEX Signal, a model developed by IEX and designed to predict when and how a stock price is about to change, and expanded the IEX Signal’s protective benefits to our Primary Peg order type. Including the IEX Discretionary Peg (D-Peg®), two out of five IEX order types are now armed with this breakthrough innovation.
Stock exchanges update functionality and offerings all the time, for any number of reasons. But unlike other exchanges, IEX is built with the express purpose of protecting investors. We focus our innovation on those investors, as well as IEX Members, and evaluate every decision based on how it will affect them.
Today’s announcement is no different.
The NBBO Today
The idea behind the National Best Bid and Offer (NBBO) is that even though there are twelve active U.S. stock exchanges, there is one “best price” for a stock across the market.
The problem with this concept is that there is no way to instantaneously update every market participant — including the exchanges themselves — when the NBBO changes. Data about price changes must literally travel the distance between participants, and that trip can be faster or slower depending on the technology a firm is using and how far a participant is from the source of the data.
While connections are faster than they have ever been, not every participant in the market can compete at the bleeding edge of speed. Some firms that handle trading on behalf of customers don’t have the leeway to use some of the fastest technologies, such as microwaves and lasers, because they are less stable and reliable, and firms that trade for customers have regulatory obligations that proprietary trading firms do not. Some are mired in legacy systems that just can’t achieve the internal processing times of the fastest participants, or have other business priorities that preclude making the investments — including huge sums of money for the fastest data and connectivity — they would need to compete in the speed race.
Regardless of the reason, these speed differentials mean that market participants — again, including the exchanges themselves — have different views of the NBBO at the exact same moment in time.
In other words, at any given moment in time, there is no universal agreement of prices. This is especially true when you’re talking about sub-second increments of time, which is the speed of today’s markets.
Traditional Latency Arbitrage
The discrepancy among different views of the NBBO creates actionable windows of time for faster traders to use their information advantage to trade with orders at stale prices, earning a low-risk profit at the expense of slower investors. Because exchanges often process market data more slowly than the fastest traders, those traders can execute at stale prices on exchanges that haven’t yet processed the change in NBBO.
Because these trades are based on knowledge of how the price of a stock is about to move, they are reliably — and nearly instantly — profitable. If you are talking about a buy order, you know that right after the trade, the price will move higher. It’s like buying a watch when you know the value of that watch has gone up in the broader market — but the store doesn’t know yet, so hasn’t updated its price. You are still taking on some risk by buying the watch, but it’s a very good bet. So while it’s not a “true arbitrage” strategy in which the position is closed out immediately, this form of speed-based trading amounts to de facto latency arbitrage.
It isn’t news that the IEX Speed Bump is designed to stop this kind of latency arbitrage by delaying orders by 350 microseconds to give IEX time to update its NBBO, and any orders pegged to the NBBO, before trades can be executed. It’s a simple and effective hardware solution that stops fast algorithmic traders from picking off orders on an exchange that is operating, and pegging, based on an older version of the NBBO.
Modern Day Latency Arbitrage
In a race where the tiniest speed advantage can translate to profit, market participants that trade based on speed are not running in place, content to cede their advantage to the IEX Speed Bump.
If an advantage can’t be gained by reacting faster to a change, an obvious strategy is to know about the changes earlier — to form your view of the new NBBO sooner. And the way to do that is to predict changes to the NBBO — essentially by observing the dominos of orders at the “best price” begin to fall before the final tile is toppled.
This type of prediction is possible because in many cases, rather than changing instantly, quotes transition between prices in a gradual way, over the course of microseconds (millionths of a second) and up to a few milliseconds (thousandths of a second). Orders quoting the best price are spread across a number of exchanges, and as counterparties arrive in waves for those quotes, a careful observer can see those dominos begin to fall. This view allows the observer to predict, with a fairly high degree of certainty, that a price is about to change. We call this kind of price change a “crumbling quote.”
Predicting a price change by observing a crumbling quote amounts to probabilistically building a future view of the NBBO. Some fast traders can then use this view of the future NBBO to trade before the price change completes and is communicated to any participants.
This new form of de facto, speed-based arbitrage — what we call “modern day latency arbitrage” — is opportunistic, reactive, and based on information only available to some participants, not fundamental analysis of a stock. Like the latency arbitrage described above, the position isn’t closed out immediately, but trades are nearly instantly profitable on average. Even if every prediction isn’t right, its accuracy is good enough to make this a very profitable strategy.
And the profits don’t come from thin air — they are extracted from the markets at the expense of slower participants, who are often large institutional investors that represent the savings of individuals through pension funds, mutual funds, and 401ks. Even some of the most sophisticated electronic market makers are exploited by these opportunistic traders who can target the most advantageous, predictable price changes, while market makers are managing massive numbers of bids and offers across hundreds to thousands of symbols.
Fighting Math with Math
The IEX Speed Bump alone isn’t enough to protect those orders against modern day latency arbitrage. Because the predictions can happen up to milliseconds before the last quote at a certain price is exhausted, stopping this form of arbitrage with our Speed Bump, which measures 350 microseconds, would require too long of a delay to be practical.
Instead, IEX decided to develop its own ability to predict the “future NBBO,” in an attempt to nullify the advantage of traders trying to conduct this form of arbitrage. Combined with the IEX Speed Bump, IEX only needs to make a prediction within 350 microseconds of another trader to be able to protect orders.
The result — the IEX Signal — is a predictive model designed to predict declines in the National Best Bid (NBB) or increases in the National Best Offer (NBO) and protects resting orders based on that view of the “future NBBO.” It’s like a yellow traffic signal that warns you when the price is about to change. Integrated into D-Peg and Primary Peg, it stops orders from trading aggressively while the Signal is on — like stopping your car from running a dangerous yellow light.
This application of the Signal is not designed to impede regular trading — it is on for little more than five seconds per symbol per day on a volume weighted average basis. However, predatory traders are extremely active during these short periods, seeking aggressively to trade when they have an information advantage over slower traders. In fact, approximately 32% of all marketable orders sent to IEX fall within this window, which represents only about 0.02% of the time in a trading day! In other words, “crumbling quotes” create a highly-sought trading opportunity. Protecting resting orders during those times makes a huge difference to the trading quality on IEX.
Exchanges are the referees of the market. In that role, they have a choice: they can enable predatory trading, or they can stand up for the interests of natural investors.
The IEX Signal joins the IEX Speed Bump as a central feature of our exchange design, sharpening our ability to protect investors, their brokers, and market makers on IEX. It’s a tool with multiple applications in our overall exchange design, starting with D-Peg and Primary Peg.
Moving forward, we’ll be expanding the ways we deploy the Signal, evolving the exchange to more comprehensively serve investors, and build a market that creates opportunity for a broader range of investing and trading strategies, not just those based on the relentless pursuit of speed at the expense of all else.
As we continue to evolve, we invite feedback from all market participants. Please reach out with any questions, and we look forward to speaking with you.
For more about the research behind the IEX Signal, please read our white paper, “The Evolution of the Crumbling Quote Signal.”
IEX is the Investors Exchange: a fair, simple, and transparent stock exchange dedicated to investor and issuer protection. Built on the belief that every investor is entitled to the same right to trade on equal terms on every single trade, IEX is on a mission to level the playing field by eliminating unfair advantages from the markets.
On September 2, 2016, IEX launched as America’s newest stock exchange, and regularly matches over 140 million shares daily with a notional value of nearly $6 billion. On February 21, 2017, IEX reached a record market share of 2.3% of total U.S. equity volume. IEX plans to begin listing publicly-traded companies in 2017.