The following was originally published on Trading Technologies Trade Talk blog. Make Surveillance Smarter and Eliminate Blind Spots With TT® Score By: Jay Biondo and Morgan Trinkaus, Product Managers, Surveillance There is nothing quite like the shot of panic you feel when you receive a letter from a regulator indicating that your trading activity, or trading activity that is under your purview, is being investigated for potential market abuse. These initial letters are usually just the prelude to a long process where the only certainty is that it will be expensive and time consuming regardless of the outcome. And now that prison time is on the table for certain malicious trading behaviors, the stakes have never been higher. In 2017, the CTFC tripled the amount of enforcement actions involving disruptive trading practices, levying tens of millions of dollars in fines and permanently barring traders from CFTC regulated markets. These investigations can seem unpredictable at best, random at worst, to both traders and compliance officers alike. Especially since much of the activity that is being questioned appears to be similar to strategies that have been running for years. The truth of the matter, though, is that major regulatory investigations are not exactly black swans. For example, spoofing behavior has been considered illegal since 2013, and spoofing cases have been some of the highest profile events in the financial news over the past five years. The number of major cases involving spoofing has grown steadily every year, and everyone in the capital markets has been sufficiently put on notice that regulators are ready, willing and able to bring enforcement actions for spoofing. This raises an important question: if spoofing investigations cannot be classified as unprecedented and unpredictable events, then why are traders and compliance officers still living in fear of an unexpected knock on the door from regulators? This raises an important question: if spoofing investigations cannot be classified as unprecedented and unpredictable events, then why are traders and compliance officers still living in fear of an unexpected knock on the door from regulators? The answer to this question is that the rule-based legacy surveillance systems that have dominated the financial industry for the past decade do not provide traders and compliance officers with the tool kit necessary to proactively identify spoofing activity. The core problem with detection of spoofing using rule-based systems is that they rely on simple “if-then” statements (i.e., parameters) that require pre-set minimum thresholds to generate alerts, and these parameters by their very definition are not flexible. For example, a rule-based system might be centered around large cancels after small fills on the opposite side of the market. It seems simple, but what constitutes a “large” cancel? What constitutes a “small” fill? In what time frame do these events have to occur? The answers are surprisingly complex. What constitutes a “large” cancel is completely dependent on what product is being traded and at what time of day. The amount of volume needed to spoof any given market is dynamic, so orders of one size that might be sufficient to affect market pressure to spoof at midday might not be enough to have the same impact near the close. Furthermore, since it is not practical to run and review a uniquely tuned surveillance for each individual product at different times of day, most rule-based surveillances are tuned to try and be a one-size-fits-all solution for all products at all times of day. This approach practically guarantees spoofing will go undetected in any product that falls outside of this “average product.” For example, if the cancel parameter is set at 50 contracts required to trigger an alert, then what about the spoofing in a more illiquid product where you only need to spoof with 40 or 30 or even 20 contracts? This activity will simply fly under the radar and go undetected because the inflexible nature of parameters has created a blind spot. What makes this situation even more precarious for traders and compliance officers is that the canned parameter-setting options that legacy surveillance providers offer to mitigate this flaw are usually modeled off of the spoofing described in the headline enforcement actions that contain activity that is typically more than five years old. This promotes a purely reactive approach to compliance. Again, traders and compliance officers find themselves confronted with a huge blind spot, with no ability to “see around corners” to the next pattern the regulators will focus on. As a result of these obvious limitations, even after spending large amounts of time and money to implement a rule-based legacy system, it is still very likely that spoofing activity will go undetected. And the worst part is that the longer the spoofing activity goes undetected, the more catastrophic the outcome of the investigation will likely end up being. No wonder traders and compliance officers in today’s capital markets have high anxiety about their potential regulatory exposure! Recognizing the dire need for a more efficient and effective approach to trade surveillance, Trading Technologies is now offering a fully hosted, machine-learning-powered compliance surveillance solution named TT Score®. TT Score is the financial industry’s first and only parameter-free trade surveillance solution that gets smarter over time as it learns from real instances of market abuse. Traders and compliance officers are no longer faced with the insoluble task of wrestling with parameters for countless hours. The machine-learning models that power TT Score have been trained with positively labeled regulatory case data by a team of data scientists and domain experts to recognize the mathematical vectors of manipulative trading activity. Human intuition is thereby removed from the equation, and TT Score’s superior approach to pattern detection is instead based on math and science. Furthermore, when exposed to new sets of positively labeled data, the machine-learning models will adapt and increase both the variety of patterns they can detect and the precision with which they can detect them. Unlike simple parameters, the machine-learning models provide traders and compliance officers with the ability to “see around corners” and identify more recent patterns of activity that are drawing the attention of regulators, not just the patterns from enforcement actions that are already five years old. Finally, one of the most exciting aspects of TT Score is how easy it will be for traders and compliance officers to access this unique and powerful technology. For traders who use the TT® platform, it will simply be another icon on your trading platform, and you can access your compliance results instantaneously to give yourself the piece of mind that you are not unknowingly violating any trading rules. For compliance officers, you don’t need to supervise traders using TT to have access to TT Score; it can be a stand-alone solution for you instead. The implementation process is also extraordinarily efficient because there is no time spent up front custom tuning hundreds of parameters. You simply feed drop-copy data through the TT platform, log on via a web browser and your results will be ready for your review. In our next blog post, we will take a deeper dive into TT Score’s machine-learning-based approach to trade surveillance and provide a glimpse at some of the state-of-the-art visual tools within TT Score that help users identify the critical “needles in the haystack” within big data sets.