Sunday, September 04, 2016

Reading the Footprints of Large Market Participants

In the last post, we took a look at a measure of upticking and downticking across all U.S. stocks as a way of gathering insight as to whether buyers or sellers were dominating market activity.  By tracking the flows of buying and selling activity in real time, we can identify the dynamics driving the auction activity of the market and profit from shifts in supply and demand as they are occurring.  The ability to perceive and act upon flows in real time is essential to the microanalysis that is an essential part of success for active traders.
Notice that the U.S. TICK measure is constructed by tracking upticks and downticks across a broad universe of stocks.  How can we track buying and selling flows for individual instruments, whether they are futures contracts or individual stocks?

The seemingly obvious answer would be to investigate every transaction in the instrument and identify whether it's occurring on an uptick or downtick and then aggregate the information.  There are several problems with that approach, however.  First, when we have a composite tape for a stock traded on multiple exchanges, it is not entirely clear when a print occurring at the same minute and second truly preceded a print occurring at the exact same time.  Second, how do we deal with transactions that occur at the same price?  Do they count neither as upticks nor downticks, or do we categorize them based upon the most recent price change?  Third, how do we distinguish between situations in which smart execution algo passively sit on bids and offers to buy and sell at best prices?  In such a situation, price may not move, but the intentions of the market participants can be very different.  To the degree that smart algos dominate execution and mask the intentions of participants, simple uptick/downtick rules can be misleading.

A valuable way of tackling this problem has been offered by Easley, Lopez de Prado, and O'Hara in their paper "Discerning Information From Trade Data".  This bulk volume classification method takes small clusters of volume or transactions and categorizes the price behavior within those to ascertain the intentions of market participants.  This provides an efficient method for inferring buying and selling pressure without relying on the ambiguities of a composite tape or remaining blind to intentions when successive transactions occur at the same price.

Above, we can see a simple implementation of this method for the ES futures (blue line) during the 9/1/2016 trading session.  The Y-axis is constructed in standard deviation units, so that we can readily see when significant buying and selling activity (red line) is occurring during the day.  Recall the questions that we can answer with the U.S. TICK data.  Through the volume classification method, these questions can be addressed without needing to track transactions across all stocks.  The questions can also be answered for individual stocks and futures contracts.  Over time, we can see shifts in buying and selling flows and participate in market activity accordingly.  The edge lies in the ability to read the footprints of large market participants, even when they make efforts to disguise their intentions.

This work is a nice illustration of how quantitative approaches to markets can inform discretionary decision-making.  It is also an important illustration of the value of information that occurs within any one-minute bar.  Many traders fail to read markets properly because they want to use a telescope instead of a microscope.  Understanding the flows occurring here and now is far more relevant to short-term trading than predicting those flows on the basis of remote events.

Further Reading:  Improving Your Trading Toolkit