Canadian Aleksandr Milrud was arrested in Florida this week on charges of orchestrating an international stock manipulation scheme. According to the FBI, Milrud was caught on tape at a New York restaurant last June bragging “that he controlled approximately 60 percent of all China-based traders presently engaged in layering.” Milrud was trying to convince a foreign broker to give him access to U.S. stock markets so he could keep his army of layerers at work. The broker happened to be wearing a wire for the FBI, hence the fairly rapid (for this sort of thing) charges.
In a parallel civil proceeding, the SEC provided additional details into Milrud’s trading, including tables of the actual order messages entered by Milrud in two instances. Milrud worked with multiple brokerage firms in an attempt to obscure the relationship between his “dirty” orders that pushed prices around and the “clean” orders that harvested the benefit of those price moves. The SEC’s tables show both the “clean” and “dirty” orders.
While the SEC’s tables do not contain any order routing information and use whole second time stamps, they were sufficient for us to manually assemble the following market views in Surveyor:
In both examples, Milrud first entered enough visible orders in his “dirty” account to cause the market to move away from those orders. Surveyor shows this effect in two places: the “local” column in the order book in the lower right corner lists all the visible orders entered by Milrud, and the lighter shade of green (sell) or blue (buy) graphically shows the portion of the overall visible order book occupied by Milrud’s local orders. Note that as the lighter blue/green increases (indicating more order pressure from Milrud), the NBBO spread (in grey) moves away from that pressure. This shows that Milrud’s “dirty” orders succeeded in moving market prices.
After the prices moved, Milrud capitalized on those moves by either buying at depressed prices or selling at elevated prices using his “clean” account. The SEC tables do not indicate whether the “clean” orders were visible or hidden, and Matt Levine at Bloomberg points out that if they were visible, Milrud’s orders on both sides of the market were of comparable size and would offset each other, raising questions as to whether any manipulation occurred. But Surveyor confirms that the “clean” orders were hidden. Note that at the “clean” price tiers ($75.25 and $22.84 in the top and bottom examples, respectively), the number of shares entered by Milrud far exceeded the number of shares quoted visibly in the actual market data. Thus, at the time of the “clean” executions (indicated by the vertical yellow stripe in each example), the order book was heavily tilted by Milrud’s “dirty” orders in the direction favoring his “clean” orders. This is textbook layering.
Because the “clean” and “dirty” accounts were at different brokerage firms, neither broker’s compliance desks had the benefit of all of the above data to help find these events. However, the broker handling the “dirty” orders did see a pattern of heavy order entry, price movement away from those orders, cancellations, and price reversion. This pattern can be used to flag potential events for further inquiry with other brokers who handled executions during the affected price intervals. This is a clumsy and time-consuming process, but it is unfortunately the best that can be done given the limitations of existing U.S. equities market structure. Solving this problem–providing a single point of access to all relevant order details–is one of the chief benefits of the Consolidated Audit Trail, but until CAT comes online, schemes like Milrud’s, when not caught on tape, will be very hard to detect. Perhaps this story will provide some urgency to speed that process along.
In the mean time, we’ve added these events into the free demo version of Surveyor. Sign up and explore further here.