Second Runner-Up

My initial impression of this competition, "Riskstorm: Datathon", was that it HAD to be a data analytics oriented hackathon. You could say that I tricked myself into joining thanks to the lack of due diligence, but I guess it worked out in the end.

I only found out that Riskstorm was actually a hackathon based on actuarial science when the problem statement started talking about insurance and risk assessments. We were provided with the EIA-923 dataset containing US Power Plant data, and were asked to identify vulnerabilities like volatility, shortages, seasonal spikes, and regional imbalance.

For the preliminary round, we focused on fuel-shortage risk. We found that natural gas was an extremely high-risk fuel, coal showed moderate risk, while petroleum and petroleum coke were generally lower risk. For the final round, we built on these findings by developing a pricing model for a power-generation insurance product.

Our final model had an annual risk-adjusted premium from $250,000 to $1.125 million based on shortage probability and expected severity. This pricing model was tested with a 10,000-run Monte Carlo claims simulation across all 981 power plants, where each plant's claim occurrence was generated as a Bernoulli-distributed event.

To me, it was a unique experience to say the least. The amount of niche terms that I learnt was insane, and my respect for actuarial scientists also grew tremendously. Quite an interesting subject, but not one for me.