So, How Do You Measure the High-Frequency Trading Arms Race?

by | 09.14.2021 | Lime Insights

latency arbitrage

  • Recent BIS study estimates from their U.K. data that global market for latency arbitrage is worth about $5 billion annually
  • Different “finish line” message data may reveal more arbitrage information
  • Competitor groups may be smaller than you think
  • Latency arbitrage has a “meaningful” price affect

Just how big and impactful is latency arbitrage across global equity markets?

A recent study published by the Bank for International Settlements (BIS) tackled the question and estimated latency arbitrage profits from global equity markets is worth about $5 billion annually. The study, “Quantifying the high-frequency trading ‘arms race’” by Matteo Aquilina, Eric Budish and Peter O’Neill, also examined the market using a different message data set, zeroed in on how many firms competed in so-called “races” and looked at frequency, race durations, and price impact on all stocks in the FTSE 350 index in Fall 2016 on the London Stock Exchange over a nine-week period.

The July 2021 study estimated that volume-only global latency arbitrage profits, using a variety of sensitivity analyses, ranged from $2.3 billion per year to $8.4 billion for races ranging from 50 microseconds to three milliseconds. Applied to 2020 markets, the volume-only estimate was $6.5 billion, while a volume-and-volatility model showed a higher estimate of $7 billion. The study’s estimate for global equity markets ranged from $3.1 billion to $11.4 billion in 2020, based on benchmark index volumes from 15 different equity markets.

Measuring Message Activity

The authors broke some new ground in the way they used data. For the study, authors acquired all message activity on all stocks on the London Stock Exchange’s FTSE 350 index from August 17, 2015 to October 16, 2015, rather than standard order book data used in similar studies. (Data from 43 days was included with one day excluded due to corrupt data.) The authors used timestamps to the microsecond to measure activity sent to the exchange’s “outer wall,” which is their term for the intended location of the exchange’s computer system for measuring speed races. With these timestamps, the authors said they could directly measure the quantity and duration of races, number of participants, and diversity and concentrations of winners and losers. The authors said once a message reaches the outer wall it is out of the participant’s hands and in the exchanges, or essentially at the “finish line.”

The study showed races made up a sizable percentage of the daily trading volume: about 22% of FTSE 100 trading volume at that time. Races also occurred frequently and were very short. The average for a FTSE 100 symbol was 537 latency-arbitrage races per day, roughly one race every minute, per LSE symbol and lasted an average of 81 microseconds. When races were measured up to a 3-millisecond window they accounted for 44% of all FTSE 100 trading volume. Modeling showed winners beat the first loser by 5 to 10 microseconds.

Also, the authors used the data to determine just how much participants in the race earned or lost. In terms of value, the average race was worth about a half tick, or roughly ₤2. Races in the 90th percentile were worth 3 ticks or about ₤7.

The top six firms take about 80% of the liquidity in the races, while providing about 42% of the liquidity that gets taken in races. Market participants outside the top six firms take about 20% of the liquidity, while providing 58%. The top three firms won about 55% of races and the top 6 firms collectively won 82% and 87%.

The study also looked at price impacts. It suggested that latency arbitrage trading in races

accounts for 33% of the effective spread and 31% of all price impact. The authors further added that changes to market design, such as including batch auctions, could reduce the cost of liquidity for investors by 17%.


This study did not include other factors such as total cost of trading (including trading fees) or how makers and takers in a particular market may ultimately be the same firm impacting prices. And it is an admittedly small sample size from which to extrapolate global market conclusions.

“Whether the numbers in our study seem big or small may depend on the vantage point from which they are viewed,” the authors noted in the study.

They suggest more studies, especially research using data from U.S. equities markets, would be valuable, although the same message data used in the report isn’t always available. The authors also support research on other markets such as fixed income, ETF futures, and currencies for further information about the size, frequency and broader impact latency arbitrage is having on those markets.

The ongoing debate and discourse over the impact of high frequency trading in equity markets seem to never end. But the authors at BIS believe they have taken a step in finding some answers and paving the way for more of them.

Check out the original study here.

Backtesting, simulation, and production trading

Featured Video

Is Your Broker Reliable?

Is Your Broker Reliable?

If you’ve dealt with unreliable trading technology in the past year, you’re not alone. Both slow reporting and complete outages happened to several trading firms during high volatility days in 2021.

read more
What is Rule 15c3-5?

What is Rule 15c3-5?

Before an order is eligible to be sent for execution, it must pass through a broker’s pre-trade risk check to ensure the order is not violating any pre-determined thresholds. This rule is known as Rule 15c3-5, but what is this rule exactly? Bob Iaccino explores.

read more