Interview with Yuri Shpolanskiy, VP Quantitative Engineering, Itiviti
Yuri, you head up Itiviti’s quantitative team. What, in your opinion, is the key to effective volatility management?
The key to effective volatility management is accurate modeling – the ability to describe the market and volatility surface, how it should transform as the market moves, and whether to use any automatic fitting techniques. Fitting is a mathematical procedure to find a curve that follows the market volatilities closely.
For comprehensive volatility management, I recommend a combination of volatility models, such as the widely used Wing, CCS (Clamped Cubic Spline) and SVI (Stochastic Volatility Inspired). Each volatility model has different behavior and parameterization, giving the flexibility to cover a wide range of derivative contracts and markets trading today. Firms will often need to apply different models depending on the scenario, as a model may fit a certain contract and market very well, but may not be suitable for another contract and market. For example, we typically find that US commodity options traders prefer CCS as they need finer point-per-point control, while European equity traders often use Wing. And sometimes, it’s just personal preference.
Regarding the curve fitting, when the mathematical problem is linear, it is fairly straightforward. However, the typical problem is not linear, comes with several constraints and thus leads to a more complicated, non-linear equation. Non-linearity requires a sophisticated and stable algorithm, which solves non-linear equations by iterations. A robust numerical procedure for all volatility models is paramount.
How do you provide stable fitting as the market changes?
One way is of course to fit the curve to the market every time it moves, but constant re-calculations have a negative impact on performance. A more efficient method is to monitor a specified set of criteria. As long as the volatility curve remains within these criteria, there is no need to re-adjust the fitting.
How do you achieve accurate volatility management without compromising performance?
– First, we ensure that the fitting procedure itself is fast.
– Second, the fitting procedure must be stable, meaning that it should not switch to an alternate route to solve the problem. The SVI model, for instance, offers a number of possible routes, but changing routes also requires changing parameters.
– The most complicated problem is to select the right criteria for when to apply, and when not to apply, the parameters found by the fitting algorithm. We have received many requests from our customers for help in this area.
You are working on a solution that will allow firms to use historical data for auto-fitting analysis. Why is this good?
It is very cool to be able to go back and examine historical data, because it shows how the algorithm reacts to changes in real life. This is crucial to understand in order to work out the criteria for when to re-adjust the curve. There are hundreds of potentially interesting parameters, but to consider them all would produce a terribly inefficient workflow. The trick is to understand which parameters are the most important, and to implement an automated solution for applying these parameters.
How do Itiviti’s customers use the volatility management models?
Some use the models out-of-the-box, others take advantage of the ability to customize them. Many actually switch between models depending on the product. There are two main scenarios:
– They specify their own parameters (and hide our suggestions)
– They base their volatility management on our suggestions, which in turn are based on typical situations specified by our customers.
Their choice really depends on the situation. They examine plots, and if they see that the curve applies inappropriately for them, they make changes.
In your opinion, what sets Itiviti’s volatility management apart?
One important aspect, from our customers’ perspective, is the ability to implement custom volatility models. It is also easy to modify the user interface of our volatility manager applications to create a customized environment.
This level of flexibility also makes it easy for us to make improvements. The application is constantly improved; not once or twice a year, but every day. A team of mathematicians and software engineers introduces new features and mathematically improves the algorithms based on customer feedback, proprietary and academic research.
Our expressed goal is to be the market leader in volatility management. Our customers should be the judges of that claim, but I believe that with customizable models, fast and accurate fitting algorithms and low impact on performance, we’re well on our way.
Blog The fast pace of derivatives markets makes their participants a demanding group. Derivatives traders must be flexible to succeed, and also need functional capabilities in their technology such as speed, global access and powerful analytics. They confront these needs in an environment affected by new regulation, plus more demands for market access and lower costs. […] April 27, 2017
Blog In the ongoing pursuit of efficiency and cost awareness among market participants, trading technology is often at the center of attention. Still there is at least one key function in trading systems left largely untouched by productivity and streamlining measures, and that is testing. Manual testing of trading infrastructure solutions remains prevalent within today’s financial […] April 20, 2017
Blog The ability to respond rapidly to a new opportunity is a prerequisite for success in derivatives trading. Adding access to a new exchange or electronic execution venue, support for a new security, or compliance to the latest regulatory wrinkle is essential to making sure opportunities are realized and profits are made. As a result, agility […] April 11, 2017
Automation & Testing Whether you are deploying FIX or proprietary protocol components within trading systems, thorough testing is crucial to success. Not only do you need to test your systems for execution venue driven events but also for managing the client flow. Manual testing is a costly and error-prone process that can be significantly improved and accelerated through automated testing. Automated regression […] April 6, 2017