I am re-reading Michael Lewis' Flash Boys book, and I have a question about how a High Frequency Trader was able to front-run an order in a particular example mentioned in the book. On page 78, chapter Ronan's Problem, it describes how the president of a large hedge fund noticed that his trading costs had increased substantially. Upon a consultation with Brad and Ronan, where he learns about what high frequency trading is and how it is increasing his costs, he attempts to capture an example using his own personal brokerage account.

After Brad and Ronan had left his office, the president of this big hedge fund, who had never before thought of himself as prey, reconsidered the financial markets. He sat at his desk watching both his personal online brokerage account and his $1,800-a-month Bloomberg terminal. In his private brokerage account he set out to buy an exchange-traded fund (ETF) comprised of Chinese construction companies. Over several hours, he watched the price of the fund on his Bloomberg terminal. It was midnight in China, nothing was happening, and the ETF's price didn't budge. He then clicked the Buy button on his online brokerage account screen, and the price on the Bloomberg screen jumped.


"I hadn't even hit Execute," says the hedge fund president. "I hadn't done anything but put in a ticker symbol and a quantity to buy. And the market popped."

The last part of this exerpt from the page is what really puzzles me. Because the implication is, that a high frequency trading algorithm was able to front-run him, and upon seeing the order, lifted the ETF's price before it entered the market. However, the president did not click the Execute button, merely typed in a ticker symbol. So, how could this high frequency trading algorithm observe an order come in, if one wasn't even made?

  • $\begingroup$ His broker must have sold data about his actions, possibly bundled with data about his person. When you enter a ticker you perform a lookup in the broker's database, when preparing a buy you might send API calls that reflect your actions. $\endgroup$
    – hroptatyr
    Commented Jan 12, 2023 at 8:02
  • $\begingroup$ Surely though that must be a breach of data protection? User activity such as a read-request to an API when querying data must be a private action? $\endgroup$ Commented Jan 12, 2023 at 9:07
  • $\begingroup$ I agree when it's not anonymous data. But if it happened as described, I'd assume the stock is traded rarely, now someone puts in a large order, the broker sends stuff like total-value-of-currently-prepared-orders thinking this is sufficiently anonymous ... can't see how that's illegal. On the other hand, the situation seems anecdotal, it'd be more convincing if he tried that experiment several times at different times of day $\endgroup$
    – hroptatyr
    Commented Jan 12, 2023 at 12:06
  • $\begingroup$ 'total-value-of-currently-prepared-ordres', conceptually I'm thinking of this as what a user has input to the UI web form, where nothing has been entered to a DB. Or am I thinking of this incorrectly? But you're right, it could very well have been sheer chance that the price jumped just as he went to buy $\endgroup$ Commented Jan 12, 2023 at 12:08
  • $\begingroup$ Oh, you're right, I wasn't thinking of a web-UI, more of a trader desk desktop tool (like the Bloomie) $\endgroup$
    – hroptatyr
    Commented Jan 12, 2023 at 12:10

1 Answer 1


Michael Lewis is either being disingenuous here or exercising creative license with this story.

  1. Spreads are wider than the edge of absorbing wholesale customer flow. Even if we take this to a further extreme and say that the market maker has a view of every market order in flight before the order hits the market, lifting the spread each time before a market order lands is too expensive.
  2. Moreover in this case it's just a frontend search for a ticker. Even if someone did sell frontend data of a user entering in a particular ticker symbol, it doesn't make sense to lift the spread just because some random person on their personal brokerage account searched a ticker symbol. The conditional probability that someone would execute on a ticker search in the direction is low, and you'd still need to predict the direction of that execution and need to correctly predict if a user will have enough size in the order to make it worth you lifting a whole price level. This is a surefire way to lose money.
  3. PFOF is regulated. Retail brokerages don't simply sell this data to order anticipators in a way that will considerably worsen their users' execution, as it would if an order anticipator were truly able to read your mind and the lift the spread in your intended direction every single time. (Note: This will lose both you and the order anticipator money — to the order anticipator they will lose a lot of money since they are wiping a whole level, and you probably a little money as you're just filling a retail portfolio position.)
  4. This contravenes the definition of frontrunning. The market maker is not the broker.

We also know after the fact that IEX's biggest customers today are HFT firms like Citadel, SIG, and HRT. So much for protecting the retail guy.

  • $\begingroup$ Thanks for the thorough answer, I have accepted it as best. I thought IEX set up the extended cable loop to prevent front-running. So wouldn't those HFT firms be operating in a market making capacity, instead of front-running? $\endgroup$ Commented Jan 17, 2023 at 11:46
  • $\begingroup$ I don't think IEX publish stats on specific market makers/broker dealers they use, so officially qualifying "biggest" might be hard. IEX does have special order types that they work with outside partners (e.g Citadel) to fulfill (e.g. Midpoint Liquidity exchange.iex.io/products/retail-program), which maybe what the OP is referring to? But that is meant to help retail traders $\endgroup$
    – Mike
    Commented Jan 18, 2023 at 4:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.