banner icon

Replaying real-world market data feeds in test environments using mdPlay

Reduce future trading losses by retransmitting real-world market data feeds in test environments to precisely             re-experience trading events

During a day of extreme market volatility, the trading systems at a leading investment bank reacted unexpectedly, resulting in the loss of millions of dollars in a matter of minutes.  The bank needed to reproduce the scenario in their test environment to understand why and ensure it didn’t happen again.  However, recreating the exact conditions that had caused their pricing systems to malfunction proved impossible.  Keen to ensure they wouldn’t face similar losses in the future, they turned to Velocimetrics who delivered a service allowing the firm to capture and precisely retransmit the trading data from any given event in their test environment.  Therefore, enabling the bank to:

  • Easily comprehend why a particular situation happened
  • Accurately assess whether system changes have had a positive impact
  • And ultimately minimise the potential for future trading losses

 

The issue

On the day in question, the bank was trading in an extremely volatile market.  An issue had emerged on a particular trading venue that had a ripple effect on the market data being published by other venues.

Our client was using fully automated trading practices.  However, unbeknown to the firm a vital component within their trading system wasn’t coping well.  It had never been tested to deal with the relatively unique scenario unfolding in the market, which unfortunately resulted in unexpected behaviour, generating significant trading losses.

Keen to avoid the future repetition of such losses, the firm wanted to reproduce the scenario again in their test environment, so they could ensure that the fix they had since made had really resolved the problem.  However, even though their test environment included multiple instances of the components that made up their production environment, they couldn’t access the market data feeds necessary to accurately reproduce the situation.  To do so would involve, not only replaying the exchange feed that experienced the issue, but also replaying all other feeds suffering from the ripple effect that were being consumed at that point-in-time, so the exact conditions could be replicated.  This proved impossible!

Never again did the bank want to be in a situation where it couldn’t reproduce a trading event and as such be unable to accurately test that the changes they had made to their systems, would actually generate a more positive response should the situation reoccur.  The firm needed a solution that would enable them to always be able to access the trading data they would need for effective testing.

 

The Solution

Velocimetrics was able to offer the client its mdPlay service that facilitates:

  • High capacity data capture of an entire day’s market data, retained for 2 months within the system
  • The automatic discovery of which market data streams are being consumed and whether they are experiencing sequence gaps
  • The ability for captured data to be replayed in a test environment

Therefore, mdPlay would enable the firm to capture a copy of all of the feeds being consumed by their trading system on an on-going basis.  The client could then automatically replay all feeds received during the previous day and if they wanted to replay a particular segment of a trading day, they could request that this specific time frame is made available to them.

With this service, the firm could accurately reproduce any previously experienced trading scenario with nanosecond range precision by either replaying the full trading day’s data or a given segment as an exact replica in their test environment.  In doing so, capturing all of the data’s richness and intricacies to precisely recreate the same set of circumstances.

The client could also request the creation of specific test scenarios.  These could include requests for a replay file where the speed the data is replayed for different trading periods is manipulated or certain trading events are replayed back-to-back.

The replays are then made available to the client via a user-friendly library that they simply search to find the particular replay they are interested in, and then select “Play”!  Just like streaming media from your playlist at home.

 

The results

The client is now able to fully test any change or enhancement to the components comprising their production trading system against a rigorous set of scenarios.  They can ensure that a system fix really has had a positive improvement and that a change hasn’t inadvertently had unintended consequences elsewhere in the trading ecosystem. In doing so, the firm has been able to:

  • Create a more robust and resilient trading environment
  • Reduce the potential for systems to send erroneous orders in usual trading situations
  • Or otherwise function in a way that contributes to unintended financial loss

Ultimately, this has enabled the client to put in place much more effective systems and risk controls.

Furthermore, Velocimetrics was able to get the client up and running with this service extremely quickly, because the market data streams being consumed could be automatically detected.  This also meant that minimal resource effort was required on behalf of the client.

Interested in learning more about mdPlay

Click here to find out more

Find out More

Issue

  • During a day of extreme market volatility, the bank’s trading systems reacted unexpectedly, generating trading losses in the millions
  • They wanted to reproduce the scenario to check if the way they had fixed the component would avoid the same issue reocurring
  • However, they couldn’t access the exact market data feeds being received when the incident occurred that would enable them to do so
  • Never again did the firm want to be in a situation where they couldn’t accurately test the changes made to their systems

Solution

  • The bank subscribed to the mdPlay service
  • With mdPlay they could capture a copy of all market data feeds received
  • They could then replay the data in their test environment with nanosecond range precision
  • They could replay the previous day’s feed, a particular segment of the trading day or request the curation of specific test scenarios
  • Replay scenarios are available via a library so users just select the test they want to run and click “Play!”

Results

  • Any system change or enhancement can now be rigorously tested using real-world market data
  • The bank can ensure a system fix really has delivered the desired improvements
  • They have been able to improve the robustness and resiliency of their trading environment
  • They’ve reduced the potential for trading systems and algos to behave in an unexpected manner in the live environment