OilX’s VP of Data Science Engineering, Minas Spyridonakos PhD, has been pioneering nowcasting techniques for oil data. Nowcasting has been successfully implemented as a core feature in the OilX Platform. In this article, he shares his insights on what nowcasting is and how OilX uses this technique to estimate the global oil supply-demand balances in near real-time.
New technologies have increased the availability of real-time data and market participants are developing novel ways to exploit this new valuable resource. Data is the new oil. In a highly topical paper (Three Quant Lessons from COVID-19), Marcos Lopez de Prado and Alex Lipton highlight three lessons that quantitative researchers could learn from COVID-19. Lesson #1 stands truly out in our opinion as it is highly pertinent to the realities of the oil markets.
Lesson #1: More Nowcasting, Less Forecasting
Forecasting models tend to use structured data to make longer term predictions. Forecasting focuses on statistical relationships between past observations and future outcomes. These relationships, however, do not always hold and are often subject to change - a truly moving target. Forecasting made a lot of sense in the past, when datasets indeed were limited and data points were scarce and infrequent.
In contrast, nowcasting models use unstructured datasets to make:
- Direct measurements: target variable is directly observed (e.g. remote sensing via satellites of oil inventories, digital twinning of the oil supply chain)
- Short range predictions:target variable is not directly observed (e.g. Apple Mobility Data to infer US gasoline demand)
Advantages relative to forecasts are evident. Direct measurements always hold true as they don’t rely on a statistical lead/lag relationship. And short range predictions are statistically more reliable than long range predictions. Or as Marcos Lopez de Prado and Alex Lipton put it Forecasting is the mathematical analogue of guessing. Do not forecast what you can nowcast.
With the increase in data availability in oil markets, due to remote sensing via satellites and sensors (e.g. SAR, optical, thermal, AIS), we think nowcasting can now also be successfully applied to oil data. Commodity markets, and oil markets in particular, have been lagging behind other financial markets in applying quantitative research and trading strategies. But now, new sources of data and advances in data analytics also create new possibilities. At OilX, we believe that nowcasting is part of this future. For this reason, we continuously develop and evolve novel nowcasting techniques and deploy them in the OilX Platform.
What is nowcasting?
Nowcasting is the prediction of the present, the very near future and the very recent past in economics and meteorology. The technique of nowcasting has been used in meteorology for a long-time. The term itself is a contraction of "now" and "forecasting" and refers to the utilisation of the readily available data sets to infer the current state of a variable. It is about predicting the present, the recent past and the near future. At OilX, we use this technique to estimate the global oil supply and demand in near real-time.
Why is it relevant for the oil market?
The crude market has been historically opaque, with measurements for the various components of the global oil supply chain normally reported with a lag of several months. As more data becomes available through remote sensing, nowcasting enables us to provide a more accurate and up-to-date view of the global oil flow.
How do we solve the problem at OilX?
At OilX we combine high technology sensor-based measurements(e.g. SAR and optical satellite imagery based stocks tracking, AIS-based vessel tracking) with our proprietary statistical and ML models to generate an accurate snapshot of the current situation of the global oil market.
Would you like to see how nowcasting works at OilX? Request a demo today.
