Eleven storm seasons in a row, her model's projected loss range has held. Not exactly — no forecast is exact — but within the band she gives her underwriting team every spring, close enough that the streak has become something of a quiet legend inside the firm.
She is the first to say this should make people nervous, not impressed. A model that's never been wrong hasn't necessarily gotten smarter; it may simply not have been tested by the kind of season that breaks it yet.
The discomfort of a perfect record
Every year she runs a version of the model deliberately stressed with a worse season than anything in the historical record, specifically so the underwriting team has a number for the year the streak ends. So far, real seasons have never come close to that stress case — which she says is exactly the problem.
Insurance modeling rewards being right until the one year it doesn't, and the payout for that single miss can erase a decade of accurate years. She reviews her own model every winter looking for the specific weakness that eleven good years have been quietly hiding.