Best Simulation Innovation: CompatibL
To meet the need for accurate risk measurement during the Covid-19 pandemic, CompatibL has developed a new innovative software feature – a market-based machine learning generator. Because historical data prior to the pandemic do not more accurately reflect the current level of risk, its ability to measure risk from shorter time series was beneficial during the pandemic.
Banks and asset managers cannot accurately calibrate traditional risk models for the short-term time series of the pandemic era. The time series for any given currency, stock or loan is too short to gain statistical confidence in the valuation. Although enough data is available across the range, combining them only provides a model for “middle currency” or “middle name”, which is not accurate enough.
Using unattended machine learning, CompatibL market generators strictly aggregate statistics and then generate data samples for each individual name. The use of machine learning algorithms reduces the inherent statistical uncertainty of a short time series by preserving and incorporating differences between names into the model. This ability was confirmed by comparison with data outside the sample – the gold standard of model verification.
CompatibL models, based on machine learning and driven by market generators, are the first in the industry to be based on groundbreaking research by quantum experts. Models can generate data for time horizons from one to 30 years that require risk models. This made it possible during the pandemic to accurately measure credit risk, limits, insurance reserves and the effectiveness of macro-strategies. Machine learning algorithms generate samples of market data when historical time series are of insufficient length without relying on any preconceived notions of data.
Unlike traditional “no model” methods, which rely on interpolation, model learning can recognize and use patterns in data. And the input data is constantly included in the model, which allows it to gradually evolve.
CompatibL’s machine-based learning solution has met the need for an accurate way to measure risk from short time series during the deployment of the Covid-19 crisis, while complying with regulatory requirements to get rid of subjective judgment.
The model is also able to obtain and use the full range of data in both crisis and routine modes for long-term risk forecasts. It is gradually evolving as the crisis subsides over time, so banks and asset managers do not need to switch to a different model when changing regimes.
The judges said:
- A market-based generator of machine learning, making optimal use of sometimes scarce data.
- An obvious problem, an interesting solution.
- Dependence on historical data before the pandemic has decreased.
- It’s really innovative. The ability to generate time series data adequate for risk management from rare and short current periods will be crucial given the declining usefulness of actual historical data as a result of events such as Negative Rates, Covid, Credit Crisis and so on.
Alexander Sokol, founder, CEO and head of Quant Research CompatibL, says:
“On behalf of CompatibL I am honored to accept Risk Markets Technology Award for Best Innovation in Modeling. Machine learning is a new transformational technology that will affect every aspect of the front, middle and back office. I am confident that by the end of this decade it will become the market standard in financial product valuation and risk management.
Last year, financial markets experienced unprecedented shocks. Managing new levels of risk required innovative solutions, and machine learning was able to cope with this challenge. The CompatibL team is proud to be at the forefront of integrating machine learning into trading and risk software. ”