Neutralize the Advantage
Make an exploit useless by removing its payoff instead of banning its tool
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 6
- Confidence
- 98%
Katsuyama rejects the broad claim that computerized or high-frequency trading is inherently bad. His design target is narrower: nobody should execute against another participant using advanced price information. IEX therefore buys fast market data but delays incoming orders through additional fiber. During that delay, IEX can receive a price update and refuse a trade at a stale price. The same rule applies regardless of participant identity, so high-frequency firms can still join if they accept a market that cannot be gamed in that way. The broader strategic rule is to avoid regulating technological evolution directly. Identify the advantage that creates the harmful payoff, redesign the system so the advantage no longer matters, and allow participants to compete under the neutralized conditions.
Origin
IEX applied this logic through its coiled-fiber speed bump, which created enough time for the venue to update prices before an informed order could trade against stale information.
Core principles
- 01Do not confuse a neutral technology with its harmful use
- 02Define the unfair outcome precisely
- 03Design the system so privileged speed cannot capture that outcome
- 04Use the same rules for every participant
- 05Judge the design by whether diverse participants willingly use it
How to run it
- 1
Separate technology from conduct
Identify which uses of the technology improve the system and which specific behavior causes harm. Avoid banning an entire category because of one exploit.
Pro tip Describe the harmful behavior without using a broad industry label.
Watch out Generic labels invite generic defenses and obscure the actual mechanism.
- 2
Locate the exploitable advantage
Determine what privileged participant knows or can do before others. Trace how that timing or information becomes a profitable action.
Pro tip Follow the complete sequence from new information to execution.
Watch out Focusing only on the final transaction can miss the earlier signal.
- 3
Define fair execution
State the condition that must hold for an interaction to proceed. At IEX, a trade should not execute at ten after the relevant market price has moved to nine.
Pro tip Use an observable condition rather than a statement of intent.
Watch out Fairness that cannot be tested cannot govern execution consistently.
- 4
Create a neutralizing window
Insert enough delay or another control for the system to update its state before the advantaged action completes. Make the exploit fail by design.
Pro tip Use the smallest intervention that reliably removes the payoff.
Watch out A delay shorter than the information gap preserves the exploit.
- 5
Apply one rule to everyone
Allow any participant to join under the same market mechanics. Exclude conduct through system design rather than excluding a class of users by name.
Pro tip Explain the unavailable advantages before onboarding participants.
Watch out Selective exceptions recreate privileged access.
- 6
Test economic redundancy
Check whether investing in ever-faster access still changes outcomes on the platform. The control succeeds when additional speed no longer buys the exploit.
Pro tip Monitor behavior and participation, not only technical latency.
Watch out An arms race may migrate to a different advantage if incentives are not watched.
In the wild
An informed trader sees a stock move from ten to nine and sends an order to sell at ten against an unaware buyer. The incoming order traverses coiled fiber while IEX receives the market update. IEX then refuses to let the stale-price trade occur, turning raw speed into an insufficient advantage.
→ The venue can price trades with current information without banning computerized traders.
Katsuyama says firms including Virtu chose to trade on IEX after the venue explained which advantages they would not receive. The company did not exclude them for being high-frequency traders; it required them to play within a market designed not to be gamed.
→ A diverse set of participants supported the venue under common constraints.
Common mistakes
Banning the category
Not everyone using computers harms the market. Treating all automated activity as equivalent rejects beneficial uses and weakens the diagnosis.
Trying to outlaw faster technology
Katsuyama argues that technological evolution cannot be regulated away. The stronger move is to make the speed investment irrelevant to the exploit.
Giving selected users exceptions
A fair market requires consistent ground rules. Preferential data or access would reintroduce the imbalance the design is meant to remove.
Is it for you?
Best for
It is best for platforms where an exploit depends on privileged timing, information, or access that can be neutralized through design.
Not ideal for
It is not ideal when the harmful action remains profitable after the proposed system control or when rules cannot be applied consistently.
From the transcript
“we never made this about shutting out high frequency traders per se. We just built a market we felt could not be gamed”
“You can't regulate evolution of technology, but you can render it useless.”
“We set the ground rules. They've decided to play.”
From the episode
IEX Group Brad Katsuyama: Masters of Business (Audio)
Brad Katsuyama