EagleAi Use Case Series: Market Data Risk – Bad market data causes a broker to reject orders from a client incorrectly which leads to a financial loss

In electronic trading, quality of market data is very critical. There has been too many instances where trading firms’ algorithmic engines trading on bad market data has caused huge losses for the trading firm. There are compliance rules that require brokers to ensure the correctness of an order by applying pre-market limit checks. It is important that these checks are applied with correct market data. In this blog we are going to consider a realistic but a hypothetical scenario where bad market data could cause huge losses even when it prevents a legitimate trade from happening.

Scenario: Trading systems subscribe to real time market data which has got bad market data for delta, gamma and other greeks of an option for one of its symbols.

  • Delta is 20x the usual delta for an option with a similar strike and contract expiry month. The limit checker checks the “Delta Adjusted Notional Value” of a client’s option order and rejects the order as it computed a 20x a typical value of the delta adjusted notional value.
  • Because of the limit breach, client is not able to send any orders to the exchange through the broker. It takes a while for the support team to figure out what is happening while the client is restless to place the orders.
  • When knowing about the trading glitch, client is upset with the broker.
  • Broker makes the client whole by agreeing to pay more than $100,000 for the trading loss suffered by the client during the 30 minute period it took to resolve.

It is obvious that a very high quality market data is critical to any trading activity at an institutional scale. When algorithms make trading decisions, it is even more important to have high quality market data. Despite the efforts of the team that supports the market data infrastructure, erroneous data does slip in. EagleAi can be an effective engine to aid in flagging when bad market data creeps in the stream.

EagleAi Market Data Risk module can process streaming data as an input. As it processes this data, it would self-learn about the characteristics of each symbol such as the typical price increase, the typical bid-ask spread at different times of the day, typical volume, typical values for the greeks if it were an option symbol etc. When EagleAi detects a huge deviation from the symbol’s patterns such as the one in this scenario, it can set a flag on the market data stream so that algos can choose to ignore this market data. EagleAi Market Data risk uses many different AI (Artificial Intelligence) techniques to detect these anomalies to compute a proprietary score named “E-score”. False positives are significantly reduced by flagging it only when EagleAi score is above a high threshold. EagleAi Market Data risk module can prevent trading decisions based on bad market data to prevent financial losses to the Client and maximize profits.

EagleAi – detects abnormal market data at lightning speed.
The EagleAi Market Data Risk module is an ensemble of Anomaly Detection techniques to detect abnormal market data across thousands of symbols.
Book a demo today @ https://eagleai.com/tradewatch-market-data-risk/

EagleAi is a collection of enterprise scale AI engines, which will watch your business’s back and prevent major losses Trading Risk, Fraud Risk, Compliance Risk, position exposures (TradeWatch Position Risk), and Market Data Risk.

EagleAi Use Case Series: Trading Risk – A single order with incompatible instructions from a client caused $100k+ loss in revenues for a major broker

In Electronic trading in large institutions, a lot can happen in a matter of seconds. More than 100,000 orders can come from its high frequency clients every second. An outage faced by the broker dealer even for a few seconds would affect many of their clients and would leave them to divert the flow to other brokers. This is a huge revenue loss for the broker as the clients may go many days before gaining confidence to fully route their flow back again to them. Many a times, the root cause of the problem may not be at the broker’s end but at the client’s themselves. Let’s take a specific example and see how EagleAi can protect a broker dealer and prevent any major disruption to their execution services.

Here is a realistic but hypothetical scenario: A client (usually a high frequency hedge fund) casually inquires about various different trading venues offered by the broker on a call with broker’s sales staff. Sales staff is eager to pass information to the client to impress on their capabilities. Each venue (such as dark liquidity pool) defines various order parameters specific to the venue and is very important to be configured properly with adequate training before it is used.

  • In this case, let us consider a curious client, who, without informing or testing with the broker, starts sending orders to a particular venue with incorrect or incompatible instructions.
  • The broker handles the orders, does the usual regulatory and compliance checks. Since in low latency flows, every piece of logic takes a few hundred microseconds and delays the order hitting the exchange, there are only minimally required checks applied.
  • Everything checks out fine in these limited checks. But what is not known at this point is that this single order is going to be the cause of a major service disruption for itself as well as for many other clients of the broker.
  • It turns out this order sent by the client had a terrible consequence at the engine processing orders for the trading venue.
  • Not only orders from the curious client got impacted, but also of hundreds of other clients that got stuck at this engine.

An outage call is initiated – on one end, senior management in business is calling up clients to pacify them and IT management is trying to find the root cause. Each and every second feels like eons as delay of not having the system working has a significant impact on the client’s confidence and also the bottom line revenues. Eventually after an hour of back and forth it was found out a single client’s order was the culprit. While the client can be faulted for not following testing guidelines before sending an order to a new venue, the broker’s fault is that it did not have a robust engine. Designing a robust system is extremely difficult but designing defenses against any deviations from the norm is not impossible.

Enter EagleAi

EagleAi’s Trading Risk will have detected the presence of orders from a client to an unusual trading venue or with unusual trading instructions to a known trading venue. EagleAi ensures that it has high confidence in predicting multiple AI (Artificial Intelligence) algorithms that it uses, thus reducing the chance of “false positives” or raising an alert when there is no need. When EagleAi raises the alert usually within a few seconds of the orders appearing in the log (or in a database), it can be triggered to warn the support staff. EagleAi bigdata analytics is designed to give enough contextual information which in this case would be client id, anomalous trading venue, anomalous trading instructions, the list of orders that fall in this category etc. to reduce the time to triage. Within minutes, the support staff can stop just this flow until it can be resolved. This helps in nipping the problem in the bud and saving the loss of hundreds of thousands of dollars in revenue to the execution desk.

How it works?

EagleAi Trade Watch Trading Risk is not one AI (Artificial Intelligence) engine. It is an ensemble of AI engines that collaborates to produce very accurate results. Conceptually each EagleAi engine passively listens to the data flows, in this case orders from the client from the log files of the engine that receives the order. While it is listening, it is also building a profile of each and every client as a relation of flows at certain times of the trading day (aka “time series profile”). Some of the characteristics it learns about the flow in this case are

  1. what is the typical size of the order sent by the client
  2. what are the typical trading venues used by the client
  3. what are the typical instructions sent by the client to the trading venues,
  4. what is the typical rate of messages sent by each client to the engine
  5. what is the participation rate of each symbol through this engine

It generates a multipart “E-Score” or “EagleAi Score” for each and every client that represents how confident it is about the client – i.e whether it has gained enough confidence from its observations so far to predict, what is the anomaly of this client compared to its peers, what is the anomaly level for this client’s overall flows and what is the anomaly level of a specific transaction.

Since EagleAi is learning automatically from a log file, there is no need for time consuming integration. EagleAi can just be pointed to a log file repository or a database of historical data. It can learn from historical data over the weekend and be ready for business on Monday.

EagleAi Trading risk module detects and alerts in real-time using bigdata analytics with enough contextual information to fix Trading risks faster avoiding thousands of dollars losses than any traditional methods.

Wouldn’t wish you had an employee who can be as efficient as EagleAi. Hire EagleAi and make EagleAi part of your team. EagleAi will watch your business’s back and prevent major losses (TradeWatch Trading Risk), frauds (TradeWatch Fraud Risk), compliance issues (TradeWatch Compliance Risk), position exposures (TradeWatch Position Risk), and market risk (TradeWatch Market Data Risk)

EagleAi – Alerts risk managers on unusual trading flows 24×7.
The EagleAi™ TradeWatch Trading Risk module is an ensemble of AI engines that observes trading transactions and learns to identify outliers that matter for risk management.
Book a demo today @ https://eagleai.com/tradewatch-trading-risk/

Check out our complete product suite –  TradeWatch Compliance RiskTradeWatch MarketData RiskTradeWatch Position RiskEagleAi Anomaly DetectorEagleAi Trend Detector