What exactly is backtesting. Its explained as using historical data to test predictive results vs actual results. In an insurance context I would use the model say using Q2 input data maybe adjusted for run off and FNB say and get estimates. I would consider this my "actual data" at that point in time. I am confused by the term historical data as my "actual data" is my historical data. So retrospectively looking back, how does my data change for backtesting to work. How is back testing different to comparing my estimated input say at YE17 looking at estimated 2018 calendar year vs my actual 2018 calendar year results?
